Home
Creators
Search Recent Random
Posts
Search DMs Popular Hash Lookup Tags Random
Importer
Import FAQ
Register Favorites Keys Review DMs Logout
Community
Telegram ThePornDude
Home Creators Posts Import Register Favorites Logout

Nvidia Driver Investigation [Part 2] (Patreon)

Published:
2021-03-26 22:20:49
Imported:
2026-06

Content

Here are all the graphs which includes the 1080p Medium/Ultra and 1440p Medium/Ultra. Crap load of work here and only 25% of it was shown in the video.

Files

Comments

Flo Rian

Impressive 5600x with Shadow of the tomb raider!

Gladamas

Thanks for your work with this. I was interested to see the (primarily) single-threaded DX11 game benchmarks

ET3D

This made me think about all the benchmarks that tech sites and techtubers do when benchmarking low end CPUs, all done with the highest end NVIDIA card to remove the GPU bottleneck. Now it turns out that perhaps these comparisons don't represent how good these CPUs are for gaming, because games would be a lot more playable with an AMD GPU.

Alberto Martinez Alonso

The reason of using the highest performing GPU had always been to avoid bottlenecks and show the maximum performance of any CPU... If you use an RTX 3080 or 3090 with low/mid range CPUs you will still see which CPU is faster (you will load the CPU with the game and with the nVidia Driver) if the 9400f is equal or faster than a r5 1400 with that GPU it will not trade places when you do the same tests with the fastest AMD GPU. What worries me is how difficult it could be to extrapolate the performance of "Low/Mid Range" GPUs from nVidia and AMD that appear to be very similar in performance when tested with the highest performing CPU but will end up being very different when paired with old or low / mid range CPUs. If I see in a chart that an nVidia RTX2060 is equal or faster than a 5600XT (and cheaper BTW) and I upgrade the RX 580 which is show to be way slower than both in the same charts and when I install the nVidia card in my AMD R5 2600 and I get no significant performance gains at my screen resolution (1080p) I would feel pretty burned, won't you?

ET3D

"and show the maximum performance of any CPU" The problem is, it doesn't show that, because of the driver overhead. People don't buy hardware based on relative performance, they buy it based on actual performance. If someone sees a CPU averaging 50 FPS with the fastest NVIDIA GPU, they might not buy it, but if it averaged 60 FPS with a similarly priced (or even cheaper) AMD GPU, then it might fit their needs. I agree that the other direction is also an issue. The point is, reviewers use the fastest CPU for GPU benchmarks and fastest GPU for CPU benchmarks, but this driver overhead problem shows that this isn't really a great way to provide information to budget buyers. When reviewing lower end stuff at least, GPUs should also be benchmarked with a low end CPU and CPUs should be benchmarked with both AMD and NVIDIA cards.

Novadoll

This issue can be seen in newer hardware but in a strange way... I downgrade from a 3900x to a 5800x. I was streaming off my 2060 super because my 5800x couldn’t handle it. Which made me really unhappy... I used faster and fast preset and my main game is apex legends. When I switched to a 6700XT I can now stream at medium settings on the cpu and no longer get stuttering and laggy while streaming. This had me rethinking getting a 5900x. I switched back to my 2060 super to test it again and yes. It seems the NVIDIA drivers are just heavy. My computer “feels” lighter overall and is snappier.