I came across a recent post by Chad Sakac (VP, VMware Alliance at EMC) discussing the issue of how vendors drive customer specifications down from broader goals to individual features or implementation sets (I'm sure VCE was not in mind at the time.) When it comes to vendors insist on framing the "client argument" in terms of specific features and proprietary approaches, I have to agree that Chad is spot on. Here's why:
First, it helps when vendors move beyond the "simple thinking" of infrastructure elements as a grid of point solutions and more of an "organic marriage of tools" - often with overlapping qualities. Some marriages begin with specific goals, some develop them along the way and others change course drastically and without much warning. The rigidness of point approaches rarely accommodates growth beyond the set of assumptions that created the it in the first place. Likewise, the "laser focus" on specific features detracts from the overall goal: the present and future value of the solution.
When I married my wife, we both knew we wanted kids. Some of our friends married and "never" wanted kids, only to discover a child on the way and subsequent fulfillment through raising them. Still, others saw a bright future strained with incompatibility and the inevitable divorce. Such is the way with marriages.
Second, it takes vision to solve complex problems. Our church (Church of the Highlands in Birmingham, Alabama) takes a very cautious position on the union between souls: requiring that each new couple seeking a marriage give it the due consideration and compatibility testing necessary to have a real chance at a successful outcome. A lot of "problems" we would encounter were identified before we were married, and when they finally popped-up we knew how to identify and deal with them properly.
Couples that see "counseling" as too obtrusive (or unnecessary) have other options. While the initial investment of money are often equivalent, the return on investment is not so certain. Uncovering incompatibilities "after the sale" provides for difficult and too often a doomed outcome (hence, 50% divorce rate.)
This same drama plays out in IT infrastructures where equally elaborate plans, goals and unexpected changes abound. You date (prospecting and trials), you marry (close) and are either fruitful (happy client), disappointed (unfulfilled promises) or divorce. Often, it's not the plan that failed but the failure to set/manage expectations and address problems that causes the split.
Our pastor could not promise that our marriage would last forever: our success is left to God and the two of us. But he did help us to make decisions that would give us a chance at a fruitful union. Likewise, no vendor can promise a flawless outcome (if they do, get a second opinion), but they can (and should) provide the necessary foundation for a successful marriage of the technology to the business problem.
Third, the value of good advice is not always obvious and never comes without risk. My wife and I were somewhat hesitant on counseling before marriage because we were "in love" and were happy to be blind to the "problems" we might face. Our church made it easy for us: no counseling, no marriage. Businesses can choose to plot a similar course for their clients with respect to their products (especially the complex ones): discuss the potential problems with the solution BEFORE the sale or there is no sale. Sometimes this takes a lot of guts - especially when the competition takes the route of oversimplification. Too often IT sales see identifying initial problems (with their own approach) as too high a risk and too great an obstacle to the sale.
Ultimately, when you give due consideration to the needs of the marriage, you have more options and are better equipped to handle the inevitable trials you will face. Whether it's an unexpected child on the way, or an unexpected up-tick in storage growth, having the tools in-hand to deal with the problem lessens its severity. The point is, being prepared is better than the assumption of perfection.
Finally, the focus has to be what YOUR SOLUTION can bring to the table: not how you think your competition will come-up short. In Chad's story, he's identified vendors disqualifying one another's solutions based on their (institutional) belief (or disbelief) in a particular feature or value proposition. That's all hollow marketing and puffery to me, and I agree completely with his conclusion: vendors need to concentrate on how their solution(s) provide present and future value to the customer and refrain from the "art" of narrowly framing their competitors.
Features don't solve problems: the people using them do. The presence (or absence) of a feature simply changes the approach (i.e. the fallacy of feature parity). As Chad said, it's the TOTALITY of the approach that derives value - and that goes way beyond individual features and products. It's clear to me that a lot of counseling takes place between Sakac's EMC team and their clients to reach those results. Great job, Chad, you've set a great example for your team!
Vision. Innovation. Results. In search of the elegant solution to complex business problems…
Showing posts with label Ethics and Technology. Show all posts
Showing posts with label Ethics and Technology. Show all posts
Friday, November 12, 2010
Wednesday, April 7, 2010
Quick Take: IBM Tops VMmark, Crushes Record with 4P Nehalem-EX
It was merely a matter of time before one of the new core-rich titans - the Intel's 8-core "Beckton" Nehalem-EX (Xeon 7500) or AMD's 12-core "Magny-Cours" (Opteron 6100) - was to make a name for itself on VMware's VMmark benchmark. Today, Intel draws first blood in the form of an 4-processor, 32-core, 64-thread, monster from IBM: the x3850 X5 running four Xeon X7560 (2.266GHz - 2.67GHz w/turbo, 130W TDP, each) and 384GB of DDR3-1066 low-power registered DIMMs. Weighing-in at 70.78@48 tiles, the 4P IBM System x3850 handily beats the next highest system - the 48-core DL785 G5 which set the record of 53.73@35 tiles back in August, 2009 - and bests it by over 30%.
At $3,800+ per socket for the tested Beckton chip, this is no real 2P alternative. In fact, a pair of Cisco UCS B250 M2 blades will get 52 tiles running for much less money. Looking at processor and memory configurations alone, this is a $67K+ enterprise server, resulting in a moderately-high $232/VM price point for the IBM x3850 X5.
SOLORI's Take: The most interesting aspect of the EX benchmark is its clock-adjusted scaling factor: between 70% and 91% versus a 2P/8-core Nehalem-EP reference (Cisco UCS, B200 M1, 25.06@17 tiles). The unpredictable nature of Intel's "turbo" feature - varying with thermal loads and per-core conditions - makes an exact clock-for-clock comparison difficult. However, if the scaling factor is 90%, the EX blows away our previous expectations about the platform's scalability. Where did we go wrong when we predicted a conservative 44@39 tiles? We're looking at three things: (1) a bad assumption about the effectiveness of "turbo" in the EP VMmark case (setting Ref_EP_Clock to 3.33 GHz), and (2) underestimating EX's scaling efficiency (assumed 70%), (3) assuming a 2.26GHz clock for EX.
Chosing our minimum QPI/HT3 scalability factor of 75%, the predicted performance was derived this way from HP Proliant BL490 G6 as a baseline:
Correcting for the as-tested clock/turbo numbers, and using AMD's 2P-to-4P VMmark scaling efficiency of 83%, and shifting to the new UCS baseline (with newer ESX version) the Nehalem-EX prediction factors to:
Clearly, this approach either overestimates the scaling efficiency or underestimates the "turbo" mode. IBM claims that a 2.93 GHz "turbo" setting is viable where Intel suggests 2.67 GHz is the maximum, so there is a source of potential bias. Looking at the tiles-per-core ratio of the VMmark result, the Nehalem-EX drops from 2.13 tiles per core on EP/2P platforms to 1.5 tiles per core on EX/4P platforms - about a 30% drop in per-core loading efficiency. That indicator matches well with our initial 75% scaling efficiency moving from 2P to 4P - something that AMD demonstrated with Istanbul last August. Given the high TDP of EX and IBM's 2.93 GHz "turbo" specification, it's possible that "turbo" is adding clock cycles (and power consumption) and compensating for a "lower" scaling efficiency than we've assumed. Looking at the same estimation with 2.93GHz "clock" and 71% efficiency (1.5/2.13), the numbers fall in line with VMmark:
This give us a good basis for evaluating 2P vs. 4P Nehalem systems: scaling factor of 71% and capable of pushing clock towards the 3GHz mark within its thermal envelope. Both of these conclusions fit typical 2P-to-4P norms and Intel's process history.
SOLORI's 2nd Take: So where does that leave AMD's newest 12-core chip? To date, no VMmark exists for AMD's Magny-Cours, and AMD chips tend not to do as well in VMmark as their Intel peers do to the benchmarks SMT-friendly loads. However, we can't resist using the same analysis against AMD/MC's 2.4GHz Opteron 6174SE (theoretical) using the 2P HP DL385 G6 as a baseline for core loading and the HP DL785 G6 for tile performance (best of the best cases):
That's nowhere near good enough to top the current 8P, 48-core Istanbul VMmark at 53.73@35 tiles, so we'll likely have to wait for faster 6100 parts to see any new AMD records. However, assuming AMD's proposition is still "value 4P" so about 200 VM's at under $18K/server gets you around $90/VM or less.
At $3,800+ per socket for the tested Beckton chip, this is no real 2P alternative. In fact, a pair of Cisco UCS B250 M2 blades will get 52 tiles running for much less money. Looking at processor and memory configurations alone, this is a $67K+ enterprise server, resulting in a moderately-high $232/VM price point for the IBM x3850 X5.
SOLORI's Take: The most interesting aspect of the EX benchmark is its clock-adjusted scaling factor: between 70% and 91% versus a 2P/8-core Nehalem-EP reference (Cisco UCS, B200 M1, 25.06@17 tiles). The unpredictable nature of Intel's "turbo" feature - varying with thermal loads and per-core conditions - makes an exact clock-for-clock comparison difficult. However, if the scaling factor is 90%, the EX blows away our previous expectations about the platform's scalability. Where did we go wrong when we predicted a conservative 44@39 tiles? We're looking at three things: (1) a bad assumption about the effectiveness of "turbo" in the EP VMmark case (setting Ref_EP_Clock to 3.33 GHz), and (2) underestimating EX's scaling efficiency (assumed 70%), (3) assuming a 2.26GHz clock for EX.
Chosing our minimum QPI/HT3 scalability factor of 75%, the predicted performance was derived this way from HP Proliant BL490 G6 as a baseline:
Est. Tiles = EP_Tiles_per_core( 2.13 ) * 32 cores * Scaling_Efficiency( 75% ) * EX_Clock( 2.26 ) / EP_Clock( 2.93 ) = 39 tiles
Est. Score = Est_Tiles( 40 ) * EP_Score_per_Tile( 1.43 ) * Est_EX_Clock( 2.26 ) / Ref_EP_Clock( 2.93 ) = 44.12
Est. Nehalem-EX VMmark -> 44.12@39 tiles
Correcting for the as-tested clock/turbo numbers, and using AMD's 2P-to-4P VMmark scaling efficiency of 83%, and shifting to the new UCS baseline (with newer ESX version) the Nehalem-EX prediction factors to:
Est. Tiles = EP_Tiles_per_core( 2.13 ) * 32 cores * Scaling_Efficiency( 83% ) * EX_Clock( 2.67 ) / EP_Clock( 2.93 ) = 51 tiles
Est. Score = Est_Tiles( 51 ) * EP_Score_per_Tile( 1.47 ) * Est_EX_Clock( 2.67 ) / Ref_EP_Clock( 2.93 ) = 68.32
Est. Nehalem-EX VMmark -> 68.3@51 tiles
Clearly, this approach either overestimates the scaling efficiency or underestimates the "turbo" mode. IBM claims that a 2.93 GHz "turbo" setting is viable where Intel suggests 2.67 GHz is the maximum, so there is a source of potential bias. Looking at the tiles-per-core ratio of the VMmark result, the Nehalem-EX drops from 2.13 tiles per core on EP/2P platforms to 1.5 tiles per core on EX/4P platforms - about a 30% drop in per-core loading efficiency. That indicator matches well with our initial 75% scaling efficiency moving from 2P to 4P - something that AMD demonstrated with Istanbul last August. Given the high TDP of EX and IBM's 2.93 GHz "turbo" specification, it's possible that "turbo" is adding clock cycles (and power consumption) and compensating for a "lower" scaling efficiency than we've assumed. Looking at the same estimation with 2.93GHz "clock" and 71% efficiency (1.5/2.13), the numbers fall in line with VMmark:
Est. Tiles = EP_Tiles_per_core( 2.13 ) * 32 cores * Scaling_Efficiency( 71% ) * EX_Clock( 2.93 ) / EP_Clock( 2.93 ) = 48 tiles
Est. Score = Est_Tiles( 48 ) * EP_Score_per_Tile( 1.47 ) * Est_EX_Clock( 2.93 ) / Ref_EP_Clock( 2.93 ) = 70.56
Est. Nehalem-EX VMmark -> 70.56@48 tiles
This give us a good basis for evaluating 2P vs. 4P Nehalem systems: scaling factor of 71% and capable of pushing clock towards the 3GHz mark within its thermal envelope. Both of these conclusions fit typical 2P-to-4P norms and Intel's process history.
