NVIDIA GeForce GTX 980 Through GeForce RTX 5080/5090 GPU Compute Performance
To put it simply, NVIDIA GPU compute performance and power efficiency has nicely evolved over the past decade when comparing Maxwell through Blackwell GPUs.
When taking the geometric mean of all the benchmarks that were able to successfully run on all of the tested GPUs, the geometric mean provides this nice representation. From the GeForce GTX 980 to GeForce RTX 5080 is 8.6x the performance while against the GeForce RTX 5090 flagship is 14.3x the performance over the past decade.
The GPU power efficiency has evolved nicely too overall but on a performance-per-Watt basis the newest GeForce RTX 50 series are roughly aligned in power efficiency with the prior GeForce RTX 40 parts.
For those curious about the GPU thermals with these mostly reference stock / Founder's Edition cards, here is that breakdown across all of the workloads tested. The NVIDIA GeForce RTX 5080 and RTX 5090 Founder's Edition thermal performance continues to be very compelling especially with a dual slot versus three slot design with the prior RTX 30/40 top-tier graphics cards.
Those wanting to dig into even more NVIDIA Linux GPU compute numbers from this multi-generational comparison can find all of my data via this result file.
If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.
