NVIDIA’s next-generation Blackwell architecture can play a significant role.
According to NVIDIA’s roadmap, it will soon launch its next-generation Blackwell architecture. The company always releases a new architecture and data center product first, then announces a reduced GeForce version a few months later, so this is also expected this time. As evidence that NVIDIA is about to launch its new data center GPU, a Dell executive has shared some interesting information about the next-generation NVIDIA hardware, saying in a recent earnings call that the company has a 1000W data center GPU in the pipeline.
Dell’s COO, Jeff Clarke, discussed Dell’s engineering advantages and the benefits of NVIDIA’s upcoming hardware at an earnings call on February 29.
He said, “We are excited about the B100 and GB200 products. The chart below shows the chip names for NVIDIA’s next-generation data center GPU and its successor products. NVIDIA currently positions the H100 as its flagship data center GPU and has just launched a second-generation product with faster HBM3e memory called the H200.” We all know that B100 is the Blackwell successor to this chip, so GB200 seems to be the second iteration of this GPU, although it is currently not on NVIDIA’s roadmap (as shown below).
Then, Clark began discussing the thermal performance of these next-generation parts, saying, “You don’t need direct liquid cooling to achieve an energy density of 1,000W per GPU. Some of the products next year will achieve this,“
The current GH200 Grace Hopper CPU+GPU’s TDP has already reached 450W to 1,000W, depending on the configuration, so it would be somewhat surprising to see the next-generation version maintain this number. At the same time, the existing H100 is a 700W GPU, but we don’t know what the power consumption requirements for its successor, the B100, are. NVIDIA seems to be able to increase its power to 1,000W, but we haven’t heard any news about the B100’s power consumption yet.
Currently, we have to wait until March 18 to see what NVIDIA has prepared for the Blackwell data center. As gamers, we might also gather some details from the announcement. Given the company’s high position in the artificial intelligence market, the whole world will be watching this year’s GTC to see what NVIDIA has up its sleeve. Although the release is getting close, we still know very little about Blackwell, only that it will use TSMC’s 3nm process and NVIDIA may use a small chip design for the first time. NVIDIA has also revealed that the demand for these chips will exceed supply in the short term.
NVIDIA has also disclosed plans for the X100 chip, scheduled to be released in 2025, which will expand the product range, including the X40 and GX200 for enterprise use, combining CPU and GPU functions in a Superchip configuration. Similarly, GB200 is expected to follow in B100‘s footsteps and integrate into the Superchip concept.
From NVIDIA’s product roadmap, the AI chip market will be revolutionized again in the next 1-2 years.
Notably, in the AI chip field where NVIDIA holds an absolute position, AMD is one of the few companies with high-end GPUs capable of training and deploying AI, positioned as a reliable alternative for generative AI and large-scale AI systems. One of AMD’s strategies to compete with NVIDIA includes the powerful MI300 series accelerator chips. Currently, AMD is challenging NVIDIA’s H100 dominance with more powerful GPUs and innovative CPU+ GPU platforms.
AMD’s latest MI300 release includes two main series, the MI300X series, a large GPU with the memory bandwidth required for leading generative AI and the training and inference performance needed for large language models; the MI300A series integrates CPU+GPU, based on the latest CDNA 3 architecture and Zen 4 CPU, providing breakthrough performance for HPC and AI workloads. Undoubtedly, the MI300 is not just a new generation of AI accelerator chips but also AMD’s vision for the next generation of high-performance computing.
Besides, NVIDIA also faces competition from self-developed AI chip players.
In February this year, tech giant Meta Platforms confirmed that it plans to deploy its latest self-developed custom chips in its data centers this year, which will work in coordination with other GPU chips to support its AI large model development. Dylan Patel, the founder of the research institution SemiAnalysis, stated that considering Meta’s operational scale if the self-developed chips are successfully deployed on a large scale, it could save hundreds of millions of dollars in energy costs and billions of dollars in chip procurement costs annually.
OpenAI is also seeking billions of dollars in funding to build a network of artificial intelligence chip factories. Foreign media reports that OpenAI is exploring manufacturing its artificial intelligence chips. OpenAI’s website has started recruiting hardware-related talents, with several positions in hardware-software co-design being recruited, and in September last year, OpenAI also recruited Andrew Tulloch, a renowned expert in the field of artificial intelligence compilers, which seems to confirm OpenAI’s investment in self-developed chips.
Not just Meta and OpenAI, according to The Information, as of now, there are over 18 chip design startups for AI large model training and inference globally, including Cerebras, Graphcore, Biren Technology, Moore Threads, and d-Matrix.
Related:
- First Benchmark Results: NVIDIA Blackwell GPU Shines
- Insider at Nvidia: 7-Day Shifts with Work Until 2 AM Nights
- Introducing NVIDIA Blackwell B200: The Ultimate AI GPU
- 23% Q4 Sales Boost Propels NVIDIA to Semiconductor Lead
- NVIDIA Driver 552.12 Causes Game Crashes: Intel to Blame?
- TSMC CoWoS Capacity Skyrockets by 150% Due to NVIDIA!
- NVIDIA CEO: The Next Industrial Revolution Begins
- NVIDIA Graduate Insight: Interviews and Salaries
- NVIDIA Rumor: RTX 50 TITAN to Outshine RTX 5090
- NVIDIA RTX 3050A with AD106 GPU: What You Need to Know
Disclaimer: This article is created by the original author. The content of the article represents their personal opinions. Our reposting is for sharing and discussion purposes only and does not imply our endorsement or agreement. If you have any objections, please contact us through the provided channels.
