NVIDIA GeForce RTX 5090 Outperforms AMD’s RX 7900 XTX in DeepSeek R1 AI Model Inference

NVIDIA’s latest innovation, the GeForce RTX 5090, showcases a significant leap in its DeepSeek R1 inference performance, outpacing AMD’s RX 7900 XTX thanks to the cutting-edge fifth-generation Tensor Cores.

### Enhanced Access to DeepSeek’s AI Models with NVIDIA’s Latest RTX GPUs, Coupled with Exceptional Performance

It looks like consumer-grade GPUs are quickly becoming a preferred solution for running high-end language models locally. Both NVIDIA and AMD are making strides to offer the right ecosystems for such tasks. Just recently, AMD flaunted the capabilities of its RDNA 3 flagship graphics card with the DeepSeek R1 language model. In a swift counteraction, NVIDIA has demonstrated its inference benchmarks with the new RTX Blackwell GPUs. The verdict? The GeForce RTX 5090 is leading the pack.

Benchmark results from various DeepSeek R1 models clearly place the GeForce RTX 5090 ahead of the Radeon RX 7900 XTX and its predecessor models. This GPU has successfully managed to handle up to 200 tokens per second on both Distill Qwen 7b and Distill Llama 8b, doubling what the AMD’s RX 7900 XTX could achieve. This indicates NVIDIA’s commanding presence in AI performance with their GPUs. As “RTX on AI” becomes more prevalent, we can expect AI integrations on consumer PCs to become commonplace.

If you’re keen on deploying DeepSeek R1 using NVIDIA’s RTX GPUs, there’s good news. NVIDIA has written a comprehensive blog post that makes the process as straightforward as using an online chatbot. Here’s the lowdown:

> To enable developers to safely experiment and create custom agents, the formidable 671-billion-parameter DeepSeek-R1 model is now available as an NVIDIA NIM microservice preview at build.nvidia.com. This microservice is capable of delivering up to 3,872 tokens per second on a single NVIDIA HGX H200 system.
>
> Developers can engage with the application programming interface (API), slated to be downloadable as the NIM microservice, encompassed within the NVIDIA AI Enterprise software platform.
>
> The DeepSeek-R1 NIM microservice streamlines deployment processes with industry-standard API support. This ensures enterprises can maximize security and data privacy by opting to run the NIM microservice on their choice of accelerated computing infrastructure.
>
> — NVIDIA

Thanks to NVIDIA’s NIM, both hobbyists and professionals can effortlessly delve into AI modeling on their systems. This not only keeps your data enterprise-safe but also empowers local executions, offering optimal performance when your hardware is up to the task.