RTX 4090 and RTX 6000 Ada
The United States government has prohibited Nvidia from importing its most powerful gaming GPU from China. As a result, the chipmaker has removed the RTX 4090 from its listings owing to worries over artificial intelligence, but they have left the RTX 6000 Ada in the market.
On the Chinese version of Nvidia’s website, you may still find listings for workstation-grade RTX 6000 Ada systems. According to the most recent set of regulations governing commerce in the United States, the company’s most powerful graphics card, the RTX 6000 Ada, is among the goods that are prohibited from being sent to China, along with certain other countries.
In point of fact, the RTX 6000 Ada is a very good fit for the training of artificial intelligence due to the fact that it has a tiny blower cooler and 48 GB of memory. In contrast, VideoCardz found that Nvidia and its partners no longer offer the GeForce RTX 4090 graphics card in China, and the company has removed any mention of the consumer product from its Chinese website.
One of the best graphics cards available, the Nvidia GeForce RTX 4090, is based on the AD102 graphics processing unit and has a total processing performance score of 5,280 (based on its FP8 Tensor FLOPS performance of 660 TFLOPS). This score makes it an export-licensable item because its TPP is higher than 4,800. Nvidia and its partners are now need to seek an export license from the United States Department of Commerce in order to send GeForce RTX 4090 devices to China.
This requirement was previously not in place. Since such licensing applications are reviewed with the expectation that they will be denied, it appears that Nvidia would prefer not to sell its GeForce RTX 4090 in China.
Given the aforementioned information, it is puzzling that the business continues to offer its RTX 6000 Ada Generation graphics board for use by professionals in China. This solution uses the AD102 graphics processing unit, which has 18,176 CUDA cores active, giving it a total processing performance score of 5,828 (based on its 728.5 FP8 TFLOPS performance without sparsity). Because this graphics card has 48 gigabytes of RAM, it is actually a better choice for artificial intelligence training and inference than the GeForce RTX 4090, which is designed for consumers.
What is perhaps more important is that Nvidia’s RTX 6000 Ada Generation comes with a compact blower cooling system. This makes it easy to use in data center environments and makes it useful both for AI training / inference jobs as well as high-performance computing, as it supports FP64 without any constraints.
RTX 6000 Ada Generation and RTX 6000 Ada specifications:
| GPU | RTX 6000 Ada Generation | GeForce RTX 4090 |
|---|---|---|
| Architecture | GPU | Ada Lovelace | AD102 | Ada Lovelace | AD102 |
| Memory | 48 GB GDDR6 w/ ECC | 24 GB GDDR6X |
| Total Processing Power (FP16/BF16) | 5,828 | 5,280 |
| Performance Density | 9.57 | 8.66 |
| Memory Bandwidth | 960 GB/s | 1008 GB/s |
| CUDA Cores | 18,176 | 16,384 |
| INT8 I FP8 Tensor | 728.5 I 1457 TFLOPS | 660 I 1320 TFLOPS |
| BF16 I FP16 Tensor | 91.06 TFLOPS | 82.58 TFLOPS |
| FP32 | 91.06 TFLOPS | 82.58 TFLOPS |
| FP64 | 1423 GFLOPS | 1290 GFLOPS |
| RT Core | Yes | Yes |
| L2 Cache | 96 MB | 72 MB |
| Power | 300W | 450W |
| Form Factor | 2-slot FHFL | 3.5-slot |
| Interface | PCle Gen4 x16: 64 GB/s | PCle Gen4 x16: 64 GB/s |
Why Nvidia still has their graphics card, the RTX 6000 Ada Generation, listed on its Chinese website is unknown. It’s possible that the business plans to file for an export license in order to continue selling this $6,800 product in the People’s Republic, or it’s possible that its partners already have a large supply of those cards in China and can keep selling stock for a while.
[…] to geopolitics, Nvidia stopped selling the conventional RTX 4090 and other AI/HPC-focused GPUs to China, prompting the new consumer RTX 40 series GPU. The new […]