The Leading Source for Global News and Information Covering the Ecosystem of High Productivity Computing
June 03, 2009
This week AMD launched its six-core "Istanbul" processor, five months ahead of its original schedule. The new 45nm Opterons will confront Intel's recently hatched Nehalem EP quad-core chips and its older 6-core Dunnington Xeon processor as AMD attempts to regain lost market share in the critical x86 server market.
Unfortunately, both Intel and AMD will struggle with high-end chip sales for the rest of the year as a result of plummeting server revenue. Gartner and IDC recently reported 64-bit x86 server sales for Q1 down by more than 25 percent due to the continued weakness in the global economy. Ironically, adding additional cores to processors will exacerbate the problem in the long term, as traditional enterprise users look to consolidate computing infrastructure via virtualization. Regardless, both analyst groups are forecasting server revenue growth to return in 2010.
The dismal news on server sales didn't prevent most major and minor OEMs from jumping on the Istanbul bandwagon, and announcing upgraded server (and workstation) platforms that incorporate AMD's new six-core Opteron. Within a couple of days of the Istanbul launch, Appro, Cray, HP, Sun Microsystems, SGI, Microway, Supermicro, Tyan, BOXX, ZT Systems, and Nor-Tech had announced or previewed new six-core server gear. Of the major system vendors, only IBM kept its plans for its new Istanbul-based servers under wraps.
Cray announced immediate support for Istanbul chips in the company's high-end XT5 and mid-range XT5m lines. "It's an easy transition for Cray," said Barry Bolding, Cray VP of Scalable Systems in a statement in support of the AMD launch. "With the XT platform, the Istanbul upgrade is going to be seamless."
Because Istanbul uses the same socket (Socket F) as the dual- and quad-core Opterons, the new six-core chip is able to plug into existing Opteron servers if users desire a field upgrade. This has been an especially useful feature for HPC setups, in which Opteron-based supercomputers are built with thousands of server nodes. Replacing the entire system would be prohibitively expensive, but swapping in newer, faster chips allows these big, expensive machines to get a second and third life. Istanbul will also be upwardly compatible with the upcoming all-AMD "Fiorano" chipset, which is scheduled to be introduced sometime later this year.
For two-processor (2P) configurations, AMD is offering the 2.2 GHz Opteron 2427 at $455; 2.4 GHz Opteron 2431 at $698; and the 2.6 GHz Opteron 2435 at $989. For 4P and 8P systems, the 2.4 GHz Opteron 8431 costs $2,149. All current models have an ACP of 75 watts. In the second half of the year, AMD plans to release additional Istanbul models, including SE (105 watts), HE (55 watts), and EE (40 watts).
For the core-hungry HPC crowd especially, Istanbul offers a lot more usable CPU cycles, while keeping power consumption in line with the previous Opteron chips. AMD is predicting 30 to 40 percent better performance compared to its quad-core Shanghai processor. To deal with the extra cores, AMD has added a number of new features. The most notable addition is HT Assist, a mechanism that speeds up interprocessor cache access by short-circuiting cache probing. This feature is expected to pay big dividends in performance in 4P and 8P configurations. The other big improvement is the implementation of HyperTransport 3 links, which supports data transfers up to 57.6 GB/second per processor.
Remaining essentially unchanged is the memory subsystem. Compared to the quad-core Nehalem's three-channel DDR3 memory, Istanbul is sticking with its predecessor's two-channel DDR2 set up. That means Intel will outrun Istanbul on most memory-bound apps, albeit at some power and cost penalty.
Perhaps even more important is the support for simultaneous multithreading in the Nehalem architecture, and the lack of it in Istanbul. Each Nehalem core can manage two threads, while AMD is sticking with the single-thread-per-core model. So a quad-core Nehalem can handle eight threads simultaneously, Istanbul just six. The actual performance boost realized by simultaneous multithreading is very application dependent, but for highly-threaded HPC workloads, the speed-up can be significant.
The Tech Report pitted the Istanbul chip against the Nehalem EP in a series of application benchmarks (including some typical HPC-like apps) and generated a nice write-up and analysis of the results. Since the tests were performed with a 2P version of the six-core Opteron to keep the playing field level (the 4P/8P Nehalem EX won't be launched until next year), AMD's strength in the greater-than-two-socket processor space was not measured. Nevertheless, the Tech Report concluded that although the latest quad-core Nehalem was superior in pure speed, Istanbul was definitely in the game, especially when it came to performance per watt:
Page: 1 of 2(Digg, Technorati, more)
PGI Accelerator™ Fortran 95/03 and C99 compilers for x64+NVIDIA
Accelerate applications on x64+GPU platforms by adding OpenMP-like compiler directives to existing Fortran and C programs. Available now for Linux, MacOS and Windows. Download a free 15 day trial.
Platform HPC Workgroup Manager
Platform HPC Workgroup Manager integrates all the cluster productivity tools you need to deploy, run and manage your HPC environment.
Mar 19 | OfficialWire | New super to support intelligence work Down Under. Read more...
Mar 18 | ChannelWeb | Westmere parts already showing up in HPC machines. Read more...
Mar 17 | The Register | But what about the tier ones? Read more...
Mar 17 | Cadalyst Magazine | A new generation of workstations is changing the nature of technical computing. Read more...
Mar 17 | Linux Magazine | Latest iteration of Sun Grid Engine able to tap into Cloud. Read more...
Jan 12 | | In-depth look at vSMP Foundation server virtualization technology, technical implementation, use cases and capabilities. The technical whitepaper provides an architectural overview and details on the three vSMP Foundation products: vSMP Foundation for SMP, vSMP Foundation for Cluster and vSMP Foundation for Cloud.
Jan 18 | | This white paper discusses Gore’s copper cable assemblies, and how they continue to exceed the standards for providing reliable, cost-effective solutions for high-performance computer applications.
Join this online panel discussion for live Q&A with leading industry experts, analysts, and end-users to discuss the latest innovations, best practices, barriers to implementation, and measurable benefits of server virtualization with a particular focus on today's real world solutions.
Learn about scalable fault-tolerant architectures and examples of energy efficient and scalable supercomputing clusters using dual QDR InfiniBand to combine capacity computing with network failover capabilities with the help of programming languages such as MPI and a robust Linux cluster management package.
LIVE@SCO9: The IBM team discusses new innovations in hardware, software and services that help clients better understand their workloads and get insight from their R&D efforts. Technology demonstrations include the soon-to-be-released Power7 HPC processor, the DCS990 system with 2.4 petabytes of storage, the xCAT management tool, secure HPC cloud computing and more. Winners of two HPCwire Readers' and Editors’ Choice Awards! Take the IBM virtual tour at SC09 or more information go online to: http://www-03.ibm.com/systems/deepcomputing/sc09.html