NVIDIA has finally started to take orders for its 'China-compliant" H20 AI GPUs as the firm tries to "re-capture" the confidence of Chinese clients in its products.
NVIDIA Takes The Leap Of Faith By Re-Entering Chinese Markets, This Time With Further Cut-Down Offerings
While 2023 has been an amazing year for Team Green's performance in the data center segment, the firm has faced setbacks in the form of US policies, which have prompted them to take measures unfavorable to their economic growth in regions such as China.
After the US government imposed bans on NVIDIA's H800 and A800 AI GPUs, the company decided to introduce further "cut-down variants that fully comply with US export policies to maintain its presence in the Chinese markets since they did play a crucial role in the financial progress NVIDIA saw in FY23. The firm unveiled a new range of AI GPUs, which included the H20 SXM, PCIe L20, and PCIe L2.
Reuters has disclosed that NVIDIA has "re-entered" the Chinese markets through its H20 AI GPU after opening its orders for potential clients. In terms of pricing, it is reported that the GPU is selling somewhere between US$12,000 and $15,000, which places it in a competitive spot with the likes of Huawei's 910B AI accelerator, which is being sold for around $16,000.
The price point is certainly well below what we saw with the firm's popular H100s, which were said to have retailed as high as $70,000 in Chinese markets. But with that, the performance degradation is also huge, which is why NVIDIA had difficulty attracting consumer interest with the new AI solutions.
Nvidia has started taking orders for its new China AI chips, with the H20 priced at US$12,000 - $15,000, on par with a rival product from China's Huawei, Reuters reports. Huawei's 910B is being sold for about 120,000 yuan (US$16,900). $NVDA #Huawei #semiconductors…
— Dan Nystedt (@dnystedt) February 1, 2024
In terms of specifications, the H20 AI GPU is a cut-down variant of the H100 GPU that features 96 GB memory capacities operating at up to 4.0 Tb/s, 296 TFLOPs Compute power, & using the GH100 die with a performance density of 2.9 TFLOPs/die versus H100's 19.4. The H100 SXM is 6.68x faster than the H20 SXM as per the listed table, but those are FP16 Tensor Core FLOPs (with Sparsity) and not the INT8 or FP8 FLOPs. The GPU has a 400W TDP and features 8-way configurations in an HGX solution. It retains the 900 GB/s NVLINK connection & also offers 7-Way MIG (Multi-Instance GPU) functionality.
The process of initiating the sales of NVIDIA's H20 AI GPUs is indeed an optimistic one since it shows that the firm hasn't given up yet. However, this time, Team Green faces much tougher competition since offerings from Huawei have been much more mature than they were previously.
News Source: Reuters
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
