Startup Delivers Technology to Boost Memory Capacity

By Michael Feldman

March 7, 2008

For some time there has been a growing disparity between processor power and memory capacity. As processors and systems have become more computationally dense, memory capacities have been growing more slowly. This has been less of an issue in PCs, where both compute power and memory are relatively abundant, but as high performance computing, virtualization, and large transactional databases take hold in the datacenter, memory has increasingly become the bottleneck to system performance. While lack of memory bandwidth may be the ultimate constraint for maximum CPU performance, for memory-constrained applications, sufficient RAM is required just to get past the formidable disk bottleneck.

Unfortunately scaling memory is not nearly as simple as scaling processing power. An economic discontinuity exists in the DRAM market. While 1 Gbit DRAM chips cost around $3, a 2 Gbit DRAM chips cost $30 to $50 dollars. And since commodity systems with DDR2 memory controllers are designed with only a limited number of DIMMs (which themselves contain a fixed number of DRAM chips), the only way to get more memory is to go with the outrageously expensive 2 Gbit chips. Since the memory controllers are designed to talk with a limited number of DRAMs, this effectively limits affordable capacity.

But MetaRAM, a 2-year-old fabless semiconductor startup, is attempting to change the game. Last week, the company unveiled a new product line that is able to aggregate 1 Gbit DRAMs so as to increase affordable main memory by a factor of two to four. The 34-person company has developed new memory technology — called MetaSDRAM — that allows 8GB and 16GB DIMMs to be built from low-cost 1 Gbit memory components. According to the company, a four-socket server can be populated with 256 GB of MetRAM memory for only $50,000. Using conventional DDR2 memory, this same system would cost $500,000 — $480,000 of which is just DRAM.

The new MetaRAM device contains a special chipset that virtualizes the 1 Gbit components so that a DDR2 memory controller thinks its talking with 2 Gbit or 4 Gbit DRAMs instead of multiple 1 Gbit DRAMs. (Fortunately, the memory controllers aren’t smart enough to know that 4 Gbit DRAMs don’t even exist yet.) Internally, the 1 Gbit DRAMs think they’re talking directly with the memory controller. The chipset in the middle contains the smarts to do the needed DDR2 protocol translations and keep the memory accesses coherent.

“So we look like a DRAM to the memory controller, and a memory controller to the DRAM,” says Suresh Rajan, MetaRAM VP Marketing.

Rajan and CEO Fred Weber founded MetaRAM in January 2006 with the idea of leapfrogging the “memory capacity gap.” Weber, a former CTO at AMD who helped drive the 64-bit x86 revolution at his former company, saw the new opportunity in the processor-memory disparity and decided that the growth of Web 2.0, datacenter virtualization, and high performance computing were going to create a large demand for x86-based systems with much larger memory footprints. In the HPC realm, memory-hungry applications in aerospace/automotive, financial services, digital content creation and rendering, oil and gas exploration, and semiconductor design and simulation would especially benefit from lots of RAM.

“Where this really gets exciting is that we can build very powerful clusters at very affordable prices to solve problems that are memory bound,” says Rajan. “Obviously some applications are compute bound, but others are memory bound, and for the latter, this is a great technology.”

The first two MetaRAM products announced are the 8 GB and 16 GB R-DIMMs. The 8 GB chipset is currently in full production and is available for $200 , while the 16 GB is qualified, but not yet in production, and will be priced at $450. Hynix Semiconductor and SMART Modular Technologies have partnered with MetaRAM to produce the memory modules for system vendors. Hynix is a large semiconductor manufacturer with relationships with many of the tier one OEMs; SMART Modular is focused on the tier two players, plus system builders and other channel customers.

Server/cluster vendors that are currently committed to carrying the first 8GB DIMMs include Appro, Rackable, Colfax International and Verari. System availability is expected in Q1 of this year, that is, any day now. Rajan says they’re talking with all the usual HPC system vendors, and he says we can anticipate more OEMs to announce MetaRAM support at some point. He notes that they currently have a working eight-socket Opteron Sun machine in-house with 512 GB of MetaRAM memory.

Some OEMs might be resistant to offering high-memory servers at commodity pricing, since this might cut into their high-end server margins. But if more system vendors offer MetaRAM-based systems, it would be difficult for their competitors to resist, especially in the highly price-sensitive x86 server market.

The MetaRAM chipset is plug compatible with any DDR2-compliant memory controller and has been qualified for both Opteron and Xeon chipsets. Currently, the MetaRAM hardware will end up mostly in AMD Opteron-based systems, since DDR2 memory controllers are standard issue there. Intel’s relatively new 5100 MCH chipset also supports DDR2, and as time passes, more Intel chipsets are likely to switch from the power-hungry, higher latency FB-DIMMs, to parallel DDR2 memory. The original idea behind the FB-DIMM was also to provide higher capacity memory, but high power usage, added latency and chipset inflexibility have proven barriers for more widespread industry adoption.

The MetaRAM technology had to overcome a few challenges as well. The chipset does add some additional latency compared to directly connected DDR2 memory. But according to MetaRAM tests, latency increases only a percent or two in a four-socket system — much less than an FB-DIMM. For memory-bound applications, the company expects any small rise in latency to be dwarfed by an average increase of between 20 and 30 percent in overall performance.

The biggest hurdle that the MetaRAM had to address was power usage. By doubling or quadrupling the number of 1 Gbit DRAMs on the DIMM and adding the extra chipset, they had to find a way to keep power usage in the same neighborhood of a low-memory DIMM. This is accomplished via their “WakeOnUse” technology, which sends power only to DRAM elements that are being accessed, while putting the rest of memory to sleep. It’s a little like turning the lights on in a house as someone walks from room to room, while switching off the lights in the remainder of the house. Rajan claims this yields about 30 percent greater power-efficiency on a watts per gigabyte basis.

According to Weber and Rajan, the MetaRAM technology accelerates affordable memory capacity by two to four years, and will ride commodity memory to stay ahead of the curve. While 1 Gbit DRAMs currently represent the best price-capacity memory today, by 2010, 2 Gbit DRAMs will be the volume solution, and by 2013, 4 Gbit devices should start to show up. MetaRAM also intends to modify the chipset to support DDR3 memory, when that technology goes mainstream.

“I think for the next five to eight years, we see a clear path ahead of us,” says Rajan. “It is our belief as time progresses that our technology will become more and more critical for the server market.”

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