GEFORCE RTX 5090 is best for all video cards for PC for launching DeepSeek AI model, NVIDIA said

NVIDIA positions its new GEFORCE RTX graphics cards of the 50th series not only as excellent games for games, but also as powerful accelerators for working with AI. The company did not stand aside from the hype around the Chinese II-model Deepseek and the other day stated that its new video cards allow “launching the DEEPSEEK language models faster than anything in the PC market”.

Image source: NVIDIA

This week, Nvidia survived the largest one -day drop in market capitalization among American companies in history. Many associate this with the rapid growth of the popularity of the R1 AI-model from the Chinese startup Deepseek. This productivity algorithm is comparable to OpenAi GPT-4O, but significantly less funds were spent on its training and less advanced equipment was used.

Deepseek algorithms are trained using the NVIDIA H800 graphic accelerators, while more productive H100 and H200 are available to American developers actively used by companies such as Openai, XII and others. Nevertheless, the success of the Chinese startup proves that access to the most advanced NVIDIA accelerators is not always necessary to achieve high results in AI.

However, in its publication, NVIDIA notes that its new GEFORCE RTX graphics cards of the 50th series are the best option for launching DeepSeek AI models, providing maximum performance. It is curious that earlier this week AMD said that her Radeon RX 7900 XT is faster than the GeForce RTX 4090, but NVIDIA data refut this.

Note that other technological companies also seek to benefit from the unexpected take -off of the popularity of Deepseek. The reasoning AI-model R1 is already available in the AWS cloud and on the Microsoft Azure platform. In parallel with this, Microsoft and Openai study the likelihood that Deepseek used Openai data to teach her AI-Models, which the media previously wrote.

NVIDIA also showed the work of DeepSeek in its NIM microservice microservice package for simple AI deployment.