NVIDIA GeForce RTX 3070 Ti / RTX 3080 Ti Compute + Renderer Performance Benchmarks

With last week's launch of the GeForce RTX 3080 Ti and this week's launch of the GeForce RTX 3070 Ti there have been plenty of Linux gaming benchmarks on Phoronix. But for those interested in these new RTX 30 Ampere graphics cards for GPU compute or rendering workloads, in this article are many benchmarks on that front compared to various RTX 20 and RTX 30 series graphics cards.

This article features many CUDA, OpenCL, and Vulkan compute benchmarks as well as various rendering workloads both with CUDA and RTX/OptiX. The real-time GPU power consumption was also monitored during testing for the performance-per-Watt. Graphics cards tested this round were from the GeForce RTX 2060 SUPER through the flagship GeForce RTX 3090 graphics card. As always, the cards were freshly re-tested under this current and same Linux software/driver stack.

No AMD Radeon graphics cards were tested for this article due to many of the benchmarks being binary-only and tied to NVIDIA's interfaces while also being blocked on the Radeon RX 6000 / RDNA2 front for the lack of ROCm support yet for running the newest AMD consumer graphics cards fully with their open-source compute stack.