For example, I use a freeware - and I want to stress, FREE - tool CapframeX https://github.com/CXWorld/CapFrameX that monitors things like FPS, CPU and GPU utilization, CPU and GPU temp, CPU and GPU clock, RAM and VRAM usage, FPS frametimes, etc. So I know from actual experience and testing which driver performs best and is stable.
Also, I’m specifically talking about RTX 50XX series GPUs here. When I had an RTX 4090, my preferred driver was 566.14, or in some cases 566.36. So the “best” driver can depend on the GPU generation and the specific driver branch.
Yesterday, I decided to compare the performance of the two latest NVIDIA drivers 596.94 vs 610.47 vs the older 577.00 driver. The reason is simple: for quite some time, something about the newer drivers hasn’t felt quite right to me.
Each driver was tested three times, with each individual flight taking approximately 22 minutes. This means I spent about 1 hour and 6 minutes per driver, resulting in a total of just over three hours of testing overall.
It may not seem like a huge sample size, but I believe it is still sufficient.Test 1.
Test 2
Test 3
These results align with what I have been observing for a long time, namely that something feels off with the newer NVIDIA drivers.
There is a clear and noticeable performance advantage for driver 577.00. Both 596.94 and 610.47 perform very similarly to each other, but both are noticeably worse than 577.00, which is not surprising based on my previous experience.
What further supports my conclusions is NVIDIA’s own changelog, where they acknowledge a potential issue: “Prefer Maximum Performance” Power Management Mode may not be applied correctly [6007998].
Personal experience also matters here. If the latest driver works fine for you, then there is no reason to change it. It is also worth remembering that every system behaves differently, so what works well on my setup may not necessarily deliver the same results on yours.
Source https://forums.flightsimulator.com/t/nvidia-geforce-game-ready-driver-610-47/767147/60