AMD RX 7900 XTX Outpaces Nvidia RTX 4090 in AI Tasks With DeepSeek

Expert Verified By

AMD RDNA 3 Options Offer Impressive Results in DeepSeek.

Story Highlight
  • AMD is promoting the RX 7000 series for AI tasks, especially the RX 7900 XTX.
  • The RX 7900 XTX outperforms the RTX 4080 SUPER and comes close to the RTX 4090.
  • AMD’s last-gen flagship GPU is powerful enough for demanding AI DeepSeek R1 models.

AMD recently released comprehensive instructions on configuring Ryzen AI CPUs and Radeon RX graphics cards to run DeepSeek R1 distilled reasoning models. Moving on to discrete graphics cards, AMD is currently solely promoting its Radeon RX 7000 series because the RDNA 3 graphics architecture incorporates AI accelerators.

Why it matters: The RX 7900 XTX’s ability to handle complex AI models efficiently means more options for users, potentially lowering the barrier for those working with AI and machine learning tasks.

While all SKUs with 12 GB to 20 GB of RAM, such as the RX 7600 XT, RX 7700 XT, RX 7800 XT, RX 7900 GRE, and RX 7900 XT, are advised till DeepSeek-R1-Distill-Qwen-14B, the flagship Radeon RX 7900 XTX is advised for DeepSeek-R1-Distill-Qwen-32B distill.

With DeepSeek-R1-Distill-Qwen-7B, DeepSeek-R1-Distill-Llama-8B, and DeepSeek-R1-Distill-Qwen-14B, the RX 7900 XTX exhibits up to 34%, 27%, and 22% better performance, respectively, than the RTX 4080 SUPER. 

Source: AMD

The major comparison between the RX 7900 XTX and the GeForce RTX 4090, which has 24 GB of RAM, comes next. Three of the four tests demonstrate that the RX 7900 XTX performs better than the competition, with DeepSeek-R1-Distill-Qwen-7B showing up to 13% higher performance, DeepSeek-R1-Distill-Llama-8B showing up to 11% higher performance, and DeepSeek-R1-Distill-Qwen-14B showing up to 2% higher performance.

The bigger DeepSeek-R1-Distill-Qwen-32B variant only has a 4% performance gap with the RTX 4090.

DeepSeek’s latest AI model has swept the industry, and although many are curious about the amount of processing power required to train it, it appears that the typical consumer can get by with a Radeon RX 7900 XTX GPU.

Was our article helpful? 👨‍💻

Thank you! Please share your positive feedback. 🔋

How could we improve this post? Please Help us. 😔

Gear Up For Latest News

Get exclusive gaming & tech news before it drops. Sign up today!

Latest News

Join Our Community

104,000FansLike
32,122FollowersFollow

Trending

Unreal Engine 5 Issues Might Go Away With RTX 60 Series Launch, Says Developer

Massive Damage's Bryan Heemskerk argues that Unreal Engine 5 launched too early for its time, but RTX 6000 will most likely fix its issues.

DC’s Wonder Woman Game Cancelled After 4 Years; Studio Also Shut Down

DC's much anticipated Wonder Woman project has been cancelled, with the company shutting down developer Monolith Productions.

Sucker Punch Abandoned Two Titles Before Settling on Ghost of Tsushima

Ghost of Tsushima developer Brian Fleming has revealed that he scrapped two projects before finally starting work on Ghost of Tsushima.

Hellblade 3: Ninja Theory to Revamp Melee Combat and Gameplay Design In Next Game

Job listing suggests Hellblade 3 will overhaul its combat, moving away from linear battles toward more player freedom.

Apex Legends Voice Actors Forced To Train Generative AI That Would Later Replace Them

Apex Legends French voice actor reveals that EA directly sent them agreements that their voices would be used to train generative AI.