NVIDIA's next-gen Blackwell platform, which includes B-series GPUs and integrates NVIDIA's own Grace Arm CPU in models such as the GB200, represents a significant development. TrendForce points out that the GB200 and its predecessor, the GH200, both feature a combined CPU+GPU solution, primarily equipped with the NVIDIA Grace CPU and H200 GPU. However, the GH200 accounted for only approximately 5% of NVIDIA's high-end GPU shipments. The supply chain has high expectations for the GB200, with projections suggesting that its shipments could exceed millions of units by 2025, potentially making up nearly 40 to 50% of NVIDIA's high-end GPU market.

Although NVIDIA plans to launch products such as the GB200 and B100 in the second half of this year, upstream wafer packaging will need to adopt more complex and high-precision CoWoS-L technology, making the validation and testing process time-consuming. Additionally, more time will be required to optimize the B-series for AI server systems in aspects such as network communication and cooling performance. It is anticipated that the GB200 and B100 products will not see significant production volumes until 4Q24 or 1Q25.

On March 29th, the United States announced another round of updates to its export controls, targeting advanced computing, supercomputers, semiconductor end-uses, and semiconductor manufacturing products. These new regulations, which took effect on April 4th, are designed to prevent certain countries and businesses from circumventing U.S. restrictions to access sensitive chip technologies and equipment. Despite these tighter controls, TrendForce believes the practical impact on the industry will be minimal.

The latest updates aim to refine the language and parameters of previous regulations, tightening the criteria for exports to Macau and D:5 countries (China, North Korea, Russia, Iran, etc.). They require a detailed examination of all technology products' Total Processing Performance (TPP) and Performance Density (PD). If a product exceeds certain computing power thresholds, it must undergo a case-by-case review. Nevertheless, a new provision, Advanced Computing Authorized (ACA), allows for specific exports and re-exports among selected countries, including the transshipment of particular products between Macau and D:5 countries.

It's official: NVIDIA delivered the world's fastest platform in industry-standard tests for inference on generative AI. In the latest MLPerf benchmarks, NVIDIA TensorRT-LLM—software that speeds and simplifies the complex job of inference on large language models—boosted the performance of NVIDIA Hopper architecture GPUs on the GPT-J LLM nearly 3x over their results just six months ago. The dramatic speedup demonstrates the power of NVIDIA's full-stack platform of chips, systems and software to handle the demanding requirements of running generative AI. Leading companies are using TensorRT-LLM to optimize their models. And NVIDIA NIM—a set of inference microservices that includes inferencing engines like TensorRT-LLM—makes it easier than ever for businesses to deploy NVIDIA's inference platform.

Raising the Bar in Generative AI
TensorRT-LLM running on NVIDIA H200 Tensor Core GPUs—the latest, memory-enhanced Hopper GPUs—delivered the fastest performance running inference in MLPerf's biggest test of generative AI to date. The new benchmark uses the largest version of Llama 2, a state-of-the-art large language model packing 70 billion parameters. The model is more than 10x larger than the GPT-J LLM first used in the September benchmarks. The memory-enhanced H200 GPUs, in their MLPerf debut, used TensorRT-LLM to produce up to 31,000 tokens/second, a record on MLPerf's Llama 2 benchmark. The H200 GPU results include up to 14% gains from a custom thermal solution. It's one example of innovations beyond standard air cooling that systems builders are applying to their NVIDIA MGX designs to take the performance of Hopper GPUs to new heights.

One key takeaway from the ongoing GTC is that Nvidia's AI empire has taken shape with strong partnerships from TSMC and other Taiwanese makers, such as those major server ODMs.

According to the news report from the technology-focused media DIGITIMES Asia, during his keynote at GTC on March 18, Huang underscored his company's partnerships with TSMC, as well as the supply chain in Taiwan. Speaking to the press later, Huang said Nvidia will have a very strong demand for CoWoS, the advanced packaging services TSMC offers.

GIGABYTE Technology and Giga Computing, a subsidiary of GIGABYTE and an industry leader in enterprise solutions, will showcase their solutions at the GIGABYTE booth #1224 at NVIDIA GTC, a global AI developer conference running through March 21. This event will offer GIGABYTE the chance to connect with its valued partners and customers, and together explore what the future in computing holds.

The GIGABYTE booth will focus on GIGABYTE's enterprise products that demonstrate AI training and inference delivered by versatile computing platforms based on NVIDIA solutions, as well as direct liquid cooling (DLC) for improved compute density and energy efficiency. Also not to be missed at the NVIDIA booth is the MGX Pavilion, which features a rack of GIGABYTE servers for the NVIDIA GH200 Grace Hopper Superchip architecture.

