The upcoming CUDA 12.9 toolkit, which will introduce first-class support for FP8 on consumer GPUs (RTX 5090). You’ll need this driver as a prerequisite. Have you run into any specific issues with the latest CUDA driver? Let me know in the comments, and I’ll update the post with community solutions.
NVIDIA’s CUDA driver is the silent workhorse behind every AI training run, 3D render, and scientific simulation. It doesn’t get the glory of new GPU hardware, but without the right driver, your expensive RTX 4090 or H100 is just a fancy space heater. cuda driver release news
If you are on a stable, headless Linux server running CUDA 11.8 code that works perfectly, for the "production branch" release. There is no security vulnerability being patched in this update. The upcoming CUDA 12
| Workload | Old Driver (545.x) | New Driver (570.x) | Difference | | :--- | :--- | :--- | :--- | | | 18.2 it/s | 18.5 it/s | +1.6% | | PyTorch Matrix Multiply (4096x4096) | 2.31 ms | 2.28 ms | +1.3% | | OctaneBench 2024 (RTX) | 425 | 428 | +0.7% | Let me know in the comments, and I’ll
is available for immediate download! To get the latest version of the requested update simply download and run Driver Reviver.
BONUS: Driver Reviver will also identify all of your other out of date drivers. To quickly and easily update them, simply choose to download and install all out of date items and you will immediately restore maximum performance and stability to your PC!