Nvidia’s RTX 4090 Crushes AMD’s 7900 XTX in AI Speed, Nearly 50% Faster!

In response to AMD’s claims that its Radeon RX 7900 XTX outperformed Nvidia in DeepSeek AI benchmarks, Nvidia has launched a strong rebuttal. In a recent blog post, Nvidia states that its RTX 5090 and RTX 4090 GPUs substantially outpace AMD’s RDNA 3 flagship, with the RTX 5090 achieving speeds up to 2.2 times faster than the RX 7900 XTX.

Nvidia conducted tests comparing the RTX 5090, RTX 4090, and RX 7900 XTX across three variations of the DeepSeek R1 AI model: Distill Qwen 7b, Llama 8b, and Qwen 32b. With the Qwen LLM at the 32b setting, results indicated that the RTX 5090 was 124% faster, and the RTX 4090 was 47% faster than the RX 7900 XTX. In the Llama 8b tests, the RTX 5090 showed a 106% speed advantage, and the RTX 4090 was also 47% faster. Using the Qwen 7b, the RTX 5090 was 103% quicker, and the RTX 4090 outperformed the RX 7900 XTX by 46%.

The figures released by Nvidia starkly contrast AMD’s earlier benchmarks, which showed the RTX 4090 and RTX 4080 lagging behind the RX 7900 XTX—except in one test where the RTX 4090 was slightly slower. AMD had reported that the RX 7900 XTX was up to 113% and 134% faster than the RTX 4090 and RTX 4080, respectively.

Nvidia’s data paints a very different picture, particularly with the RTX 4090 which they highlight as being considerably faster than the RX 7900 XTX. Even more impressive, according to Nvidia, is that the RTX 5090 more than doubles the performance of the RTX 4090, decisively overshadowing the RX 7900 XTX.

This situation serves as a reminder to view benchmarks released by manufacturers with skepticism. Even if Nvidia’s results are accurate, the conflicting data from AMD and Nvidia show that each company’s internal tests can lead to vastly different conclusions.

DeepSeek R1 is a groundbreaking AI model that has taken the industry by storm, delivering competitive performance against top AI models while requiring 11 times less computing power. The announcement of DeepSeek’s capabilities was primarily responsible for a staggering $600 billion loss in Nvidia’s market capitalization in just one day.