SOLORI's 2nd Take: So where does that leave AMD's newest 12-core chip? To date, no VMmark exists for AMD's Magny-Cours, and AMD chips tend not to do as well in VMmark as their Intel peers do to the benchmarks SMT-friendly loads. However, we can't resist using the same analysis against AMD/MC's 2.4GHz Opteron 6174SE (theoretical) using the 2P HP DL385 G6 as a baseline for core loading and the HP DL785 G6 for tile performance (best of the best cases):
Est. Tiles = HP_Tiles_per_core( 0.92 ) * 48 cores * Scaling_Efficiency( 83% ) * MC_Clock( 2.3 ) / HP_Clock( 2.6 ) = 33 tiles
Est. Score = Est_Tiles( 34 ) * HP_Score_per_Tile( 1.54 ) * Est_MC_Clock( 2.3 ) / Ref_HP_Clock( 2.8 ) = 41.8
Est. 4P Magny-Cours VMmark -> 41.8@33 tiles
That's nowhere near good enough to top the current 8P, 48-core Istanbul VMmark at 53.73@35 tiles, so we'll likely have to wait for faster 6100 parts to see any new AMD records. However, assuming AMD's proposition is still "value 4P" so about 200 VM's at under $18K/server gets you around $90/VM or less.
Wednesday, January 13, 2010
UCS and VCE vBlock Type 1 Challenge Top VMmark
[caption id="attachment_1306" align="alignright" width="105" caption="VMmark "Tile" Loads "]
[/caption]
Last November we reported on the Fujitsu RX300 S5 rack server taking the top VMmark spot for 8-core systems. Yesterday (January 11, 2010) Cisco's UCS B200-M1 using VMware ESX 4.0 (build 164009) came within 0.5% of the top spot with a score of 25.06@17 tiles. While falling only slightly short of the mark set by the brute force RX300/DX80 combo, the UCS system did so with a very different solution, unsurprisingly similar to the vBlock Type 1 architecture described by Chad Sakacc in his blog post about the VMware, Cisco and EMC alliance.
Given that VMmark is a single node test harness, the difference between rack server and blade server architectures is a non-issue. However, more than just rack vs. blade is going on in this comparison. The Cisco UCS system is being fed by a pair of 10GE converged network adapters - used both for host network access and Fiber Channel bus access - and a monolithic storage array in the guise of a CLARiiON CX4-240 complete with a complement of 20, 73GB STEC SSD's - just to sweeten the pot.
[caption id="attachment_1385" align="aligncenter" width="450" caption="VMmark Network Configuration for the UCS B200-M1"]
[/caption]
While it is clear from past VMmark posts that the network speed (beyond 1Gbps) has little to do with the results, it is nice to see the confidence Cisco has in the CNA approach (Cisco UCS M71KR-Q) to go with the "eggs in a basket" solution. Given the storage demands on the CNA, the VMmark result should remove any doubt about the viability (performance) of high-capacity tandems (we'll leave the physical link security concerns for another day.)
However, where the "rubber meets the road" in this contest is storage I/O and this solution - in our opinion is just plain showing off. With just 41 disks to build from, the CX4-240 has been configured to deliver 37 LUNs - nearly one LUN per unit disk. Before any awards are given out for storage of the year, we need to consider that 36 of those LUNs are RAID0 - yielding a testing platform with no real-world analog (hence "showing off".)
[caption id="attachment_1388" align="aligncenter" width="449" caption="CLARiiON CX4-240 Storage Build-out for UCS B200-M1 VMmark"]
[/caption]
Given the ease at which RAID0 can be replaced by RAID1+0, it may be safe to assume that the same results could have been obtained by using 77 disks instead of 41 - at which point the CX4-240 would still be less than half the size of the top VMmark's 172-disk solution. The reason is clear: SSD's accelerate I/O loads incredibly well in architectures that support them. If anything, this "runner-up" proves that SSD adoption is on the verge of becoming mainstream.
But what does this test show about UCS? Firstly, it shows that Cisco's platform can compete with the best solutions out there on CPU and I/O performance (what's a half a percentage point across 102 virtual machines?) It's not really a surprise given that the UCS platform was designed to do just that - and within a neatly managed framework. Secondly, it shows that the choice of EMC as a partner was an excellent one. As Martin Glassborow commented on his Storagebod's Blog, EMC's involvement in VMware has energized the storage vendor to take bold and innovative steps towards Cloud Computing solutions that it might not have done otherwise (like the RAID0 SSD array). Thirdly and most importantly, it underscores the importance of predictable performance in a virtualization solution. Given the UCS/vBlock approach to systems organization, it can be very difficult not to draw solid parallels between the benchmarks and expectations for net new builds based on the criterion.
[/caption]Last November we reported on the Fujitsu RX300 S5 rack server taking the top VMmark spot for 8-core systems. Yesterday (January 11, 2010) Cisco's UCS B200-M1 using VMware ESX 4.0 (build 164009) came within 0.5% of the top spot with a score of 25.06@17 tiles. While falling only slightly short of the mark set by the brute force RX300/DX80 combo, the UCS system did so with a very different solution, unsurprisingly similar to the vBlock Type 1 architecture described by Chad Sakacc in his blog post about the VMware, Cisco and EMC alliance.
Given that VMmark is a single node test harness, the difference between rack server and blade server architectures is a non-issue. However, more than just rack vs. blade is going on in this comparison. The Cisco UCS system is being fed by a pair of 10GE converged network adapters - used both for host network access and Fiber Channel bus access - and a monolithic storage array in the guise of a CLARiiON CX4-240 complete with a complement of 20, 73GB STEC SSD's - just to sweeten the pot.
[caption id="attachment_1385" align="aligncenter" width="450" caption="VMmark Network Configuration for the UCS B200-M1"]
[/caption]While it is clear from past VMmark posts that the network speed (beyond 1Gbps) has little to do with the results, it is nice to see the confidence Cisco has in the CNA approach (Cisco UCS M71KR-Q) to go with the "eggs in a basket" solution. Given the storage demands on the CNA, the VMmark result should remove any doubt about the viability (performance) of high-capacity tandems (we'll leave the physical link security concerns for another day.)
However, where the "rubber meets the road" in this contest is storage I/O and this solution - in our opinion is just plain showing off. With just 41 disks to build from, the CX4-240 has been configured to deliver 37 LUNs - nearly one LUN per unit disk. Before any awards are given out for storage of the year, we need to consider that 36 of those LUNs are RAID0 - yielding a testing platform with no real-world analog (hence "showing off".)
[caption id="attachment_1388" align="aligncenter" width="449" caption="CLARiiON CX4-240 Storage Build-out for UCS B200-M1 VMmark"]
[/caption]Given the ease at which RAID0 can be replaced by RAID1+0, it may be safe to assume that the same results could have been obtained by using 77 disks instead of 41 - at which point the CX4-240 would still be less than half the size of the top VMmark's 172-disk solution. The reason is clear: SSD's accelerate I/O loads incredibly well in architectures that support them. If anything, this "runner-up" proves that SSD adoption is on the verge of becoming mainstream.
But what does this test show about UCS? Firstly, it shows that Cisco's platform can compete with the best solutions out there on CPU and I/O performance (what's a half a percentage point across 102 virtual machines?) It's not really a surprise given that the UCS platform was designed to do just that - and within a neatly managed framework. Secondly, it shows that the choice of EMC as a partner was an excellent one. As Martin Glassborow commented on his Storagebod's Blog, EMC's involvement in VMware has energized the storage vendor to take bold and innovative steps towards Cloud Computing solutions that it might not have done otherwise (like the RAID0 SSD array). Thirdly and most importantly, it underscores the importance of predictable performance in a virtualization solution. Given the UCS/vBlock approach to systems organization, it can be very difficult not to draw solid parallels between the benchmarks and expectations for net new builds based on the criterion.
Thursday, November 19, 2009
NEC Offers "Dunnington" Liposuction, Tops 64-Core VMmark
NEC’s venerable Express5800/A1160 is back at the top VMmark chart, this time establishing the brand-new 64-core category with a score of 48.23@32 tiles - surpassing its 48-core 3rd place posting by over 30%. NEC’s new 16-socket, 64-core, 256GB “Dunnington” X7460 Xeon-based score represents a big jump in performance over its predecessor with a per tile ratio of 1.507 - up 6% from the 48-core ratio of 1.419.
To put this into perspective, the highest VMmark achieved, to date, is the score of 53.73@35 tiles (tile ratio 1.535) from the 48-core HP DL785 G6 in August, 2009. If you are familiar with the "Dunnington" X7460, you know that it's a 6-core, 130W giant with 16MB L2 cache and a 1000's price just south of $3,000 per socket. So that raises the question: how does 6-cores X 16-sockets = 64? Well, it's not pro-rationing from the Obama administration's "IT fairness" czar. NEC chose to disable the 4th and 6th core of each socket to reduce the working cores from 96 to 64.
At $500/core, NEC's gambit may represent an expensive form of "core liposuction" but it was a necessary one to meet VMware's "logical processor per host" limitation of 64. That's right, currently VMware's vSphere places a limit on logical processors based on the following formula:
CPU_Sockets X Cores_Per_Socket X Threads_Per_Core =< 64
According to VMware, the other 32 cores would have been "ignored" by vSphere had they been enabled. Since "ignored" is a nebulous term (aka "undefined"), NEC did the "scientific" thing by disabling 32 cores and calling the system a 64-core server. The win here: a net 6% improvement in performance per tile over the 6-core configuration - ostensibly from the reduced core loading on the 16MB of L3 cache per socket and reduction in memory bus contention.
Moving forward to 2010, what does this mean for vSphere hardware configurations in the wake of 8-core, 16-thread Intel Nehalem-EX and 12-core, 12-thread AMD Magny-Cours processors? With a 4-socket Magny-Cours system limitation, we won't be seeing any VMmarks from the boys in green beyond 48-cores. Likewise, the boys in blue will be trapped by a VMware limitation (albeit, a somewhat arbitrary and artificial one) into a 4-socket, 64-thread (HT) configuration or an 8-socket, 64-core (HT-disabled) configuration for their Nehalem-EX platform - even if using the six-core variant of EX. Looks like VMware will need to lift the 64-LCPU embargo by Q2/2010 just to keep up.
Wednesday, October 7, 2009
Quick Take: Nehalem/Istanbul Comparison at AnandTech
Johan De Gelas and crew present an interesting comparison of Dunnington, Shanghai, Istanbul and Nehalem in a new post at AnandTech this week. In the test line-up are the "top bin" parts from Intel and AMD in 4-core and 6-core incarnations:
Most importantly for virtualization systems architects is how the vCPU scheduling affects "measured" performance. The telling piece comes from the difference in comparison results where vCPU scheduling is equalized:
[caption id="attachment_1276" align="aligncenter" width="450" caption="AnandTech's Quad Sockets v. Dual Sockets Comparison. Oct 6, 2009."]
[/caption]
When comparing the results, De Gelas hits on the I/O factor which chiefly separates VMmark from vAPUS:
This is yet another issue that VMware architects struggle with in complex deployments. The latency in "Dunnington" is a huge contributor to its downfall and why the Penryn architecture was a dead-end. Combined with 8 additional threads in the 2P form factor, Nehalem delivers twice the number of hardware execution contexts than Shanghai, resulting in significant efficiencies for Nehalem where small working data sets are involved.
When larger sets are used - as in vAPUS - the Istanbul's additional cores allows it to close the gap to within the clock speed difference of Nehalem (about 12%). In contrast to VMmark which implies a 3:2 advantage to Nehalem, the vAPUS results suggest a closer performance gap in more aggressive virtualization use cases.
