AMD started the spat with benchmarks that showed its RX 7900 XTX graphics card outperforming Nvidia RTX RTX RTX RTX RTX RTX RTX RTX RTX RTX RTx 4090 Super in DeepSeek R1 test. Nvidia, as expected, responded with results that showed the opposite. They do.
Accordingly David McAfee, on Xsaid AMD’s RDNA3 GPU was up to 13% quicker than the RTX 4090. It was also 34% faster than the RTX 4080 super. RX 7900 XTX was the best performing RX 7900 XTX against RTX 4090 when using DeepSeek R1 Qwen 7B. It led by 13%. AMD’s GPU was 11% faster in Distill Llama 7B and Distill Qwen 8B. The RTX 4090 was 4% faster in Distill Qwen32B.
AMD’s GPU had a larger lead over the RTX 4080 Super. The RX 7900 was 34% faster with DeepSeek R1 Distill Qwen 8B. It was 27% faster with Distill Llama 7B. We can’t be certain how Nvidia’s graphics cards were configured for these tests, and as AMD ran the tests, it’s fair that Team Red didn’t go out of its way optimizing things for Team Green hardware.
Nvidia fights Back
After AMD’s benchmarks were published, Nvidia countered with a Blog post and its own results, as reported by Tom’s Hardwareclaims that the RTX is 2.2 times faster than RX 7900 XTX. Nvidia reported a 124% edge using Qwen 32B. The RTX 4090 had a 47% advantage. RTX 5090 was faster with Llama 8B by 106%, while the RTX 4090 was 47% ahead.
The back-and-forth inevitably highlights the importance of treating manufacturer-released benchmarks with caution. Different optimizations, drivers versions, and test conditions can lead to different results.
Even if Nvidia benchmarks are accurate, AMD still wins in one important aspect: price. Its GPU is cheaper than both Nvidia’s high-end offerings.
