NVIDIA RTX 4000 SFF Ada professional graphics card test: AI inference artifact, power consumption is only 70W

NVIDIA has released many GPUs based on the Ada architecture. In addition to GeForce game graphics cards, it also includes a large number of professional graphics cards. Compared with traditional game graphics cards, the biggest feature of these professional graphics cards is the use of ECC memory, which improves the system The stability of the operation and the energy consumption ratio are also more outstanding. In addition, there are special optimizations for specialized productivity software. The RTX 4000 SFF Ada professional graphics card, currently priced at 9,999 yuan, has been launched in the domestic market, and its power consumption is much lower than traditional game graphics cards.

Currently, some media have obtained the RTX 4000 SFF Ada professional graphics card and conducted tests. They found that the graphics computing performance of the RTX 4000 SFF Ada professional graphics card is similar to that of the RTX 3060 Ti. For example, TimeSpy’s score is 10603 points, which is higher than that of the RTX 3060 Ti. The score of 11518 is about 10% lower, and the actual game is also about 6-8 frames lower than the RTX 3060 Ti, which is much higher than the RTX 3060. The most important feature of this professional-grade card is its power consumption, which averages only 71W, which is far lower than the RTX 3060 Ti's 199W.

RTX 4000 SFF Ada professional graphics card is equipped with AD104 GPU, which has 6144 CUDA cores and 20GB capacity of GDDR6 memory. The main focus is AI performance. The computing power of INT8 reaches 306.8 TOPS. It can be said to be a model for AI training. Professional graphics card. Of course, if you want to spend 9,999 yuan to buy a computer that is only used for games, it may not be cost-effective. But you can certainly do that if you want.

The above is the detailed content of NVIDIA RTX 4000 SFF Ada professional graphics card test: AI inference artifact, power consumption is only 70W. For more information, please follow other related articles on the PHP Chinese website!