SOLORI's Take: We differ with De Gelas on the reduction in vAPUS' data set to accommodate the "cheaper" memory build of the Nehalem system. While this offers some advantages in testing, it also diminishes one of Opteron's greatest strengths: access to cheap and abundant memory. Here we have the testing conundrum: fit the test around the competitors or the competitors around the test. The former approach presents a bias on the "pure performance" aspect of the competitors, while the latter is more typical of use-case testing.
We do not construe this issue as intentional bias on AnandTech's part, however it is another vector to consider in the evaluation of the results. De Gelas delivers a report worth reading in its entirety, and we view this as a primer to the issues that will define the first half of 2010.
- Intel Nehalem-EP Xeon, X5570 2.93GHz, 4-core, 8-thread
- Intel "Dunnington" Xeon, X7460, 2.66GHz, 6-core, 6-thread
- AMD "Shanghai" Opteron 2389/8389, 2.9GHz, 4-core, 4-thread
- AMD "Istanbul" Opteron 2435/8435, 2.6GHz, 6-core, 6-thread
Most importantly for virtualization systems architects is how the vCPU scheduling affects "measured" performance. The telling piece comes from the difference in comparison results where vCPU scheduling is equalized:
[caption id="attachment_1276" align="aligncenter" width="450" caption="AnandTech's Quad Sockets v. Dual Sockets Comparison. Oct 6, 2009."]
[/caption]When comparing the results, De Gelas hits on the I/O factor which chiefly separates VMmark from vAPUS:
The result is that VMmark with its huge number of VMs per server (up to 102 VMs!) places a lot of stress on the I/O systems. The reason for the Intel Xeon X5570's crushing VMmark results cannot be explained by the processor architecture alone. One possible explanation may be that the VMDq (multiple queues and offloading of the virtual switch to the hardware) implementation of the Intel NICs is better than the Broadcom NICs that are typically found in the AMD based servers.Johan De Gelas, AnandTech, Oct 2009
This is yet another issue that VMware architects struggle with in complex deployments. The latency in "Dunnington" is a huge contributor to its downfall and why the Penryn architecture was a dead-end. Combined with 8 additional threads in the 2P form factor, Nehalem delivers twice the number of hardware execution contexts than Shanghai, resulting in significant efficiencies for Nehalem where small working data sets are involved.
When larger sets are used - as in vAPUS - the Istanbul's additional cores allows it to close the gap to within the clock speed difference of Nehalem (about 12%). In contrast to VMmark which implies a 3:2 advantage to Nehalem, the vAPUS results suggest a closer performance gap in more aggressive virtualization use cases.
SOLORI's Take: We differ with De Gelas on the reduction in vAPUS' data set to accommodate the "cheaper" memory build of the Nehalem system. While this offers some advantages in testing, it also diminishes one of Opteron's greatest strengths: access to cheap and abundant memory. Here we have the testing conundrum: fit the test around the competitors or the competitors around the test. The former approach presents a bias on the "pure performance" aspect of the competitors, while the latter is more typical of use-case testing.
We do not construe this issue as intentional bias on AnandTech's part, however it is another vector to consider in the evaluation of the results. De Gelas delivers a report worth reading in its entirety, and we view this as a primer to the issues that will define the first half of 2010.
Friday, September 25, 2009
Quick Take: HP Blade Tops 8-core VMmark w/OC'd Memory
HP's ProLiant BL490c G6 server blade now tops the VMware VMmark table for 8-core systems - just squeaking past rack servers from Lenovo and Dell with a score of 24.54@17 tiles: a new 8-core record. The half-height blade was equipped with two, quad-core Intel Xeon X5570 (Nehalem-EP, 130W TDP) and 96GB ECC Registered DDR3-1333 (12x 8GB, 2-DIMM/channel) memory.
In our follow-up, we found that HP's on-line configuration tool does not allow for DDR3-1333 memory so we went to the street for a comparison. For starters, we examined the on-line price from HP with DDR3-1066 memory and the added QLogic QMH2462 Fiber Channel adapter ($750) and additional NC360m dual-port Gigabit Ethernet controller ($320) which came to a grand total of $28,280 for the blade (about $277/VM, not including Blade chassis or SAN storage).
Stripping memory from the build-out results in a $7,970 floor to the hardware, sans memory. Going to the street to find 8GB sticks with DDR3-1333 ratings and HP support yielded the Kingston KTH-PL313K3/24G kit (3x 8GB DIMMs) of which we would need three to complete the build-out. At $4,773 per kit, the completed system comes to $22,289 (about $218/VM, not including chassis or storage) which may do more to demonstrate Kingston's value in the market place rather than HP's penchant for "over-priced" memory.
Now, the interesting disclosure from HP's testing team is this:
[caption id="attachment_1203" align="aligncenter" width="450" caption="Notes from HP's VMmark submission."]
[/caption]
While this appears to boost memory performance significantly for HP's latest run (compared to the 24.24@17 tiles score back in May, 2009) it does so at the risk of running the Nehalem-EP memory controller out of specification - essentially, driving the controller beyond the rated load. It is hard for us to imagine that this specific configuration would be vendor supported if used in a problematic customer installation.
SOLORI's Take:Those of you following closely may be asking yourselves: "Why did HP choose to over-clock the memory controller in this run by pushing a 1066MHz, 2DPC limit to 1333MHz?" It would appear the answer is self-evident: the extra 6% was needed to put them over the Lenovo machine. This issue raises a new question about the VMmark validation process: "Should out of specification configurations be allowed in the general benchmark corpus?" It is our opinion that VMmark should represent off-the-shelf, fully-supported configurations only - not esoteric configuration tweaks and questionable over-clocking practices.
Should there be as "unlimited" category in the VMmark arena? Who knows? How many enterprises knowingly commit their mission critical data and processes to systems running over-clocked processors and over-driven memory controllers? No hands? That's what we thought... Congratulations anyway to HP for clawing their way to the top of the VMmark 8-core heap...
In our follow-up, we found that HP's on-line configuration tool does not allow for DDR3-1333 memory so we went to the street for a comparison. For starters, we examined the on-line price from HP with DDR3-1066 memory and the added QLogic QMH2462 Fiber Channel adapter ($750) and additional NC360m dual-port Gigabit Ethernet controller ($320) which came to a grand total of $28,280 for the blade (about $277/VM, not including Blade chassis or SAN storage).
Stripping memory from the build-out results in a $7,970 floor to the hardware, sans memory. Going to the street to find 8GB sticks with DDR3-1333 ratings and HP support yielded the Kingston KTH-PL313K3/24G kit (3x 8GB DIMMs) of which we would need three to complete the build-out. At $4,773 per kit, the completed system comes to $22,289 (about $218/VM, not including chassis or storage) which may do more to demonstrate Kingston's value in the market place rather than HP's penchant for "over-priced" memory.
Now, the interesting disclosure from HP's testing team is this:
[caption id="attachment_1203" align="aligncenter" width="450" caption="Notes from HP's VMmark submission."]
[/caption]While this appears to boost memory performance significantly for HP's latest run (compared to the 24.24@17 tiles score back in May, 2009) it does so at the risk of running the Nehalem-EP memory controller out of specification - essentially, driving the controller beyond the rated load. It is hard for us to imagine that this specific configuration would be vendor supported if used in a problematic customer installation.
SOLORI's Take:Those of you following closely may be asking yourselves: "Why did HP choose to over-clock the memory controller in this run by pushing a 1066MHz, 2DPC limit to 1333MHz?" It would appear the answer is self-evident: the extra 6% was needed to put them over the Lenovo machine. This issue raises a new question about the VMmark validation process: "Should out of specification configurations be allowed in the general benchmark corpus?" It is our opinion that VMmark should represent off-the-shelf, fully-supported configurations only - not esoteric configuration tweaks and questionable over-clocking practices.
Should there be as "unlimited" category in the VMmark arena? Who knows? How many enterprises knowingly commit their mission critical data and processes to systems running over-clocked processors and over-driven memory controllers? No hands? That's what we thought... Congratulations anyway to HP for clawing their way to the top of the VMmark 8-core heap...
Sunday, September 13, 2009
Quick Take: Magny-Cours Spotted, Pushed to 3GHz for wPrime
Andreas Galistel at NordicHardware posted an article showing a system running a pair of engineering samples of the Magny-Cours processor running at 3.0GHz. Undoubtedly these images were culled from a report "leaked" on XtremeSystems forums showing a "DINAR2" motherboard with SR5690 chipset - in single and dual processor installation - running Magny-Cours at the more typical pre-release speed of 1.7GHz.
We know that Magny-Cours is essentially a MCM of Istanbul delivered in the rectangular socket G34 package. One thing illuminating about the two posts is the reported "reduction" in L3 cache from 12MB (6MB x 2 in MCM) to 10MB (2 x 5MB in MCM). Where did the additional cache go? That 's easy: since a 2P Magny-Cours installation is essentially a 4P Istanbul configuration, these processors have the new HT Assist feature enabled - giving 1MB of cache from each chip in the MCM to HT Assist.
Another thing intriguing about the XtremeSystems post in particular is the reported wPrime 32M and 1024M completion times. Compared to the hyper-threading-enabled 2P Xeon W5590 (130W TDP) running wPrime 32M at 3.33GHz (3.6GHz turbo) in 3.950 seconds, the 2P 3.0GHz Magny-Cours completed wPrime 32M in an unofficial 3.539 seconds - about 10% quicker while running a 10% slower clock. From the myopic lens of this result, it would appear AMD's choice of "real cores" versus hyper-threading delivers its punch.
SOLORI's Take: As a "reality check" we can compared the reigning quad-socked, quad-core Opteron 8393 SE result in wPrime 32M and wPrime 1024M at 3.90 and 89.52 seconds, respectively. Adjusted for clock and core count versus its Shanghai cousin, the Magny-Cours engineering samples - at 3.54 and 75.77 seconds, respectively - turned-in times about 10% slower than our calculus predicted. While still "record breaking" for 2P systems, we expected the Magny-Cours/Istanbul cores to out-perform Shanghai clock-per-clock - even at this stage of the game.
Due to the multi-threaded nature of the wPrime benchmark, it is likely that the HT Assist feature - enabled in a 2P Magny-Cours system by default - is the cause of the discrepancy. By reducing the available L3 cache by 1MB per die - 4MB of L3 cache total - HT Assist actually could be creating a slow-down. However, there are several things to remember here:
Looking at the more "typical" engineering sample speed tests posted on the XtremeSystems' forum tracks with the 3.0GHz overclock results at a more "typical" clock speed of 2.6GHz for 2P Magny-Cours: 3.947 seconds and 79.625 seconds for wPrime 32M and 1024M, respectively. Even at that speed, the 24-core system is on par with the 2P Nehalem system clocked nearly a GHz faster. Oddly, Intel reports the W5590 as not supporting "turbo" or hyper-threading although it is clear that Intel's marketing is incorrect based on actual testing.
Assuming Magny-Cours improves slightly on its way to market, we already know how 24-core Istanbul stacks-up against 16-thread Nehalem in VMmark and what that means for Nehalem-EP. This partly explains the marketing shift as Intel tries to position Nehalep-EP as a destined for workstations instead of servers. Whether or not you consider this move a prelude to the ensuing Nehalem-EX v. Magny-Cours combat to come or an attempt to keep Intel's server chip power average down by eliminating the 130W+ parts from the "server" list, Intel and AMD will each attempt win the war before the first shot is fired. Either way, we see nothing that disrupts the price-performance and power-performance comparison models that dominate the server markets.
[Ed: The 10% difference is likely due to the fact that the author was unable to get "more than one core" clocked at 3.0GHz. Likewise, he was uncertain that all cores were reliably clocking at 2.6GHz for the longer wPrime tests. Again, this engineering sample was designed to run at 1.7GHz and was not likely "hand picked" to run at much higher clocks. He speculated that some form of dynamic core clocking linked to temperature was affecting clock stability - perhaps due to some AMD-P tweaks in Magny-Cours.]
We know that Magny-Cours is essentially a MCM of Istanbul delivered in the rectangular socket G34 package. One thing illuminating about the two posts is the reported "reduction" in L3 cache from 12MB (6MB x 2 in MCM) to 10MB (2 x 5MB in MCM). Where did the additional cache go? That 's easy: since a 2P Magny-Cours installation is essentially a 4P Istanbul configuration, these processors have the new HT Assist feature enabled - giving 1MB of cache from each chip in the MCM to HT Assist.
