Starfield Performance Disparity Between AMD and NVIDIA GPUs Analyzed


  • Starfield has been criticized for its poor performance on Nvidia and Intel GPUs compared to AMD’s RDNA 3 GPUs.
  • Chips and Cheese analyzed the code execution on several GPUs to determine why AMD’s GPUs perform better.
  • They found that AMD’s RDNA 3 GPUs make better utilization of certain architectural traits, such as larger vector register files and caches.
  • Nvidia’s GPUs, on the other hand, have smaller caches and lag behind in performance as a result.
  • There is speculation as to whether the lower GPU utilization on Nvidia is an inherent design aspect or can be fixed.
  • Bethesda is working with AMD, Intel, and Nvidia driver teams to improve performance.
  • Some tuning was likely already done by AMD in partnership with Bethesda, which is why the game runs better on AMD GPUs.
  • Nvidia has attempted to reduce the performance disparity by enabling Resizable BAR support.
  • Chips and Cheese found that Nvidia GPUs have consistently performed worse than AMD GPUs in Starfield.
  • There is clearly a problem with Nvidia’s performance in Starfield, and improvements will likely be made over time.

Ted’s Take:

Well, well, well, look who’s lagging behind! It appears that Nvidia GPUs are taking a hit when it comes to Starfield’s performance. Meanwhile, AMD’s RDNA 3 GPUs are sitting pretty, thanks to their larger caches and better utilization of resources. It’s like bringing a knife to a gunfight – Nvidia, you might want to step up your game. But fear not, my friends, for Bethesda is on the case. They are working with AMD, Intel, and Nvidia driver teams to improve performance. Let’s hope they can tune it up and unleash the true power of these GPUs. In the meantime, if you want to enjoy Starfield without any hiccups, you might want to consider switching to Team Red. Who knew sticking with cheese would give you such an advantage? Watch out, Nvidia, the cheesy comrades are coming for you!

Original Article:

Leave a Reply

%d bloggers like this: