AMD Radeon RX 7900 XTX Outperforms NVIDIA GeForce RTX 4090 in DeepSeek AI Inference Benchmarks: How to Run R1 on Your Local AMD System

AMD’s Radeon RX 7900 XTX has emerged as a powerful contender in the arena of AI inference, outperforming NVIDIA’s GeForce RTX 4090 in benchmark tests using DeepSeek’s R1 AI model. This development signals a significant shift in the computing landscape, making advanced AI capabilities more accessible to everyday users.

DeepSeek’s R1 AI Model: AMD’s Competitive Edge

The release of DeepSeek’s latest AI model has generated considerable excitement within the tech community. For those questioning the computational demands of running this model, the AMD Radeon RX 7900 XTX GPU, utilizing its “RDNA 3″architecture, proves sufficient for the task. Recent benchmarks demonstrate a marked performance advantage of AMD’s flagship RX 7000 series over its NVIDIA counterpart across various models.

DeepSeek performing very well on @AMDRadeon 7900 XTX. Learn how to run on Radeon GPUs and Ryzen AI APUs here: https://t.co/FVLDLJ18Ov pic.twitter.com/5OKEkyJjh3

— David McAfee (@McAfeeDavid_AMD) January 29, 2025

Advantages of Local AI Processing with AMD GPUs

For individuals leveraging consumer GPUs for AI tasks, the AMD Radeon series presents a compelling option, delivering excellent performance relative to price compared to traditional AI accelerators. Running AI models locally not only enhances performance but also addresses significant privacy concerns associated with utilizing cloud services, particularly regarding data handled by DeepSeek’s AI solutions.

Step-by-Step Guide to Running DeepSeek R1 on Your AMD Hardware

Getting started with DeepSeek’s R1 model is straightforward. Follow these steps for optimal performance:

Step 1: Ensure that you are using the Adrenalin driver version 25.1.1 Optional or later.

Step 2: Download LM Studio version 0.3.8 or higher from lmstudio.ai/ryzenai.

Step 3: Install LM Studio and bypass the onboarding screen.

Step 4: Click on the “discover”tab.

Step 5: Select your desired DeepSeek R1 distill. Smaller models like Qwen 1.5B are recommended for speed, while larger models offer enhanced reasoning capabilities.

Step 6: On the right-hand side, select the “Q4 K M” quantization and click “Download.”

Step 7: After downloading, return to the chat tab, choose the DeepSeek R1 distill from the dropdown, and ensure “manually select parameters” is checked.

Step 8: In the GPU offload layers, move the slider to the maximum setting.

Step 9: Click on “model load.”

Step 10: Begin interacting with your locally running reasoning model!

Additional Resources and Future Trends

If you encounter difficulties, AMD has also created a detailed tutorial on YouTube that walks through the setup process for running DeepSeek’s large language models on local AMD machines. This is particularly beneficial for users concerned about data security and privacy.

As we look ahead, the anticipated launch of new GPUs from both AMD and NVIDIA promises substantial advancements in inferencing capabilities, thanks in part to dedicated AI engines designed to handle demanding workloads more efficiently.