"wPrime uses a recursive call of Newton's method for estimating functions, with f(x)=x2-k, where k is the number we're sqrting, until Sgn(f(x)/f'(x)) does not equal that of the previous iteration, starting with an estimation of k/2. It then uses an iterative calling of the estimation method a set amount of times to increase the accuracy of the results. It then confirms that n(k)2=k to ensure the calculation was correct. It repeats this for all numbers from 1 to the requested maximum."
Another thing intriguing about the XtremeSystems post in particular is the reported wPrime 32M and 1024M completion times. Compared to the hyper-threading-enabled 2P Xeon W5590 (130W TDP) running wPrime 32M at 3.33GHz (3.6GHz turbo) in 3.950 seconds, the 2P 3.0GHz Magny-Cours completed wPrime 32M in an unofficial 3.539 seconds - about 10% quicker while running a 10% slower clock. From the myopic lens of this result, it would appear AMD's choice of "real cores" versus hyper-threading delivers its punch.
SOLORI's Take: As a "reality check" we can compared the reigning quad-socked, quad-core Opteron 8393 SE result in wPrime 32M and wPrime 1024M at 3.90 and 89.52 seconds, respectively. Adjusted for clock and core count versus its Shanghai cousin, the Magny-Cours engineering samples - at 3.54 and 75.77 seconds, respectively - turned-in times about 10% slower than our calculus predicted. While still "record breaking" for 2P systems, we expected the Magny-Cours/Istanbul cores to out-perform Shanghai clock-per-clock - even at this stage of the game.
Due to the multi-threaded nature of the wPrime benchmark, it is likely that the HT Assist feature - enabled in a 2P Magny-Cours system by default - is the cause of the discrepancy. By reducing the available L3 cache by 1MB per die - 4MB of L3 cache total - HT Assist actually could be creating a slow-down. However, there are several things to remember here:
- These are engineering samples qualified for 1.7GHz operation
- Speed enhancements were performed with tools not yet adapted to Magny-Cours
- The author indicated a lack of control over AMD's Cool 'n Quiet technology which could have made "as tested" core clocks somewhat lower than what CPUz reported (at least during the extended tests)
- It is speculated that AMD will release Magny-Cours at 2.2GHz (top bin) upon release, making the 2.6+ GHz results non-typical
- The BIOS and related dependencies are likely still being "baked"
Looking at the more "typical" engineering sample speed tests posted on the XtremeSystems' forum tracks with the 3.0GHz overclock results at a more "typical" clock speed of 2.6GHz for 2P Magny-Cours: 3.947 seconds and 79.625 seconds for wPrime 32M and 1024M, respectively. Even at that speed, the 24-core system is on par with the 2P Nehalem system clocked nearly a GHz faster. Oddly, Intel reports the W5590 as not supporting "turbo" or hyper-threading although it is clear that Intel's marketing is incorrect based on actual testing.
Assuming Magny-Cours improves slightly on its way to market, we already know how 24-core Istanbul stacks-up against 16-thread Nehalem in VMmark and what that means for Nehalem-EP. This partly explains the marketing shift as Intel tries to position Nehalep-EP as a destined for workstations instead of servers. Whether or not you consider this move a prelude to the ensuing Nehalem-EX v. Magny-Cours combat to come or an attempt to keep Intel's server chip power average down by eliminating the 130W+ parts from the "server" list, Intel and AMD will each attempt win the war before the first shot is fired. Either way, we see nothing that disrupts the price-performance and power-performance comparison models that dominate the server markets.
[Ed: The 10% difference is likely due to the fact that the author was unable to get "more than one core" clocked at 3.0GHz. Likewise, he was uncertain that all cores were reliably clocking at 2.6GHz for the longer wPrime tests. Again, this engineering sample was designed to run at 1.7GHz and was not likely "hand picked" to run at much higher clocks. He speculated that some form of dynamic core clocking linked to temperature was affecting clock stability - perhaps due to some AMD-P tweaks in Magny-Cours.]
Wednesday, September 9, 2009
Quick Take: Dell/Nehalem Take #2, 2P VMmark Spot
The new 1st runner-up spot for VMmark in the “8 core” category was taken yesterday by Dell's R710 - just edging-out the previous second spot HP ProLiant BL490 G6 by 0.1% - a virtual dead heat. Equipped with a pair of Xeon X5570 ($1386/ea, bulk list) and 96GB registered DDR3/1066 (12x8GB), the 2U, rack mount R710 weighs-in with a tile ratio of 1.43 over 102 VMs. :
- Dell R710 w/redundant high-output power supply, ($18,209)
- 2 x Intel Xeon X5570 Processors (included)
- 96GB ECC DDR3/1066 (12×8GB) (included)
- 2 x Broadcom NexXtreme II 5709 dual-port GigabitEthernet w/TOE (included)
- 1 x Intel PRO 1000VT quad-port GigabitEthernet (1x PCIe-x4 slot, $529)
- 3 x QLogic QLE2462 FC HBA (1x PCIe slot, $1,219/ea)
- 1 x LSI1078 SAS Controller (on-board)
- 8 x 15K SAS OS drive, RAID10 (included)
- Required ProSupport package ($2,164)
- Total as Configured: $24,559 ($241/VM, not including storage)
Three Dell/EMC CX3-40f arrays were used as the storage backing of the test. The storage system included 8GB cache, 2 enclosures and 15, 15K disks per array delivering 19 LUNs at about 300GB each. Intel's Hyper-Threading and "Turbo Boost" were enabled for 8-thread, 3.33GHz core clocking as was VT; however embedded SATA and USB were disabled as is common practice.
At about $1,445/tile ($241/VM) the new “second dog” delivers its best at a 20% price premium over Lenovo's "top dog" - although the non-standard OS drive configuration makes-up a half of the difference, with Dell's mandatory support package making-up the remainder. Using a simple RAID1 SAS and eliminating the support package would have droped the cost to $20,421 - a dead heat with Lenovo at $182/VM.
Comparing the Dell R710 the 2P, 12-core benchmark HP DL385 G6 Istanbul system at 15.54@11 tiles:
- HP DL385 G6 ($5,840)
- 2 x AMD 2435 Istanbul Processors (included)
- 64GB ECC DDR2/667 (8×8GB) ($433/DIMM)
- 2 x Broadcom 5709 dual-port GigabitEthernet (on-board)
- 1 x Intel 82571EB dual-port GigabitEthernet (1x PCIe slot, $150/ea)
- 1 x QLogic QLE2462 FC HBA (1x PCIe slot, $1,219/ea)
- 1 x HP SAS Controller (on-board)
- 2 x SAS OS drive (included)
- $10,673/system total (versus $14,696 complete from HP)
Direct pricing shows Istanbul’s numbers at $1,336/tile ($223/VM) which is a 7.5% savings per-VM over the Dell R710. Going to the street - for memory only - changes the Istanbul picture to $970/tile ($162/VM) representing a 33% savings over the R710.
SOLORI's Take: Istanbul continues to offer a 20-30% CAPEX value proposition against Nehalem in the virtualization use case - even without IOMMU and higher memory bandwidth promised in upcoming Magny-Cours. With the HE parts running around $500 per processor, the OPEX benefits are there for Istanbul too. It is difficult to understand why HP wants to charge $900/DIMM for 8GB PC-5300 sticks when they are available on the street for 50% less - that's a 100% markup. Looking at what HP charges for 8GB DDR3/1066 - $1,700/DIM - they are at least consistent. HP's memory pricing practice makes one thing clear - customers are not buying large memory configurations from their system vendors...
On the contrary, Dell appears to be happy to offer decent prices on 8GB DDR3/1066 with their R710 at approximately $837/DIMM - almost par with street prices. Looking to see if this parity held up with Dell's AMD offerings, we examined the prices offered with Dell's R805: while - at $680/DIMM - Dell's prices were significantly better than HP's, they still exceeded the market by 50%. Still, we were able to configure a Dell R805 with AMD 2435's for much less than the equivalent HP system:
- Dell R805 w/redundant power ($7,214)
- 2 x AMD 2435 Istanbul Processors (included)
- 64GB ECC DDR2/667 (8×8GB) ($433/ea, street)
- 4 x Broadcom 5708 GigabitEthernet (on-board)
- 1 x Intel PRO 100oPT dual-port GigabitEthernet (1x PCIe slot, included)
- 1 x QLogic QLE2462 FC HBA (1x PCIe slot, included)
- 1 x Dell PERC SAS Controller (on-board)
- 2 x SAS OS drive (included)
- $10,678/system total (versus $12,702 complete from Dell)
This offering from Dell should be able to deliver equivalent performance with HP's DL385 G6 and likewise savings/VM compared to the Nehalem-based R710. Even at the $12,702 price as delivered from Dell, the R805 represents a potential $192/VM price point - about $50/VM (25%) savings over the R710.
Wednesday, August 26, 2009
Quick Take: HP's Sets Another 48-core VMmark Milestone
Not satisfied with a landmark VMmark score that crossed the 30 tile mark for the first time, HP's performance team went back to the benches two weeks later and took another swing at the performance crown. Well, the effort paid off, and HP significantly out-paced their two-week-old record with a score of 53.73@35 tiles in the heavy weight, 48-core category.
Using the same 8-processor HP ProLiant DL785 G6 platform as in the previous run - complete with 2.8GHz AMD Opteron 8439 SE 6-core chips and 256GB DDR2/667 - the new score comes with significant performance bumps in the javaserver, mailserver and database results achieved by the same system configuration as the previous attempt - including the same ESX 4.0 version (164009). So what changed to add an additional 5 tiles to the team's run? It would appear that someone was unsatisfied with the storage configuration on the mailserver run.
Given that the tile ratio of the previous run ran about 6% higher than its 24-core counterpart, there may have been a small indication that untapped capacity was available. According to the run notes, the only reported changes to the test configuration - aside from the addition of the 5 LUNs and 5 clients needed to support the 5 additional tiles - was a notation indicating that the "data drive and backup drive for all mailserver VMs" we repartitioned using AutoPart v1.6.
The change in performance numbers effectively reduces the virtualization cost of the system by 15% to about $257/VM - closing-in on its 24-core sibling to within $10/VM and stretching-out its lead over "Dunnington" rivals to about $85/VM. While virtualization is not the primary application for 8P systems, this demonstrates that 48-core virtualization is definitely viable.
SOLORI's Take: HP's performance team has done a great job tuning its flagship AMD platform, demonstrating that platform performance is not just related to hertz or core-count but requires balanced tuning and performance all around. This improvement in system tuning demonstrates an 18% increase in incremental scalability - approaching within 3% of the 12-core to 24-core scaling factor, making it actually a viable consideration in the virtualization use case.
In recent discussions with AMD about the SR5690 chipset applications for Socket-F, AMD re-iterated that the mainstream focus for SR5690 has been Magny-Cours and the Q1/2010 launch. Given the close relationship between Istanbul and Magny-Cours - detailed nicely by Charlie Demerjian at Semi-Accurate - the bar is clearly fixed for 2P and 4P virtualization systems designed around these chips. Extrapolating from the similarities and improvements to I/O and memory bandwidth, we expect to see 2P VMmarks besting 32@23 and 4P scores over 54@39 from HP, AMD and Magny-Cours.
SOLORI's 2nd Take: Intel has been plugging away with its Nehalem-EX for 8-way systems and - delivering 128-threads - promises to deliver some insane VMmarks. Assuming Intel's EX scales as efficiently as AMD's new Opterons have, extrapolations indicate performance for the 4P, 64-thread Nehalem-EX shoud fall between 41@29 and 44@31 given the current crop of speed and performance bins. Using the same methods, our calculus predicts an 8P, 128-thread EX system should deliver scores between 64@45 and 74@52.
