Quick Take: PC Pro Recommends 4-node-in-2U Platform

Boston Limited UK has recently received a "recommended" rating from PC Pro UK for its 4-node-in-2U platform with AMD's Istanbul processor on-board. Dubbed the "Boston Quattro 6000GP" and following-up on the 2-node-in-1U "Boston 3000GP" platform, this systems allows for 4-nodes with 2x AMD Istanbul processors per node. This formula yields 8 processors (48 cores) in 2U resulting in a core density of over 1,000 cores per standard 42U rack.

Computational density like this is bound for virtualization and HPC clusters. Judging from the recent reports on Istanbul's virtualization potential and HPL performance, this combination offers a compelling platform alternative to blade computing. In its review, PC Pro UK touched on the platform's power consumption, saying:

"In idle we saw one, two, three and four nodes draw a total of 234W, 349W, 497W and 630W. Under pressure these figures rose to 345W, 541W, 802W and 1026W respectively. Even if you could find an application that pushed the cores this hard you'll find each server node draws a maximum of 256W - not bad for a 12-core system. Dell's PowerEdge R900, reviewed in our sister title IT Pro, has four 130W X7450 six-core processors and that consumes 778W under heavy load."

- PC Pro, UK

Advanced Clustering HPL Comparison: Instanbul vs Nehalem

Advanced Clustering Technologies, based in Kansas City, KS and specializing in HPC solutions, has just released a High-Performance Linpack (HPL) performance report comparing "equivalent" Xeon X5550 and Opteron 2435 systems. According to Advanced Clustering, their goal was "to show the peak performance in terms of GFLOPS (billion floating point operations per second)" of the comparison systems.

In their tests, Advance Clustering attempted to keep platform specifications as uniform as possible (OS, power supply, hard drive). Due to Nehalem's tripple-channel memory, differing amounts of memory were used in the comparison and Advanced Clustering compensated by making adjustments to the problem size in an attempt to utilize 100% of available systems RAM accordingly.

The results showed that AMD's Istanbul delivers 15% more GFLOPS  at a 30% savings in effective system cost ($/GFLOP). Advanced Clustering comments that while Istanbul delivered a higher GFLOP rating than Nehalem, it did so at only 79% of its theoretical potential due to the weaker memory bandwidth of the Socket-F system. From our conversation with AMD's Mike Goddard, we are told that a lot of Istanbul's potential - including much higher memory bandwidth - will realizable only in its sockted G34 incarnation. By that time, the comparison will likely be between Intel's 8-core Nehalem-EX and AMD's 12-core Magny-Cours product.

AMD G34 Motherboards Spotted

Charlie Demerjian was first to post a couple of "spy photos" of two G34 motherboards on his site "SemiAccurate." Both of the boards sighted are dual-socket G34 and appear to carry on-board 10GE ports.

The G34 from Quanta appears to be 2-DPC (16 DIMM slots) with what could be Broadcom's BCM84812 100/1G/10G chipset, while the other looks to be 3-DPC (24 DIMM slots). The larger of the two, from Inventec, looks to be designed for low-profile/HPC applications and appears to support 10GE with a pair of on-board SFP+ connectors - possibly using Intel's 82599ES for 100/1G/10G compatibility.

As we know from our briefings with AMD on the subject of G34, these motherboards should be able to run quad-channel DDR3/1333 at 2-DPC. This allows AMD/G34 to offer more than twice the amount of DDR3/1333 than Nehalem-EP (8-DIMMs/CPU on G34 versus 3-DIMMs/CPU on Nehalem). With 2-sockets yielding 24-cores, 64GB RAM (16 x 4GB DDR3/1333) and on-board, redundant 10GE, AMD's IOMMU and these G34 boards, as Charlie puts it, "scream HPC and heavy-load virtualization."

Based on what we can see, expect to see these boards in the $375 to $450 price range in Q1/2010.

Testing the Tyan Transport GT28 (B2935G28V4H)

Over the next few posts I'll be investigating the Tyan GT28, 1U, 2-socket, dual-node (4-socket total) barebones system. This cost effective system comes some impressive features built-in:

• 1U, Dual-node Chassis (shared 1000W supply)


• Dual AMD Opteron 2000-series processor (per node)


• Shanghai processor support, (AGESA v3.3.0.0 required, BIOS v1.03+)


• 16 DRAM slots (per node, DDR2/REG/ECC)


• 2 x Dual Gigabit Ethernet (per node, Intel 82571EB)


• 20Gbps Infiniband 4x port (per node, B2935G28VHI only, Mellanox MT25204)


• 2 x Hot-Swap SATA/SAS bays (per node)


• 1 x Low-profile PCI Express x 8 slot (per node)


• XGI Volari Z9S Graphics Controller


• nVIDIA nForce Professional 3600 (NFP 3600)


• IPMI/SMDC slot (per node, Tyan M3295-2/M3296)

I am testing mine with the M3296 SMDC card with IP/KVM support (Raritan KIRA100, firmware version 1.00, build 5772, GT28 r01). WARNING: The units I received did not come with AGESA v3.3.0.0 updated BIOS and would not boot with a Shanghai processor - requiring a BIOS update with an older 2000-series chip.

After BIOS update on both the SMDC and motherboard, system boots without issue in my test configuration (per node):

• 4 x Kingston 4GB 240-Pin DDR2 SDRAM DDR2 667 (PC2 5300) ECC Registered Server Memory Model KVR667D2D4P5/4G


• 1 x AMD Opteron 2376 Shanghai 2.3GHz 45nm Socket F 75W Quad-Core Server Processor


• 1 x LSI LSI00110 8-lane, 2.5Gb/s PCI Express SATA / SAS SAS3442E-R 3Gb/s 8-port


• 1 x Tyan M3296 IP/KVM SMDC Card (IPMI 2.0)

I'll have some screen shots of BIOS settings and configurations (courtesy of the M3296's) soon. Since the Gigabit Ethernet controllers support both "jumbo frame" and TCP off-load, we'll try to squeeze-in a few performance comparisons of these features as well.