SK Hynix believes that it leads the industry with the development and production of High Bandwidth Memory (HBM) solutions, but rival memory manufacturers are working hard on equivalent fifth generation packages. NVIDIA was expected to select SK Hynix as the main supplier of HBM3E parts for utilization on H200 "Hopper" AI GPUs, but a surprise announcement was issued by Micron's press team last month. The American firm revealed that HBM3E volume production had commenced: ""(our) 24 GB 8H HBM3E will be part of NVIDIA H200 Tensor Core GPUs, which will begin shipping in the second calendar quarter of 2024. This milestone positions Micron at the forefront of the industry, empowering artificial intelligence (AI) solutions with HBM3E's industry-leading performance and energy efficiency."

According to a Korea JoongAng Daily report, this boast has reportedly "shocked" the likes of SK Hynix and Samsung Electronics. They believe that Micron's: "announcement was a revolt from an underdog, as the US company barely held 10 percent of the global market last year." The article also points out some behind-the-scenes legal wrangling: "the cutthroat competition became more evident when the Seoul court sided with SK Hynix on Thursday (March 7) by granting a non-compete injunction to prevent its former researcher, who specialized in HBM, from working at Micron. He would be fined 10 million won for each day in violation." SK Hynix is likely pinning its next-gen AI GPU hopes on a 12-layer DRAM stacked HBM3E product—industry insiders posit that evaluation samples were submitted to NVIDIA last month. The outlook for these units is said to be very positive—mass production could start as early as this month.

TrendForce highlights the current landscape of the HBM market, which as of early 2024, is primarily focused on HBM3. NVIDIA's upcoming B100 or H200 models will incorporate advanced HBM3e, signaling the next step in memory technology. The challenge, however, is the supply bottleneck caused by both CoWoS packaging constraints and the inherently long production cycle of HBM—extending the timeline from wafer initiation to the final product beyond two quarters.

The current HBM3 supply for NVIDIA's H100 solution is primarily met by SK hynix, leading to a supply shortfall in meeting burgeoning AI market demands. Samsung's entry into NVIDIA's supply chain with its 1Znm HBM3 products in late 2023, though initially minor, signifies its breakthrough in this segment.

During the 2024 SIEPR Economic Summit, NVIDIA CEO Jensen Huang acknowledged that the company's next-generation DGX systems, designed for AI and high-performance computing workloads, will require liquid cooling due to their immense power consumption. Huang also hinted that these new systems are set to be released in the near future. The revelation comes as no surprise, given the increasing power of GPUs needed to satisfy AI and machine learning applications. As computational requirements continue to grow, so does the need for more powerful hardware. However, with great power comes great heat generation, necessitating advanced cooling solutions to maintain optimal performance and system stability. Liquid cooling has long been a staple in high-end computing systems, offering superior thermal management compared to traditional air cooling methods.

By implementing liquid cooling in the upcoming DGX systems, NVIDIA aims to push the boundaries of performance while ensuring the hardware remains reliable and efficient. Although Huang did not provide a specific release date for the new DGX systems, his statement suggests that they are on the horizon. Whether the next generation of DGX systems uses the current NVIDIA H200 or the upcoming Blackwell B100 GPU as their primary accelerator, the performance will undoubtedly be delivered. As the AI and high-performance computing landscape continues to evolve, NVIDIA's position continues to strengthen, and liquid-cooled systems will certainly play a crucial role in shaping the future of these industries.

Brett Simpson, Partner and Co-Founder of Arete Research, sat down with AMD CTO Mark Papermaster during the former's "Investor Webinar Conference." A transcript of the Arete + AMD question and answer session appeared online last week—the documented fireside chat concentrated mostly on "AI compute market" topics. Papermaster was asked about his company's competitive approach when taking on NVIDIA's very popular range of A100 and H100 AI GPUs, as well as the recently launched GH200 chip. The CTO did not reveal any specific pricing strategies—a "big picture" was painted instead: "I think what's important when you just step back is to look at total cost of ownership, not just one GPU, one accelerator, but total cost of ownership. But now when you also look at the macro, if there's not competition in the market, you're going to see not only a growth of the price of these devices due to the added content that they have, but you're -- without a check and balance, you're going to see very, very high margins, more than that could be sustained without a competitive environment."

Papermaster continued: "And what I think is very key with -- as AMD has brought competition market for these most powerful AI training and inference devices is you will see that check and balance. And we have a very innovative approach. We've been a leader in chiplet design. And so we have the right technology for the right purpose of the AI build-out that we do. We have, of course, a GPU accelerator. But there's many other circuitry associated with being able to scale and build out these large clusters, and we're very, very efficient in our design." Team Red started to ship its flagship accelerator, Instinct MI300X, to important customers at the start of 2024—Arete Research's Simpson asked about the possibility of follow-up models. In response, AMD's CTO referenced some recent history: "Well, I think the first thing that I'll highlight is what we did to arrive at this point, where we are a competitive force. We've been investing for years in building up our GPU road map to compete in both HPC and AI. We had a very, very strong harbor train that we've been on, but we had to build our muscle in the software enablement."