With EX expected to clock at 2.66GHz with 140W TDP and AMD's MCM-based Magny-Cours doing well to hit 130W ACP in the same speed bins, CIO's balancing power and performance considerations will need to break-out the spreadsheets to determine the winners here. With both systems running 4-channel DDR3, there will be no power or price advantage given on either side to memory differences: relative price-performance and power consumption of the CPU's will be major factors. Assuming our extrapolations are correct, we're looking at a slight edge to AMD in performance-per-watt in the 2P segment, and a significant advantage in the 4P segment.
Using the same 8-processor HP ProLiant DL785 G6 platform as in the previous run - complete with 2.8GHz AMD Opteron 8439 SE 6-core chips and 256GB DDR2/667 - the new score comes with significant performance bumps in the javaserver, mailserver and database results achieved by the same system configuration as the previous attempt - including the same ESX 4.0 version (164009). So what changed to add an additional 5 tiles to the team's run? It would appear that someone was unsatisfied with the storage configuration on the mailserver run.
Given that the tile ratio of the previous run ran about 6% higher than its 24-core counterpart, there may have been a small indication that untapped capacity was available. According to the run notes, the only reported changes to the test configuration - aside from the addition of the 5 LUNs and 5 clients needed to support the 5 additional tiles - was a notation indicating that the "data drive and backup drive for all mailserver VMs" we repartitioned using AutoPart v1.6.
The change in performance numbers effectively reduces the virtualization cost of the system by 15% to about $257/VM - closing-in on its 24-core sibling to within $10/VM and stretching-out its lead over "Dunnington" rivals to about $85/VM. While virtualization is not the primary application for 8P systems, this demonstrates that 48-core virtualization is definitely viable.
SOLORI's Take: HP's performance team has done a great job tuning its flagship AMD platform, demonstrating that platform performance is not just related to hertz or core-count but requires balanced tuning and performance all around. This improvement in system tuning demonstrates an 18% increase in incremental scalability - approaching within 3% of the 12-core to 24-core scaling factor, making it actually a viable consideration in the virtualization use case.
In recent discussions with AMD about the SR5690 chipset applications for Socket-F, AMD re-iterated that the mainstream focus for SR5690 has been Magny-Cours and the Q1/2010 launch. Given the close relationship between Istanbul and Magny-Cours - detailed nicely by Charlie Demerjian at Semi-Accurate - the bar is clearly fixed for 2P and 4P virtualization systems designed around these chips. Extrapolating from the similarities and improvements to I/O and memory bandwidth, we expect to see 2P VMmarks besting 32@23 and 4P scores over 54@39 from HP, AMD and Magny-Cours.
SOLORI's 2nd Take: Intel has been plugging away with its Nehalem-EX for 8-way systems and - delivering 128-threads - promises to deliver some insane VMmarks. Assuming Intel's EX scales as efficiently as AMD's new Opterons have, extrapolations indicate performance for the 4P, 64-thread Nehalem-EX shoud fall between 41@29 and 44@31 given the current crop of speed and performance bins. Using the same methods, our calculus predicts an 8P, 128-thread EX system should deliver scores between 64@45 and 74@52.
With EX expected to clock at 2.66GHz with 140W TDP and AMD's MCM-based Magny-Cours doing well to hit 130W ACP in the same speed bins, CIO's balancing power and performance considerations will need to break-out the spreadsheets to determine the winners here. With both systems running 4-channel DDR3, there will be no power or price advantage given on either side to memory differences: relative price-performance and power consumption of the CPU's will be major factors. Assuming our extrapolations are correct, we're looking at a slight edge to AMD in performance-per-watt in the 2P segment, and a significant advantage in the 4P segment.
Wednesday, August 12, 2009
Quick Take: HP Plants the Flag with 48-core VMmark Milestones
Following on the heels of last month we predicted that HP could easily claim the VMmark summit with its DL785 G6 using AMD's Istanbul processors:
Well, HP didn't make us wait too long. Today, the PC maker cleared two significant VMmark milestones: crossing the 30 tile barrier in a single system (180 VMs) and exceeding the 40 mark on VMmark score. With a score of 47.77@30 tiles, the HP DL785 G6 - powered by 8 AMD Istanbul 8439 SE processors and 256GB of DDR2/667 memory - set the bar well beyond the competition and does so with better performance than we expected - most likely due to AMD's "HT assist" technology increasing its scalability.
Not available until September 14, 2009, the HP DL785 G6 is a pricey competitor. We estimate - based on today's processor and memory prices - that a system as well appointed as the VMmark-configured version (additional NICs, HBA, etc) will run at least $54,000 or around $300/VM (about $60/VM higher than the 24-core contender and about $35/VM lower than HP's Dunnnigton "equivalent").
SOLORI's Take: While the September timing of the release might imply a G6 with AMD's SR5690 and IOMMU, we're doubtful that the timing is anything but a coincidence: even though such a pairing would enable PCIe 2.0 and highly effective 10Gbps solutions. The modular design of the DL785 series - with its ability to scale from 4P to 8P in the same system - mitigates the economic realities of the dwindling 8P segment, and HP has delivered the pinnacle of performance for this technology.
We are also impressed with HP's performance team and their ability to scale Shanghai to Istanbul with relative efficiency. Moving from DL785 G5 quad-core to DL785 G6 six-core was an almost perfect linear increase in capacity (95% of theoretical increase from 32-core to 48-core) while performance-per-tile increased by 6%. This further demonstrates the "home run" AMD has hit with Istanbul and underscores the excellent value proposition of Socket-F systems over the last several years.
Unfortunately, while they demonstrate a 91% scaling efficiency from 12-core to 24-core, HP and Istanbul have only achieved a 75% incremental scaling efficiency from 24-cores to 48-cores. When looking at tile-per-core scaling using the 8-core, 2P system as a baseline (1:1 tile-to-core ratio), 2P, 4P and 8P Istanbul deliver 91%, 83% and 62.5% efficiencies overall, respectively. However, compared to the %58 and 50% tile-to-core efficiencies of Dunnington 4P and 8P, respectively, Istanbul clearly dominates the 4P and 8P performance and price-performance landscape in 2009.
In today's age of virtualization-driven scale-out, SOLORI's calculus indicates that multi-socket solutions that deliver a tile-to-core ratio of less than 75% will not succeed (economically) in the virtualization use case in 2010, regardless of socket count. That said - even at a 2:3 tile-to-core ratio - the 8P, 48-core Istanbul will likely reign supreme as the VMmark heavy-weight champion of 2009.
SOLORI's 2nd Take: HP and AMD's achievements with this Istanbul system should be recognized before we usher-in the next wave of technology like Magny-Cours and Socket G34. While the DL785 G6 is not a game changer, its footnote in computing history may well be as a preview of what we can expect to see out of Magny-Cours in 2H/2010. If 12-core, 4P system price shrinks with the socket count we could be looking at a $150/VM price-point for a 4P system: now that would be a serious game changer.
If AMD’s Istanbul scales to 8-socket at least as efficiently as Dunnington, we should be seeing some 48-core results in the 43.8@30 tile range in the next month or so from HP’s 785 G6 with 8-AMD 8439 SE processors. You might ask: what virtualization applications scale to 48-cores when $/VM is doubled at the same time? We don’t have that answer, and judging by Intel and AMD’s scale-by-hub designs coming in 2010, that market will need to be created at the OEM level.
Well, HP didn't make us wait too long. Today, the PC maker cleared two significant VMmark milestones: crossing the 30 tile barrier in a single system (180 VMs) and exceeding the 40 mark on VMmark score. With a score of 47.77@30 tiles, the HP DL785 G6 - powered by 8 AMD Istanbul 8439 SE processors and 256GB of DDR2/667 memory - set the bar well beyond the competition and does so with better performance than we expected - most likely due to AMD's "HT assist" technology increasing its scalability.
Not available until September 14, 2009, the HP DL785 G6 is a pricey competitor. We estimate - based on today's processor and memory prices - that a system as well appointed as the VMmark-configured version (additional NICs, HBA, etc) will run at least $54,000 or around $300/VM (about $60/VM higher than the 24-core contender and about $35/VM lower than HP's Dunnnigton "equivalent").
SOLORI's Take: While the September timing of the release might imply a G6 with AMD's SR5690 and IOMMU, we're doubtful that the timing is anything but a coincidence: even though such a pairing would enable PCIe 2.0 and highly effective 10Gbps solutions. The modular design of the DL785 series - with its ability to scale from 4P to 8P in the same system - mitigates the economic realities of the dwindling 8P segment, and HP has delivered the pinnacle of performance for this technology.
We are also impressed with HP's performance team and their ability to scale Shanghai to Istanbul with relative efficiency. Moving from DL785 G5 quad-core to DL785 G6 six-core was an almost perfect linear increase in capacity (95% of theoretical increase from 32-core to 48-core) while performance-per-tile increased by 6%. This further demonstrates the "home run" AMD has hit with Istanbul and underscores the excellent value proposition of Socket-F systems over the last several years.
Unfortunately, while they demonstrate a 91% scaling efficiency from 12-core to 24-core, HP and Istanbul have only achieved a 75% incremental scaling efficiency from 24-cores to 48-cores. When looking at tile-per-core scaling using the 8-core, 2P system as a baseline (1:1 tile-to-core ratio), 2P, 4P and 8P Istanbul deliver 91%, 83% and 62.5% efficiencies overall, respectively. However, compared to the %58 and 50% tile-to-core efficiencies of Dunnington 4P and 8P, respectively, Istanbul clearly dominates the 4P and 8P performance and price-performance landscape in 2009.
In today's age of virtualization-driven scale-out, SOLORI's calculus indicates that multi-socket solutions that deliver a tile-to-core ratio of less than 75% will not succeed (economically) in the virtualization use case in 2010, regardless of socket count. That said - even at a 2:3 tile-to-core ratio - the 8P, 48-core Istanbul will likely reign supreme as the VMmark heavy-weight champion of 2009.
SOLORI's 2nd Take: HP and AMD's achievements with this Istanbul system should be recognized before we usher-in the next wave of technology like Magny-Cours and Socket G34. While the DL785 G6 is not a game changer, its footnote in computing history may well be as a preview of what we can expect to see out of Magny-Cours in 2H/2010. If 12-core, 4P system price shrinks with the socket count we could be looking at a $150/VM price-point for a 4P system: now that would be a serious game changer.
Wednesday, July 29, 2009
NEC Adds Top 48-Core, Dell Challenges 24-Core in VMmark Race
NEC's venerable Express5800/A1160 tops the 48-core VMmark category today with a score of 34.05@24 tiles to wrest the title away from IBM who established the category back in June, 2009. NEC's new "Dunnington" X7460 Xeon-based score represents a performance per tile ratio of 1.41 and a tile to core efficiency of 50% using 128GB of ECC DDR2 RAM.
Compared to the leading 24-core "Dunnington" results - held by IBM's x3850 M2 at 20.41@14 tiles - the NEC benchmark sets a scalability factor of 85.7% when moving from 4-socket to 8-socket systems. Both servers from NEC and IBM are scalable systems allowing for multiple chassis to be interconnected to achieve greater CPU-per-system numbers - each scaling in 4-CPU increments - ostensibly for OLTP advantages. The NEC starts at around $70K for 128GB and 48-cores resulting in a $486/VM cost to VMmark.
Also released today, Dell's PowerEdge R905 - with 24 2.8GHz Istanbul cores (8439 SE) and 128GB of ECC DDR2 RAM - secures the number two slot in the 24-category with a posting of 29.51@20 tiles. This represents a tile ratio of 1.475 and tile efficiency of 83.3% for the $29K rack server from Dell at about $240/VM. Compared to its 12-core counterpart, this represents a 91% scalability factor.
If AMD's Istanbul scales to 8-socket at least as efficiently as Dunnington, we should be seeing some 48-core results in the 43.8@30 tile range in the next month or so from HP's 785 G6 with 8-AMD 8439 SE processors. You might ask: what virtualization applications scale to 48-cores when $/VM is doubled at the same time? We don't have that answer, and judging by Intel and AMD's scale-by-hub designs coming in 2010, that market will need to be created at the OEM level.
