AMD’s Radeon RX 7900 XTX truly outshines its competition, delivering extraordinary performance with the DeepSeek R1 AI model, and outperforming NVIDIA’s GeForce RTX 4090 in inference benchmarks.
Let’s delve into why AMD’s support for DeepSeek’s R1 LLM models is making waves. In the gaming and tech world, DeepSeek’s newest AI model is the talk of the town. While some might speculate about the kind of computing power it takes to run this model, AMD’s RDNA 3 Radeon RX 7900 XTX is a revelation, offering more than enough performance for the average user. In benchmarks, AMD’s flagship RX 7000 series GPU has shown itself to be a superior choice over NVIDIA’s offerings, across several models.
AMD has quickly responded to the excitement, providing instructions for running DeepSeek on their GPUs. David McAfee from AMD recently highlighted their success with a tweet, revealing how well the RX 7900 XTX handles DeepSeek tasks, and pointing users to resources for setting up their Radeon GPUs and Ryzen AI APUs.
For users looking to leverage their consumer GPUs for AI workloads, AMD is proving to be a cost-effective option, especially as AI processing accelerators can become expensive. Running these models locally also means users can better protect their privacy—an understandable concern with AI technology advancing so rapidly. Thankfully, AMD has put together a comprehensive guide on setting up DeepSeek R1 distillations on their hardware. Here’s a simplified run-through of the process:
1. First off, ensure your driver version is 25.1.1 Optional or newer.
2. Next, grab the LM Studio 0.3.8 or a more recent version from lmstudio.ai/ryzenai.
3. Install the LM Studio and bypass the onboarding screen.
4. Navigate to the discover tab.
5. Select the appropriate DeepSeek R1 distillation. For speedy performance, start with a smaller distill like the Qwen 1.5B; larger distills provide better reasoning capabilities.
6. Ensure the “Q4 K M” quantization option is checked, then click “Download”.
7. Once the download is complete, return to the chat tab, choose the DeepSeek R1 distill from the menu, and tick “manually select parameters”.
8. Maximize the GPU offload layers by moving the slider to its end.
9. Hit model load and you’re good to go!
10. Engage with your reasoning model, all powered by your local AMD setup.
If you find these steps challenging, don’t worry—AMD has a YouTube tutorial that breaks down each step clearly. It’s worth watching to ensure your data remains safe while you run DeepSeek’s LLMs on your AMD hardware. As AMD and NVIDIA gear up to release their next-gen GPUs, we anticipate a surge in inferencing power, thanks in part to the inclusion of dedicated AI engines designed to handle these types of demanding computing tasks.