NVIDIA today announced that Australia's Pawsey Supercomputing Research Centre will add the NVIDIA CUDA Quantum platform accelerated by NVIDIA Grace Hopper Superchips to its National Supercomputing and Quantum Computing Innovation Hub, furthering its work driving breakthroughs in quantum computing.

Researchers at the Perth-based center will leverage CUDA Quantum - an open-source hybrid quantum computing platform that features powerful simulation tools, and capabilities to program hybrid CPU, GPU and QPU systems - as well as, the NVIDIA cuQuantum software development kit of optimized libraries and tools for accelerating quantum computing workflows. The NVIDIA Grace Hopper Superchip - which combines the NVIDIA Grace CPU and Hopper GPU architectures - provides extreme performance to run high-fidelity and scalable quantum simulations on accelerators and seamlessly interface with future quantum hardware infrastructure.

The Barcelona Supercomputing Center (BSC) and the State University of New York (Stony Brook and Buffalo campuses) have pitted NVIDIA's relatively new CG100 "Grace" Superchip against several rival products in a "wide variety of HPC and AI benchmarks." Team Green marketing material has focused mainly on the overall GH200 "Grace Hopper" package—so it is interesting to see technical institutes concentrate on the company's "first true" server processor (ARM-based), rather than the ever popular GPU aspect. The Next Platform's article summarized the chip's internal makeup: "(NVIDIA's) Grace CPU has a relatively high core count and a relatively low thermal footprint, and it has banks of low-power DDR5 (LPDDR5) memory—the kind used in laptops but gussied up with error correction to be server class—of sufficient capacity to be useful for HPC systems, which typically have 256 GB or 512 GB per node these days and sometimes less."

Benchmark results were revealed at last week's HPC Asia 2024 conference (in Nagoya, Japan)—Barcelona Supercomputing Center (BSC) and the State University of New York also uploaded their findings to the ACM Digital Library (link #1 & #2). BSC's MareNostrum 5 system contains an experimental cluster portion—consisting of NVIDIA Grace-Grace and Grace-Hopper superchips. We have heard plenty about the latter (in press releases), but the former is a novel concept—as outlined by The Next Platform: "Put two Grace CPUs together into a Grace-Grace superchip, a tightly coupled package using NVLink chip-to-chip interconnects that provide memory coherence across the LPDDR5 memory banks and that consumes only around 500 watts, and it gets plenty interesting for the HPC crowd. That yields a total of 144 Arm Neoverse "Demeter" V2 cores with the Armv9 architecture, and 1 TB of physical memory with 1.1 TB/sec of peak theoretical bandwidth. For some reason, probably relating to yield on the LPDDR5 memory, only 960 GB of that memory capacity and only 1 TB/sec of that memory bandwidth is actually available."

NVIDIA's intrepid CEO, Jensen Huang, has spent a fair chunk of January travelling around China—news outlets believe that Team Green's leader has conducted business meetings with very important clients in the region. Insiders proposed that his low-profile business trip included visits to NVIDIA operations in Shenzhen, Shanghai and Beijing. The latest updates allege that a stopover in Taiwan was also planned, following the conclusion of Mainland activities. Photos from an NVIDIA Chinese new year celebratory event have been spreading across the internet lately—many were surprised to see Huang appear on-stage in Shanghai and quickly dispense with his trademark black leather jacket. He swapped into a colorful "Year of the Wood Dragon" sleeveless shirt for a traditional dance routine.

It was not all fun and games during Huang's first trip to China in four years—inside sources have informed the Wall Street Journey about growing unrest within the nation's top ranked Cloud AI tech firms. Anonymous informants allege that leadership, at Alibaba Group and Tencent, are not happy with NVIDIA's selection of compromised enterprise GPUs—it is posited that NVIDIA's President has spent time convincing key clients to not adopt natively-developed solutions (unaffected by US Sanctions). The short hop over to Taiwan is reported not to be for R&R purposes—insiders had Huang's visiting key supply partners; TSMC and Wistron. Industry experts think that these meetings are linked to NVIDIA's upcoming "Blackwell" B100 AI GPU, and "supercharged" H200 "Hopper" accelerator. It is too early for the rumor mill to start speculation about nerfed versions of NVIDIA's 2024 enterprise products reaching Chinese shores, but Jensen Huang is seemingly ready to hold diplomatic talks with all sides.