Based on the performance we're seeing in 8-socket systems relative to 4-socket and the upcoming "massively mult-core" processors in 2010, the law of diminishing returns seems to favor the 4-socket system as the limit for anything but massive OLTP workloads. Even then, we expect to see 48-core in a "4-way" box more efficient than the same number of cores in an 8-way box. The choice in virtualization will continue to be workload biased, with 2P systems offering the best "small footprint" $/VM solution and 4P systems offering the best "large footprint" $/VM solution.
Compared to the leading 24-core "Dunnington" results - held by IBM's x3850 M2 at 20.41@14 tiles - the NEC benchmark sets a scalability factor of 85.7% when moving from 4-socket to 8-socket systems. Both servers from NEC and IBM are scalable systems allowing for multiple chassis to be interconnected to achieve greater CPU-per-system numbers - each scaling in 4-CPU increments - ostensibly for OLTP advantages. The NEC starts at around $70K for 128GB and 48-cores resulting in a $486/VM cost to VMmark.
Also released today, Dell's PowerEdge R905 - with 24 2.8GHz Istanbul cores (8439 SE) and 128GB of ECC DDR2 RAM - secures the number two slot in the 24-category with a posting of 29.51@20 tiles. This represents a tile ratio of 1.475 and tile efficiency of 83.3% for the $29K rack server from Dell at about $240/VM. Compared to its 12-core counterpart, this represents a 91% scalability factor.
If AMD's Istanbul scales to 8-socket at least as efficiently as Dunnington, we should be seeing some 48-core results in the 43.8@30 tile range in the next month or so from HP's 785 G6 with 8-AMD 8439 SE processors. You might ask: what virtualization applications scale to 48-cores when $/VM is doubled at the same time? We don't have that answer, and judging by Intel and AMD's scale-by-hub designs coming in 2010, that market will need to be created at the OEM level.
Based on the performance we're seeing in 8-socket systems relative to 4-socket and the upcoming "massively mult-core" processors in 2010, the law of diminishing returns seems to favor the 4-socket system as the limit for anything but massive OLTP workloads. Even then, we expect to see 48-core in a "4-way" box more efficient than the same number of cores in an 8-way box. The choice in virtualization will continue to be workload biased, with 2P systems offering the best "small footprint" $/VM solution and 4P systems offering the best "large footprint" $/VM solution.
Wednesday, July 15, 2009
RIP Dunnington: HP's 4P/24-core Istanbul Takes VMmark Summit
HP has simultaneously achieved two near identical VMmark scores with their ProLiant DL585 G6 rack server and ProLiant BL685c G6 blade, claiming the summit from the reigning 24-core champion. Since first establishing the 24-core tier VMmark in September 2009, the Intel "Dunnington" 6-core processor (FSB architecture) has gone unchallenged. Now, with the release of the Opteron 8439SE raising the performance bar and the Opteron 8435 making a clear price-performance case, Dunnington's vacation is over.
Today's Istanbul-based achievements - established in the same memory footprint as the top Dunnington - renders the venerable processor all but obsolete, besting the champ by 4 tiles (24 more virtual machines) with a score-tile ratio of 1.5 for the rack system and 1.46 (same as the Dunnington at 14 tiles) for the blade. Using the HP and IBM on-line configuration tools, we established the retail (on-line) price for each system - down to the Fiber Channel HBA's - and compared them for $/VM value. Here are the results:
| HP DL685 G6 | HP BL685c G6 | IBM x3850 M2 | |
| Processor | 4x Opteron 8439SE 2.8GHz | 4x Opteron 8435 2.6GHz | 4x Xeon X7460 2.67GHz |
| Memory | 128GB (16x8GB PC2-5300 Reg ECC) | 128GB (16x8GB PC2-5300 Reg ECC) | 128GB (32x4GB PC2-5300 Reg ECC) |
| LAN Controllers | 1x Dual-Port NC371i 1Gbps, 3x Dual-Port NC380T 1Gbps | 2x Dual-Port NC532i Flex-10 10Gbs, 1x Dual-Port NC360m 1Gbps | 2x Intel PRO 1000PT Dual-Port 1Gbps |
| HBA | Qlogic QMH2462 Dual-Port FC | Qlogic QMH2462 Dual-Port FC | 2x Qlogic QMH2462 Dual-Port FC |
| OS RAID Controller | HP Smart Array P800 | HP Smart Array P400i | HBA |
| OS Disks | 2x 73Gb SAS 10K | 2x 73Gb SAS 10K | SAN |
| On-line Price | $36,862.00 | $35,296.00 | $34,269.00 |
| On-line w/3rd Party Memory | $28,712.00 | $27,356.00 | $33,207.00 |
| VMmark Results | 29.95@20 tiles | 29.19@20 tiles | 20.5@14 tiles |
| VMmark Tile Ratio | 1.5 | 1.46 | 1.46 |
| Cost/VM Retail | $307.18 | $294.13 | $407.96 |
| Cost/VM 3rd Party | $239.27 | $227.97 | $276.73 |
The results indicate a 21-38% savings per-VM for Istanbul over Dunnington in the 4P/24-core virtualization space. This is bread-and-butter territory for VDI implementations and SQL virtualizations, and Intel's last remaining market place for the Dunnington processor. With the top-bin Istanbul weighing-in with 3% better performance, 18% less power consumption and 30% more capacity against Dunnington at the same price point, Intel's 4P gambit is played-out and Nehalem-EX cannot arrive too soon for Intel.
It is worth asking the question: does the HP ProLiant 4P/24-core offer the best value? The answer depends on the value proposition. From a straight $/VM vantage point, the HP DL385 G6 comparison demonstrated a more economical $182/VM - a difference of $40/VM lower than the BL685c G6 - so the 2P rack system still comes out on top for the absolute bottom-line concious. However, for applications like SQL consolidations, the additional savings in licensing on 4P platforms versus 2P platforms dwarfs this differential.
What is clear: AMD's Istanbul solution will remain unchallenged in the 4P space both in raw performance and in price-performance until Nehalem-EX is delivered. That means if Nehalem-EX does not arrive in Q3/2009, the market will likely wait for Q1/2010 to make any long-term purchasing decisions in anticipation of the new platforms slated to break-in the new year.
Tuesday, July 14, 2009
Quick Take: AMD Releases SE/HE Six-Cores
Today AMD published pricing for 5 new Istanbul SKUs - two designated as 105W APC high-performance SE and three as 55W APC low-power HE models.
In the SE category, the 2439SE and 8439SE at 2.8GHz replace the top-bin 2435/8435 Istanbul which occupies the 2.6GHz, 75W APC bin. Besides the clock frequency changes, maximum CPU temp is reduced from 76C to 71C. As with all other Istanbul's so far, these are HT3 bus parts running at 4.8GT/s. Price per socket has been announced at $1,019 and $2,649 for the 2439SE and 8439SE, respectively.
While the new SE parts do little to help the Opteron surpass the X5560 in raw performance, they fit well into the price-performance picture for AMD so long as street prices for the X5560 continue to hover in the $1,200-1,300 range.
[caption id="attachment_783" align="aligncenter" width="450" caption="SPECint_rate2006 - AMD Istanbul SE SKU's"]
[/caption]
In the HE category, the 2425HE/8425HE and 2423HE are new clock speed bins running at 2.1GHz and 2.0GHz, respectively. These parts maintain the same 76C maximum CPU temp as the normal 75W ACP parts, but are selected to consume just 55W ACP. Again, these SKU's also carry the 4.8GT/s HT3 bus of their Istanbul brethren. Pricing per socket has been announced at $523 and $1,514 for the 2425HE and 8425HE, respectively, with the 2423 HE targeted at $455 each.
[caption id="attachment_785" align="aligncenter" width="450" caption="SPECint_rate2006 - AMD Istanbul HE SKU's"]
[/caption]
Here, AMD's lower power target and pricing help the chip maker do some profit-taking as the price-performance of the HE parts appear to offer a measurable advantage over the L5506 (60W TDP) which is circling the $475 region (street price). See AMD's official press release about High Energy Efficiency and the Processing Power of Six-Cores for more details.
SOLORI's Take: AMD has expanded the Istanbul line with both high-performance and low-power SKU's as promised. With DDR3 prices inching downward, AMD's price-performance position is eroding slowly as Q3/2009 approaches. However, the 2-to-1 price penalty for top-bin Xeon/Nehalem platforms will take a lot more time to overcome, leaving the AMD the solid choice for budget conscious virtualization.
What's perhaps more exciting for AMD followers - especially in the good-enough performance market - is sitting in the HE bin. The HE shows weakness in the 2P space, however, against the 2.26GHz L5520 part from Intel which sports 8 thread per CPU and can burst core speeds in excess of 3GHz with its "turbo" feature. This places the 2P 2425 HE somewhere in-between L5506 and L5520 in performance-per-watt, with 2425 HE maintaining a reasonable price-performance advantage.
In the unchallenged 4P space, the 8425 HE, at 2.1GHz and $1,580 (est. street price) offers nearly 3:2 power savings over the standard part offering 24-cores at a little over 200W ACP (4P configurations). This savings will help scale-out clouds both private and public.
[gallery link="file" columns="2"]
In the SE category, the 2439SE and 8439SE at 2.8GHz replace the top-bin 2435/8435 Istanbul which occupies the 2.6GHz, 75W APC bin. Besides the clock frequency changes, maximum CPU temp is reduced from 76C to 71C. As with all other Istanbul's so far, these are HT3 bus parts running at 4.8GT/s. Price per socket has been announced at $1,019 and $2,649 for the 2439SE and 8439SE, respectively.
While the new SE parts do little to help the Opteron surpass the X5560 in raw performance, they fit well into the price-performance picture for AMD so long as street prices for the X5560 continue to hover in the $1,200-1,300 range.
[caption id="attachment_783" align="aligncenter" width="450" caption="SPECint_rate2006 - AMD Istanbul SE SKU's"]
[/caption]In the HE category, the 2425HE/8425HE and 2423HE are new clock speed bins running at 2.1GHz and 2.0GHz, respectively. These parts maintain the same 76C maximum CPU temp as the normal 75W ACP parts, but are selected to consume just 55W ACP. Again, these SKU's also carry the 4.8GT/s HT3 bus of their Istanbul brethren. Pricing per socket has been announced at $523 and $1,514 for the 2425HE and 8425HE, respectively, with the 2423 HE targeted at $455 each.
[caption id="attachment_785" align="aligncenter" width="450" caption="SPECint_rate2006 - AMD Istanbul HE SKU's"]
[/caption]Here, AMD's lower power target and pricing help the chip maker do some profit-taking as the price-performance of the HE parts appear to offer a measurable advantage over the L5506 (60W TDP) which is circling the $475 region (street price). See AMD's official press release about High Energy Efficiency and the Processing Power of Six-Cores for more details.
SOLORI's Take: AMD has expanded the Istanbul line with both high-performance and low-power SKU's as promised. With DDR3 prices inching downward, AMD's price-performance position is eroding slowly as Q3/2009 approaches. However, the 2-to-1 price penalty for top-bin Xeon/Nehalem platforms will take a lot more time to overcome, leaving the AMD the solid choice for budget conscious virtualization.
What's perhaps more exciting for AMD followers - especially in the good-enough performance market - is sitting in the HE bin. The HE shows weakness in the 2P space, however, against the 2.26GHz L5520 part from Intel which sports 8 thread per CPU and can burst core speeds in excess of 3GHz with its "turbo" feature. This places the 2P 2425 HE somewhere in-between L5506 and L5520 in performance-per-watt, with 2425 HE maintaining a reasonable price-performance advantage.
In the unchallenged 4P space, the 8425 HE, at 2.1GHz and $1,580 (est. street price) offers nearly 3:2 power savings over the standard part offering 24-cores at a little over 200W ACP (4P configurations). This savings will help scale-out clouds both private and public.
[gallery link="file" columns="2"]
(Note: SPEC CPU results gathered from published tables at http://spec.org.)
Updated 7/15/2009: Added link to AMD's press release.
