NVIDIA GeForce RTX 3080 Ti Linux Performance

NVIDIA this week at Computex Taipei announced the RTX 3070 Ti and RTX 3080 Ti while today the actual review embargo expires on the GeForce RTX 3080 Ti as this new flagship gaming graphics card. Here are our initial benchmarks of the NVIDIA GeForce RTX 3080 Ti under Linux.

NVIDIA has stated that the GeForce RTX 3080 Ti should be about 1.5x faster than the prior generation RTX 2080 Ti or 2x the performance of the GeForce GTX 1080 Ti for rasterized games. The RTX 3080 Ti is rated for 34 shader TFLOPS, 67 RT TFLOPS, and 273 Tensor TFLOPS. The GeForce RTX 3080 Ti features 10,240 CUDA cores and 12GB of GDDR6X video memory. The 10,240 CUDA cores is up from the 8,704 CUDA cores found on the RTX 3080 and there is also 2GB of additional vRAM compared to that non-Ti model from last year.

The retail availability of the RTX 3080 Ti is set for tomorrow, 3 June, though actual availability still is likely to be tight. Pricing on the RTX 3080 Ti starts out at $1199 USD.

Due to only finding out about the cards being sent out last week and the time it takes for testing plus to freshly re-test all comparison cards on the very latest Linux drivers, this is a more brief review than normal. In this article is looking at the GeForce RTX 3080 Ti performance for native Linux games and Steam Play titles plus some Vulkan ray-tracing benchmarks. The GPU compute with OptiX, CUDA, and other workloads will be coming in a follow-up article in the coming days.

For this launch-day testing on a Ryzen 9 5900X system with Ubuntu 21.04 was using NVIDIA 465.31 driver for the GeForce cards tested while the AMD cards were tested using Linux 5.13 Git with Mesa 21.2-devel from the Oibaf PPA. When running the Vulkan ray-tracing benchmarks, the Radeon Software for Linux 21.30 PRO packaged Vulkan driver was used as the only AMD Vulkan driver option currently having ray-tracing support.