According to TechSpot , after AMD released benchmark results showing that the Radeon RX 7900 XTX card outperformed Nvidia's graphics cards, the 'green team' quickly 'responded' with its own benchmark results, confirming the superior power of the GeForce RTX 5090 graphics card when running DeepSeek AI models.
Nvidia's powerful RTX 5090 graphics card
PHOTO: TECH4GAMERS SCREENSHOT
Nvidia Claims RTX 5090 Is More Powerful at AI Than AMD's Radeon Cards
According to Nvidia, the RTX 5090 is up to 2.2 times faster than the RX 7900 XTX when running DeepSeek R1 AI models. The company conducted detailed benchmark tests with three versions of the DeepSeek R1 AI model: Distill Qwen 7b, Llama 8b, and Qwen 32b. The results showed that the RTX 5090 significantly outperformed its AMD rival in all tests.
Specifically, with Qwen LLM 32b, the RTX 5090 is 124% faster, while the previous generation RTX 4090 is also 47% faster. Similarly, with Llama 8b, the RTX 5090 is 106% faster, the RTX 4090 is 47% faster. Even in the Qwen 7b test, the RTX 5090 is still 103% faster, the RTX 4090 is 46% faster.
This result is in stark contrast to AMD's previous benchmarks, which showed the RX 7900 XTX outperforming Nvidia's RTX 4090 and 4080 in many cases.
Nvidia's DeepSeek AI Models Performance Is Absolutely Superior
PHOTO: TECHSPOT SCREENSHOT
Nvidia also claims its GeForce RTX 50 Series GPUs are the only option to run DeepSeek's 'distilled' models faster than any other option on the PC market.
The reason for this superiority, according to Nvidia, is that the DeepSeek R1 model line is designed to simulate the human problem-solving process, allocating more computational resources to 'thinking' and 'reflecting.' This approach, called "test-time scaling," allows the model to automatically allocate computational resources during inference to solve problems more efficiently.
Nvidia's RTX 50 Series GPUs, with dedicated fifth-generation Tensor Cores, are built on the same Blackwell GPU architecture used in data centers. This architecture allows RTX to fully accelerate DeepSeek models, delivering the highest inference performance on the PC.
Nvidia also highlighted its RTX AI platform, which enables DeepSeek-R1 to be available on more than 100 million Nvidia RTX AI PCs worldwide.
Ultimately, Nvidia argues that high-performance RTX GPUs ensure AI capabilities are always accessible, even without an internet connection, delivering low latency and enhanced privacy.
With these impressive benchmark results, Nvidia continues to assert its leadership in the AI and GPU fields. However, the competition between Nvidia and AMD is still very fierce and we can expect more 'wars' and new benchmark results in the future.
Source: https://thanhnien.vn/nvidia-phan-phao-sau-khi-bi-amd-tuyen-bo-vuot-mat-185250205090519067.htm
Comment (0)