Wednesday, July 1, 2009
Lenovo Claims Top VMmark Spot: 2P, 8C
The new top spot for VMmark in the "8 core" category is now held by Lenovo's R525 G2 rack server with a score of 24.35@17 tiles (tile ratio of 1.43 over 102 VMs). As this server appears to be available in the overseas (China) markets only, we can only estimate the street price of the system used in the benchmark based on the reported build-out at to be around $20,330 per server (street):
An EMC CX3-40f was used as the storage backing of the test. The storage system included 4GB cache, 4 enclosures and 55 146GB 15K FC disks (10, 15, 15, 15), and 17 LUNs at 100GB each. Interestingly, a Cisco Linksys SR2024 GigabitEthernet switch was used for the network interconnection (about $299/each at NewEgg) which implies that test results are not being influenced on network performance or latency. Given the use of a 2-port FC HBA for storage, iSCSI network performance is not a factor.
At about $1,094/tile ($182/VM) the new "top dog" delivers its best at a 5% price-per-VM premium over Istanbul's only VMmark results (1.41 tile ratio) and an 80% system price premium (assuming memory sourced by third parties). Since we had to go to the street to configure the Lenovo system, the Istanbul system saves about $1,570 [in mark-up] under similar (non-vendor pricing) circumstances:
Street pricing changes Istanbul's numbers to $892/tile ($149/VM) signifying a 22% per-VM savings and a 52% savings in system price. Given that virtualization systems are generally sold in pairs, this comparison shows that a redundant Istanbul system can be had for less than the cost of a non-redundant Nehalem. For SMB's getting started in virtualization, Istanbul continues to offer a compelling system value proposition over Nehalem.
- Base Lenovo R525 G2 ($4,900 - 30,000 yuan)
- 2 x Intel Xeon X5570 Processors ($1,500/ea)
- 96GB ECC DDR3/1066 (12x8GB) ($900/DIMM from Kingston)
- 1 x Intel 82575EB dual-port GigabitEthernet (on-board)
- 2 x Intel 82571EB dual-port GigabitEthernet (2x PCIe slot, $150/ea)
- 1 x QLogic QLE2462 FC HBA (1x PCIe slot, $1,300)
- 1 x LSI1078 SAS Controller (on-board)
- 2 x SAS OS drive ($300 est.)
An EMC CX3-40f was used as the storage backing of the test. The storage system included 4GB cache, 4 enclosures and 55 146GB 15K FC disks (10, 15, 15, 15), and 17 LUNs at 100GB each. Interestingly, a Cisco Linksys SR2024 GigabitEthernet switch was used for the network interconnection (about $299/each at NewEgg) which implies that test results are not being influenced on network performance or latency. Given the use of a 2-port FC HBA for storage, iSCSI network performance is not a factor.
At about $1,094/tile ($182/VM) the new "top dog" delivers its best at a 5% price-per-VM premium over Istanbul's only VMmark results (1.41 tile ratio) and an 80% system price premium (assuming memory sourced by third parties). Since we had to go to the street to configure the Lenovo system, the Istanbul system saves about $1,570 [in mark-up] under similar (non-vendor pricing) circumstances:
- Base HP DL385 G6 ($5,100)
- 2 x AMD 2435 Istanbul Processors (included)
- 64GB ECC DDR2/800 (8x8GB) ($370/DIMM)
- 2 x Broadcom 5709 dual-port GigabitEthernet (on-board)
- 1 x Intel 82571EB dual-port GigabitEthernet (1x PCIe slot, $150/ea)
- 1 x QLogic QLE2462 FC HBA (1x PCIe slot, $1,300)
- 1 x HP SAS Controller (on-board)
- 2 x SAS OS drive (included)
- $9,810/system total (versus $11,378 complete from HP)
Street pricing changes Istanbul's numbers to $892/tile ($149/VM) signifying a 22% per-VM savings and a 52% savings in system price. Given that virtualization systems are generally sold in pairs, this comparison shows that a redundant Istanbul system can be had for less than the cost of a non-redundant Nehalem. For SMB's getting started in virtualization, Istanbul continues to offer a compelling system value proposition over Nehalem.
Monday, June 22, 2009
AMD Istanbul and Intel Nehalem-EP: Street Prices
It's been three weeks after the official launch of AMD's 6-core Istanbul processor and we wanted to take a look at prevailing street prices for the DIY upgrade option.
After almost two months on the market, the Nehalem has been on the street long enough to see a 1-3% drop in prices. How does Istanbul stack-up against the Nehalem-EP/Xeon pricing?
Compared to the competing Nehalem SKU's, the Istanbul is fetching a premium price. This is likely due to the what AMD perceives to be the broader market that Istanbul is capable of serving (and its relative newness relative to demand, et al). Of course, there are no Xeon Nehalem-EX SKU's in supply to compare against Istanbul in the 4P and 8P segments, but in 2P, it appears Istanbul is running 6% higher at the top bin SKU and 27% higher at the lower bin SKU - with the exception of the 60W TDP part, upon which Intel demands a 13% premium over the 2.2GHz Istanbul part.
This last SKU is the "green datacenter" battleground part. Since the higher priced 2.6GHz Istanbul rates a 15W (ACP) premium over the L5520, it will be interesting to see if system integrators will compare it to the low-power Xeon in power-performance implementations. Comparing SPECpower_ssj2008 between similarly configured Xeon L5520 and X5570, the performance-per-watt is within 2% for relatively anemic, dual-channel 8GB memory configurations.
In a virtualization system, this memory configuration would jump from an unusable 8GB to at least 48GB, increasing average power consumption by another 45-55W and dropping the performance-per-watt ratio by about 25%. Looking at the relative performance-per-watt of the Nehalem-EP as compared to the Istanbul in TechReport's findings earlier this month, one could extrapolate that the virtualization performance-per-watt for Istanbul is very competitive - even with the lower-power Xeon - in large memory configurations. We'll have to wait for similar SPECpower_ssj2008 in 4P configurations to know for sure.
System memory represents 15-20% of system pricing - more in very large memory foot prints. We've indicated that Istanbul's time-to-market strategy shows a clear advantage (CAPEX) in memory pricing alone - more than compensating for the slight premium in CPU pricing.
These parts show a 28%, 40% and 62% premium price for DDR3 components versus DDR2 which indicates Istanbul's savings window is still wide-open. Since DDR3 prices are not expected to fall until Q3 at the earliest, this cost differential is expected to influence "private cloud" virtualization systems more strongly. However, with the 0.3V lower voltage requirement on the DDR3 modules, Nehalem-EP actually has a slight adavantage from a operational power perspective in dual-channel configurations. When using tripple-channel for the same memory footprint, Nehalem-EP's memory consumes about 58% more power (4x8GB vs. 9x4GB).
Istanbul Pricing (Street)
| AMD "Istanbul" Opteron™ Processor Family | |||
| 2400 Series | Price | 8400 Series | Price |
| 2.6GHz Six-Core, 6-Thread AMD Opteron 2435 (75W ACP) | $1060.77 | 2.6GHz Six-Core, 6-Thread AMD Opteron 8435 (75W ACP) | $2,842.14 |
| 2.4GHz Six-Core, 6-Thread AMD Opteron 2431 (75W ACP) | $743.74 $699.00 | 2.4GHz Six-Core, 6-Thread AMD Opteron 8431 (75W ACP) | $2,305.70 |
| 2.2GHx Six-Core, 6-Thread AMD Opteron 2427 (75W ACP) | $483.82 $499.99 | ||
Nehalem-EP/EX Pricing (Street)
After almost two months on the market, the Nehalem has been on the street long enough to see a 1-3% drop in prices. How does Istanbul stack-up against the Nehalem-EP/Xeon pricing?
| Intel "Nehalem" Xeon Processor Family | |||
| EP Series | Price | EX Series | Price |
| 2.66GHz Quad-Core, 8-Thread Intel Xeon EP X5550 (95W TDP) | $999.95 $999.99 | Quad-Core, 8-Thread Intel Xeon EX | TDB |
| 2.4GHz Quad-Core, 8-Thread Intel Xeon EP E5530 (80W TDP) | $548.66 $549.99 | Quad-Core, 8-Thread Intel Xeon EX | TBD |
| 2.26GHz Quad-Core, 8-Thread Intel Xeon EP E5520 (80W TDP) | $400.15 $379.99 | ||
| 2.26GHz Quad-Core, 8-Thread Intel Xeon EP L5520 (60W TDP) | $558.77 $559.99 | ||
Compared to the competing Nehalem SKU's, the Istanbul is fetching a premium price. This is likely due to the what AMD perceives to be the broader market that Istanbul is capable of serving (and its relative newness relative to demand, et al). Of course, there are no Xeon Nehalem-EX SKU's in supply to compare against Istanbul in the 4P and 8P segments, but in 2P, it appears Istanbul is running 6% higher at the top bin SKU and 27% higher at the lower bin SKU - with the exception of the 60W TDP part, upon which Intel demands a 13% premium over the 2.2GHz Istanbul part.
This last SKU is the "green datacenter" battleground part. Since the higher priced 2.6GHz Istanbul rates a 15W (ACP) premium over the L5520, it will be interesting to see if system integrators will compare it to the low-power Xeon in power-performance implementations. Comparing SPECpower_ssj2008 between similarly configured Xeon L5520 and X5570, the performance-per-watt is within 2% for relatively anemic, dual-channel 8GB memory configurations.
In a virtualization system, this memory configuration would jump from an unusable 8GB to at least 48GB, increasing average power consumption by another 45-55W and dropping the performance-per-watt ratio by about 25%. Looking at the relative performance-per-watt of the Nehalem-EP as compared to the Istanbul in TechReport's findings earlier this month, one could extrapolate that the virtualization performance-per-watt for Istanbul is very competitive - even with the lower-power Xeon - in large memory configurations. We'll have to wait for similar SPECpower_ssj2008 in 4P configurations to know for sure.
System Memory Pricing (Street)
System memory represents 15-20% of system pricing - more in very large memory foot prints. We've indicated that Istanbul's time-to-market strategy shows a clear advantage (CAPEX) in memory pricing alone - more than compensating for the slight premium in CPU pricing.
| System Memory Pricing | |||||
| DDR2 Series (1.8V) | Price | DDR3 Series (1.5V) | Price | ||
| $100.00 |
| $138.00 | ||
| $80.00 |
| $132.00 | ||
| $396.00 |
| $1035.00 | ||
These parts show a 28%, 40% and 62% premium price for DDR3 components versus DDR2 which indicates Istanbul's savings window is still wide-open. Since DDR3 prices are not expected to fall until Q3 at the earliest, this cost differential is expected to influence "private cloud" virtualization systems more strongly. However, with the 0.3V lower voltage requirement on the DDR3 modules, Nehalem-EP actually has a slight adavantage from a operational power perspective in dual-channel configurations. When using tripple-channel for the same memory footprint, Nehalem-EP's memory consumes about 58% more power (4x8GB vs. 9x4GB).
ExtremeTech: Shanghai to Istanbul
ExtremeTech runs some tests on the AMD Istanbul 6-core processor and compares the 2435 (2.6GHz) part to the 2384 (2.7GHz) part in a drop-in replacement using PassMark and Spec_JBB2005. Testing was performed in the same Supermicro system running an updated (AGESA 3.5.0.0) BIOS supporting Istanbul processors.
While not as thorough as Scott Wassman's drop-in testing at TechReport (reported earlier this month), ExtremeTech's results and conclusions were about the same: Istanbul makes a great upgrade processor.
Both ExtremeTech and TechReport make compelling upgrade arguments in their testing. Compared to a new system architecture like Nehalem, it is logistically less disruptive - technologically and economically - to certify a CPU upgrade versus as platform replacement. After internal certification, a BIOS and CPU upgrade takes about 20-minutes per system to implement. In a virtualized datacenter where low-level differences are abstracted-away by the hypervisor certification testing should be much less invasive. Likewise, rolling upgrades in a virtualized datacenter with vMotion technology can provide a non-disruptive path from 4-core to 6-core. As Case puts it:
However, this "upgrade proposition" is a difficult position for AMD as it does little to sell new systems. Historically, CPU upgrades only happen in 10-15% of the installed base, making CPU sales based on BIOS/drop-in upgrades an interesting footnote. Integrators want to move new hardware with Instanbul pre-installed, not sell "upgrade packages." Perhaps the dynamics of the new economy will drive a statistical anomoly based on the strength of the Istanbul proposition. Datacenter managers face a familiar dilemma with some new twists.
