NVIDIA's tweaked H20 AI GPU enters mass production in Q2 2024, destined for China

NVIDIA reportedly plans mass production of its new H20 AI GPU designed exclusively for China, in Q2 2024, meeting the updated US sanctions.

2 minutes & 16 seconds read time

NVIDIA has been forced to tweak its AI GPUs to meet constantly updated US sanctions, with its new H20 AI GPU set to enter mass production in Q2 2024, ready for the Chinese market.

NVIDIA's tweaked H20 AI GPU enters mass production in Q2 2024, destined for China 305

We reported -- last year, now that we're in 2024 -- that NVIDIA was in the process of making new AI GPUs for the Chinese market. These AI GPUs included the HGX H20, L20 PCIe, and L2 PCIe AI GPUs, which acted as alternatives to the AI GPUs that NVIDIA has on the market already.

The new H20 AI GPU is a cut-down version of the H100, with 96GB of HBM3 memory with up to 4.0TB/sec of memory bandwidth. There are 296 TFLOPs of compute power, and the H100 AI GPU die with a performance density of 2.9 TFLOPs/die, compared to the 19.4 TFLOPs/die of the regular H100 AI GPU.

NVIDIA's current H100 SXM is 6.68x faster than the upcoming H20 SXM, according to NVIDIA's own detailed specifications of the upcoming H20 AI GPU. The new H20 AI GPU has a 400W TDP, and is capable of being installed into an 8-way configuration inside of an HGX solution. NVIDIA keeps the 900GB/sec NVLink connectivity on its new H20 AI GPU, with 7-way MIG (Multi-Instance GPU) functionality.

NVIDIA was meant to have its new H20 AI GPU ready to go in late 2023, but it was delayed... the US government has been playing with its sanctions. NVIDIA has been jumping through (AI) hoops to get things right for the Chinese (and other) markets that are affected by the Biden administration and its constantly evolving US sanctions.

We've seen NVIDIA's flagship GeForce RTX 4090 banned in China, with prices for the RTX 4090 skyrocketing because of it. There have been some Chinese companies ripping apart the RTX 4090 and retrofitting them into AI accelerators. Then NVIDIA announced -- and now has finally released -- its edited GeForce RTX 4090D (the "D" is important there) for the Chinese market, a slightly cut-down RTX 4090 exclusive for China.

It seems that NVIDIA is now back on track, with mass production of its H20 AI GPU for the Chinese market ready to start in Q2 2024, meaning we have a few more months to wait before the H20 is out of the oven.

NVIDIA's new H20 AI GPU features 96GB of HBM3 memory with 4.0TB/sec of memory bandwidth, 296 TFLOPs of FP8 Tensor Core performance, 148 TFLOPs of FL16 Tensor Core performance, and 74 TFLOPs of TF32 Tensor Core performance. There are 44 TFLOPs of FP32 compute performance, 1 TFLOPs of FP64 performance, and no RT cores on the H20 AI GPU.

It will fall onto the PCIe Gen5 x16 standard, with 128GB/sec of bandwidth, and 900GB/sec of bandwidth across its NVLink connector.

Buy at Amazon

NVIDIA H100 80 GB Graphic Card PCIe HBM2e (NVIDIA H100 80 GB)

TodayYesterday7 days ago30 days ago
Buy at Newegg
* Prices last scanned on 2/27/2024 at 8:37 pm CST - prices may not be accurate, click links above for the latest price. We may earn an affiliate commission.

Anthony joined the TweakTown team in 2010 and has since reviewed 100s of graphics cards. Anthony is a long time PC enthusiast with a passion of hate for games built around consoles. FPS gaming since the pre-Quake days, where you were insulted if you used a mouse to aim, he has been addicted to gaming and hardware ever since. Working in IT retail for 10 years gave him great experience with custom-built PCs. His addiction to GPU tech is unwavering and has recently taken a keen interest in artificial intelligence (AI) hardware.

Newsletter Subscription

Related Tags