AMD’s Radeon RX 7900 XTX Surpasses NVIDIA’s GeForce RTX 4090 in DeepSeek’s AI Inference Benchmarks: Learn How to Execute R1 on Your Local AMD Systems

AMD has made impressive strides with its Radeon RX 7900 XTX, as it breezes through running the DeepSeek R1 AI model, even surpassing NVIDIA’s GeForce RTX 4090 in inference benchmarks.

AMD Raises the Bar: DeepSeek’s R1 Models Get a Boost

DeepSeek’s latest AI model is causing quite a stir in the tech world. While there’s curiosity about the hefty computing power used during its development, it turns out that even the average user can achieve remarkable results using AMD’s “RDNA 3” Radeon RX 7900 XTX GPU. AMD put the spotlight on this by showcasing DeepSeek’s R1 inference benchmarks, where their flagship GPU outshines NVIDIA’s competing model across various tests.

A recent tweet from David McAfee, AMD’s Corporate Vice President and General Manager, shows DeepSeek’s robust performance on the Radeon 7900 XTX and links to resources for running the models on Radeon GPUs and Ryzen AI APUs.

For those interested in leveraging consumer GPUs for AI tasks, this guide is invaluable. Local model execution has become particularly appealing, not just for performance but also for privacy reasons, which many users of DeepSeek’s AI models deeply appreciate. AMD has crafted a detailed manual on running DeepSeek R1 models on their GPUs:

1. Update to at least the 25.1.1 Optional Adrenalin driver.
2. Download LM Studio 0.3.8 or a newer version from lmstudio.ai/ryzenai.
3. Install LM Studio and bypass the onboarding screen.
4. Head to the discover tab.
5. Pick your desired DeepSeek R1 Distill; Qwen 1.5B is a recommended start for its speed. Larger distills improve reasoning skills significantly and are all highly competent.
6. Ensure “Q4 K M” quantization is selected on the right and hit “Download”.
7. After downloading, return to the chat tab, select your DeepSeek R1 distill from the menu, and check “manually select parameters”.
8. Max out the GPU offload layers by sliding the control all the way.
9. Load the model.
10. Begin interacting with a reasoning model, fully operational on your AMD local setup!

If you find these steps a tad tricky, AMD has thoughtfully released a YouTube tutorial breaking down each part. Check it out to smoothly run DeepSeek’s LLMs on your local AMD systems, safeguarding your data in the process.

With upcoming releases from NVIDIA and AMD, we anticipate a leap forward in inference capabilities, especially with the integration of dedicated AI engines to support these demanding tasks.