"Perhaps, the most tell tale result comes from the BOPS rating scored using SpecJBB2005, which simulates a server's ability to process JAVA code. Here, there was a 20% increase in performance, with BOPS increasing from 380721 to 471440. That 20% performance boost would [definitely] be noticed on a busy server in a data center."
While not as thorough as Scott Wassman's drop-in testing at TechReport (reported earlier this month), ExtremeTech's results and conclusions were about the same: Istanbul makes a great upgrade processor.
"It all comes down to simple math, where one has to consider the cost of the CPUs and the time needed to perform an upgrade to see if the return on investment is worthwhile. Most will find that in this case, it is..."- Lloyd Case, ExtremeTech
"And if you have existing, compatible Socket F servers, the Istanbul Opterons should be an excellent drop-in upgrade. They're a no-brainer, really, when one considers energy costs and per-socket/per-server software licensing fees."
Both ExtremeTech and TechReport make compelling upgrade arguments in their testing. Compared to a new system architecture like Nehalem, it is logistically less disruptive - technologically and economically - to certify a CPU upgrade versus as platform replacement. After internal certification, a BIOS and CPU upgrade takes about 20-minutes per system to implement. In a virtualized datacenter where low-level differences are abstracted-away by the hypervisor certification testing should be much less invasive. Likewise, rolling upgrades in a virtualized datacenter with vMotion technology can provide a non-disruptive path from 4-core to 6-core. As Case puts it:
"Simply put, by just upgrading five servers in a data center, data center managers can eliminate the need to purchase an additional server to meet performance needs."
However, this "upgrade proposition" is a difficult position for AMD as it does little to sell new systems. Historically, CPU upgrades only happen in 10-15% of the installed base, making CPU sales based on BIOS/drop-in upgrades an interesting footnote. Integrators want to move new hardware with Instanbul pre-installed, not sell "upgrade packages." Perhaps the dynamics of the new economy will drive a statistical anomoly based on the strength of the Istanbul proposition. Datacenter managers face a familiar dilemma with some new twists.
Sunday, June 21, 2009
Dell Posts Top 4P/16-core VMmark
Dell has posted a new VMmark for its PowerEdge M905 series with a score of 22.90@17 tiles for a 4P Opteron 8393 SE based system. Although newly posted on VMware's VMmark scoreboard, this test was performed on ESX 4.0 (build 159706) and completed May 19, 2009. This is the the first time a 4P Opteron system has exceeded 1-tile-per-core to achieve the highest composite score and bests the previous high score - a Dell R905 - by 1% and 1 tile (102 virtual machines total).
Dell's M905 was fitted with 128GB of PC2-5300 (DDR2/667) registered ECC memory, 4 on-board Broadcom NetXtreme 1Gbps, and a QLogic QLE2462 FC adapter for virtual machine storage. Three Dell/EMC CX3-4of's were used including 6 enclosures with 15 disks per enclosure to deliver 18 LUNx (6 per enclosure, 90 physical disks total.)
Dell's M905 was fitted with 128GB of PC2-5300 (DDR2/667) registered ECC memory, 4 on-board Broadcom NetXtreme 1Gbps, and a QLogic QLE2462 FC adapter for virtual machine storage. Three Dell/EMC CX3-4of's were used including 6 enclosures with 15 disks per enclosure to deliver 18 LUNx (6 per enclosure, 90 physical disks total.)
Thursday, June 18, 2009
First 48-core VMmark Appears
Following in the footsteps of the first 12-core VMmark comes the current champion at 33.85@24 tiles using 48-cores - and, despite the timing, it is not an Istanbul server. In fact, today's leader is the IBM System x3950 M2 running 8, 6-core Intel Xeon MP "Dunnington" X7460 processors with 256GB DDR2/667 RAM (5.3GB/core).
This score edges-out the previous champion - the HP ProLiant DL785 G5 with 8, 4-core Opteron 8393SE processors - which reigned at 31.56@21 tiles. In contrast to the 4-socket, 24-core IBM System x3850 M2 Xeon leading the 24-core category, this doubling of socket/core count resulted in only a 50% increase in capacity. This scaling inefficiency is less typical in 2P-to-4P transition but seems to plague the 4p-to-8P segment.
Looking at the virtualization notes we discover what it takes to keep 48-cores fed to achieve such a benchmark:
While the Dunnington tops the list by sheer brute force, it's safe to assume that - given the 32-core Opteron is nipping at its heels - the 48-core Istanbul results will displace it soon (possibly alluded to in Elisabeth Stahl's "Benchmarking and Performance Blog" reference above). More interestingly, will AMD's much touted "HT Assist" allow the 8P Istanbul to break the 4P-to-8P "curse" of scaling inefficiency? If not, it would show that much work is needed before the relatively "massive " core counts of 2010 are upon us.
This score edges-out the previous champion - the HP ProLiant DL785 G5 with 8, 4-core Opteron 8393SE processors - which reigned at 31.56@21 tiles. In contrast to the 4-socket, 24-core IBM System x3850 M2 Xeon leading the 24-core category, this doubling of socket/core count resulted in only a 50% increase in capacity. This scaling inefficiency is less typical in 2P-to-4P transition but seems to plague the 4p-to-8P segment.
"The x3950 M2 is based on the fourth generation of IBM Enterprise X-Architecture®, and is designed to deliver innovation with enhanced reliability and availability features that enable optimal performance for databases, enterprise applications and virtualized environments.""I'm really looking forward to even more virtualization benchmarks which are coming very soon."
- Elisabeth Stahl, IBM Benchmarking and Systems Performance Blog
Looking at the virtualization notes we discover what it takes to keep 48-cores fed to achieve such a benchmark:
- 4-QLogic QLE2462 HBA's (Dual-port, 4-Gbps FC)
- 1-IBM DS4800 with 4GB cache
- 19 EXP 810 storage expansion units for
- 1.8TB in 49 LUNs
- 280 15K disks total
- 21 IBM x336 clients
- DP 3.2GHz Xeon
- 3GB RAM
- Server 2003 R2
- 2 IBM x335 clients
- DP Xeon 3.06GHz
- 2.5GB RAM
- Server 2003 R2
- Eight vSwitches
- 120 ports total
- 4 Intel PRO 1000PT Dual-port 1Gb Ethernet controllers
- one per vSwitch
While the Dunnington tops the list by sheer brute force, it's safe to assume that - given the 32-core Opteron is nipping at its heels - the 48-core Istanbul results will displace it soon (possibly alluded to in Elisabeth Stahl's "Benchmarking and Performance Blog" reference above). More interestingly, will AMD's much touted "HT Assist" allow the 8P Istanbul to break the 4P-to-8P "curse" of scaling inefficiency? If not, it would show that much work is needed before the relatively "massive " core counts of 2010 are upon us.
Wednesday, June 10, 2009
First 12-core VMmark for Istanbul Appears
VMware has posted the VMmark score for the first Istanbul-based system and it's from HP: the ProLiant DL385 G6. While it's not at the top of the VMmark chart at 15.54@11 tiles (technically it is at the top of the 12-core benchmark list), it still shows a compelling price-performance picture.
For comparison's sake, we've chosen the HP DL385 G5 and HP DL380 G6 as they were configured for their VMmark tests. In the case of the ProLiant DL380 G6, we could only configure the X5560 and not the X5570 as tested so the price is actually LOWER on the DL380 G6 than the "as tested" configuration. Likewise, we chose the PC-6400 (DDR2/667, 8x8GB) memory for the DL 385 G5 versus the more expensive PC-5300 (533) memory as configured in 2008.
As configured for pricing, each system comes with processor, memory, 2-SATA drives and VMware Infrastructure Standard for 2-processors. Note that in testing, additional NIC's, HBA, and storage are configured and such additions are not included herein. We have omitted these additional equipment features as they would be common to a deployment set and have no real influence on relative pricing.
Here's some good news: 50% more cores for only 5% more (sound like an economic stimulus?) The comparison Nehalem-EP is nearly 3x the Istanbul system in price.
Comparing Istanbul's VMmark Scores
For comparison's sake, we've chosen the HP DL385 G5 and HP DL380 G6 as they were configured for their VMmark tests. In the case of the ProLiant DL380 G6, we could only configure the X5560 and not the X5570 as tested so the price is actually LOWER on the DL380 G6 than the "as tested" configuration. Likewise, we chose the PC-6400 (DDR2/667, 8x8GB) memory for the DL 385 G5 versus the more expensive PC-5300 (533) memory as configured in 2008.
As configured for pricing, each system comes with processor, memory, 2-SATA drives and VMware Infrastructure Standard for 2-processors. Note that in testing, additional NIC's, HBA, and storage are configured and such additions are not included herein. We have omitted these additional equipment features as they would be common to a deployment set and have no real influence on relative pricing.
Systems as Configured for Pricing Comparison
| System | Processor | Speed | Cores | Threads | Memory | Speed | Street | ||||||||
| HP ProLiant DL385 G5 | Opteron 2384 | 2.7 | 8 | 8 | 64 | 667 | $10,877.00 | ||||||||
| HP ProLiant DL385 G6 | Opteron 2435 | 2.6 | 12 | 12 | 64 | 667 | $11,378.00 | ||||||||
| HP ProLiant DL380 G6 | Xeon X5560* | 2.93 | 8 | 16 | 96 | 1066 | $30,741.00 |
Here's some good news: 50% more cores for only 5% more (sound like an economic stimulus?) The comparison Nehalem-EP is nearly 3x the Istanbul system in price.
Friday, May 22, 2009
Operton vs. Nehalem-EP at AnandTech
AnandTech's Johan DeGelas has an interesting article on what he calls "real world virtualization" using a benchmark process his team calls "vApus Mk I" and runs it on ESX 3.5 Update 4. Essentially, it is a suite of Web 2.0 flavored apps running entirely on Windows in a mixed 32/64 structure. We're cautiously encouraged by this effort as it opens the field of potential reviewers wide open.
Additionally, he finally comes to the same conclusion we've presented (in an economic impact context) about Shanghai's virtualization value proposition. While his results are consistent with what we have been describing - that Shanghai has a good price-performance position against Nehalem-EP - there are some elements about his process that need further refinement.
Our biggest issue comes with his handling of 32-bit virtual machines (VM) and disclosure of using AMD's Rapid Virtualization Indexing (RVI) with 32-bit VMs. In the DeGalas post, he points out some well known "table thrashing" consequences of TLB misses:
However, the MCS eFMS web portal (2 VMs) is running in a 32-bit OS. What makes this problematic is VMware's default handling of page tables in 32-bit VM's is "shadow page table" using VMware's binary translation engine (BT). In otherwords, RVI is not enabled by default for ESX 3.5.x:
Additionally, he finally comes to the same conclusion we've presented (in an economic impact context) about Shanghai's virtualization value proposition. While his results are consistent with what we have been describing - that Shanghai has a good price-performance position against Nehalem-EP - there are some elements about his process that need further refinement.
Our biggest issue comes with his handling of 32-bit virtual machines (VM) and disclosure of using AMD's Rapid Virtualization Indexing (RVI) with 32-bit VMs. In the DeGalas post, he points out some well known "table thrashing" consequences of TLB misses:
"However, the web portal (MCS eFMS) will give the hypervisor a lot of work if Hardware Assisted Paging (RVI, NPT, EPT) is not available. If EPT or RVI is available, the TLBs (Translation Lookaside Buffer) of the CPUs will be stressed quite a bit, and TLB misses will be costly."
However, the MCS eFMS web portal (2 VMs) is running in a 32-bit OS. What makes this problematic is VMware's default handling of page tables in 32-bit VM's is "shadow page table" using VMware's binary translation engine (BT). In otherwords, RVI is not enabled by default for ESX 3.5.x:
"By default, ESX automatically runs 32bit VMs (Mail, File, and Standby) with BT, and runs 64bit VMS (Database, Web, and Java) with AMD-V + RVI."
- VROOM! Blog, 3/2009
Subscribe to:
Posts (Atom)