Chipmakers Look to Rock High-End Server Biz

By Michael Feldman

April 1, 2010

The x86 CPU festivities are over for now, but the party’s just getting started. The debut of the latest Intel Xeon and AMD Opteron processors over the last few weeks marks something of a turning point for server makers. For one thing, the introduction of Intel’s 6-core Westmere EP and 8-core Nehalem EX CPUs, and AMD’s 12-core Magny-Cours processor marks the beginning of the end of the quad-core era. Given that, HPC servers with fewer than double-digit core counts will soon be the exception rather than the rule.

AMD and Intel are attacking the high-server space somewhat differently, though. With Westmere EP (now the Xeon 5600), Intel is continuing its traditional 2P server business. But with Nehalem EX (now the Xeon 7500), Intel is charting new territory — big shared memory SMP machines. Intel also introduced the Xeon 6500, a 2P-only variant of Nehalem EX, ostensibly aimed at the HPC market. Meanwhile AMD has consolidated its performance-oriented 2P and 4P products into a single Opteron 6000 product line, starting with Magny-Cours (now the Opteron 6100).

From a price-performance perspective AMD has a good story. With the 6100 Opterons, AMD will go head-to-head with the 2P 5600 Xeons, which have faster cores, but fewer of them. The mid-range Opteron 6174, which sports 12 cores and runs at 2.2 GHz, costs $1,165 in quantity. A Xeon with comparable performance is the 6-core X5680, which is clocked at 3.33 GHz and costs $1664. Although the individual Xeon cores run faster, for many types of parallel workloads, the additional six cores on the Opteron will make up the difference, and then some. The fact that the Intel architecture implements HyperThreading, which handles two threads per core, only boosts performance by 10 to 20 percent. And in some cases, such as Linpack, it doesn’t help at all. Since the 6100 Opterons have four channels of memory and support up to 12 DIMM slots per socket, compared to three channels and nine DIMMs for the Xeon 5600s, the AMD CPUs have an additional advantage on memory-loving apps.

The 6100 Opterons will also go up against the 6500 Xeons in the 2P arena, as well as the 7500 Xeons in the 4P space. Here the Xeons go up to eight cores, the memory channel differential has been equalized at four apiece, and the memory capacity advantage is now with Intel at 16 DIMMs per socket. But the EX-class parts are even more expensive than Xeon 5600 chips. For example, the 8-core 6500 and 7500 products cost between $2,461 and $3,692, which is more than two and three times the price, respectively, of the Opteron 6174 mentioned above. Even the least expensive 6-core EX, which is the 1.86 GHz Xeon x7530, costs $200 more than the 6174.

The bottom line is that the new Magny-Cours processors look like a very competitive solution for 2P and 4P servers. But the 4P story is particularly interesting. AMD is pushing this 6000 series as a platform that does away with the “4P tax.” The tax refers to the traditional premium vendors charged for CPUs and chipsets that support 4-socket servers. Since the 6000 hardware can be used in both 2P and 4P boxes, you can actually save money by consolidating dual-socket servers (as long as you don’t need to spread out the processors over more boxes to get at more I/O). “The only reason 4P processors have been priced like they have is because there’s a guy in the business who owns a large chunk of the market and has been pricing that way for 20 years,” says John Fruehe, who heads AMD’s Product Marketing of the Server and Workstation Division. “It’s more tradition than technology that has forced that price.”

That “guy,” that Fruehe is referring to is, of course, Intel. But prior to Magny-Cours, AMD also priced its 4P/8P Opteron 8000 CPUs at a premium in relation to its 2P Opteron 2000 parts. But according to him, they eventually came to the conclusion that the demand for 4P servers was being inhibited by this pricing model. In fact, according to Fruehe, the quad-socket Opteron-based supercomputers on the TOP500 list came about because AMD gave the system vendors a nice volume discount on Opteron 8000 CPUs. “Generally speaking those were deals where an 8000 processor was priced like a 2000,” he told me. “Suddenly the economics made sense.”

Although he wouldn’t point to any specific systems, the half-petaflop “Ranger” Sun Constellation cluster at TACC, which uses quad-socket Opteron-based blades, almost certainly fits in this category. Fruehe maintains AMD still turned a profit on these supercomputer deals, but it gave them the idea that it could move a lot more product by pricing 4P parts like 2P parts. They believe that this strategy will unleash this market in HPC and across enterprise computing.

On the other hand, AMD has decided leave the 8P (and above), at least for the time being. At 60K or so processors per year, the company has calculated this is too small a market to give special consideration to. One might ask, though: If the 4P servers are such a good idea, why not 8P, 16P and so on? As you keep adding processors, or cores for that matter, memory bandwidth and capacity become the limiting factor. As AMD and Intel keep pouring on the cores, they’re forced to rebalance the memory subsystem.

The idea behind the new Xeon 7500 line is to max out both compute and memory in a familiar x86 package. As of this week, OEMs can build 8-socket commodity boxes with 1 TB of memory. With this approach, not only does Intel think it can edge out proprietary RISC CPUs in SMP servers used for mission-critical computing, it also believes it can grow the SMP market overall.

According to David Kanter at Real World Technologies, that might indeed come to pass. Although in the past there were multiple reasons that 8P servers represented a specialty market, a confluence of commodity technologies, including the new Xeons themselves, are changing the economics. In a recent article, Kanter writes:

The primary barriers to adoption for large x86 servers are software, maturity and cost/benefit. Scalable applications that would benefit from 8S servers are not common. Some classic examples include I/O heavy workloads like ERP, transactional or analytic databases and also select HPC workloads that favor shared memory rather than message passing. More recently, server consolidation using virtualization has emerged as an important workload. In 2010, there are simply more scalable workloads than were previously available.

Kanter goes on to analyze how the different pieces of the enterprise ecosystem are evolving, and how they could favor a shift to commodity 8P servers. For now, AMD seems content to play it conservative and let Intel test the SMP waters. If successful, perhaps the junior member of the x86 franchise will jump in after Intel has built the market. In the meantime, AMD is focused on rebuilding its server mojo in the 2P and 4P sweet spots. Magny-Cours looks like a fine start.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire