NVIDIA drivers. Older versions are better?

[Rei-sen]

I usually install latest versions of NVIDIA drivers. Today I was doing little research and came across this topic:
https://www.reddit.com/r/nvidia/comments/aq4at3/best_performing_driver_for_gtx_1070ti/
TLDR: NVIDIA 399.24 is one of the best performing drivers.

Then I found this video:

I'm going to roll back to check it, but won't have time to check it in IL-2 VR till weekend. Before rolling back I'll measure FPS with IL-2 built-in FPS meter and then compare performance. Currently I'm running 418.91 version. Has anyone tried older NVIDIA drivers recently? If yes, feel free to chime in. I'll post my results later.

[JonRedcorn]

Gotta be careful and mindful that a lot of these small channels run these tests with pretty much 0 standards, this is more akin to a seat of the pants estimate in my mind. They might have other software interfering, during certain runs, have unmatched settings, it's all very non scientific. From what I've heard from official channels there has been no downgrade in performance from the later drivers.

[Dutch2]

On 3/15/2019 at 8:12 AM, Arthur-A said:

I usually install latest versions of NVIDIA drivers. Today I was doing little research and came across this topic:
https://www.reddit.com/r/nvidia/comments/aq4at3/best_performing_driver_for_gtx_1070ti/
TLDR: NVIDIA 399.24 is one of the best performing drivers.

Then I found this video:

I'm going to roll back to check it, but won't have time to check it in IL-2 VR till weekend. Before rolling back I'll measure FPS with IL-2 built-in FPS meter and then compare performance. Currently I'm running 418.91 version. Has anyone tried older NVIDIA drivers recently? If yes, feel free to chime in. I'll post my results later.

Good finding. This could be true, because and logical, Nvidia is focussing on their new RTX20XX and GTX16XX and all the other like the GTX10XX, 9XX and older will have a lesser prior. 

edit: I see you also run on VR, also that is not the prior for Nvidia anymore, I suspect they even abandoned VR, the new prior for Nvidia is now Raytracing.   

Edited March 16, 2019 by Dutch2

[SCG_Fenris_Wolf]

On 3/15/2019 at 12:56 PM, JonRedcorn said:

Gotta be careful and mindful that a lot of these small channels run these tests with pretty much 0 standards, this is more akin to a seat of the pants estimate in my mind. They might have other software interfering, during certain runs, have unmatched settings, it's all very non scientific. From what I've heard from official channels there has been no downgrade in performance from the later drivers. 

^ this

 

Changing something on the driver of the RTX20XX doesn't change the driver for the GTX1080 @Dutch2 They are separate drivers. It just defines what's in the package.

 

You can walk through 20 releases, and still have the same driver for a GTX1070 as it was 12 months ago. Just saying.

[Dutch2]

18 hours ago, SCG_Fenris_Wolf said:

^ this

 

Changing something on the driver of the RTX20XX doesn't change the driver for the GTX1080 @Dutch2 They are separate drivers. It just defines what's in the package.

 

You can walk through 20 releases, and still have the same driver for a GTX1070 as it was 12 months ago. Just saying.

We will see if he does do his testing in VR, without prove we can not judge. 

BTW. I maybe complete wrong inhere and I do not have any reason to have any doubt to your remarks. Only I do see a different creation and modification date on my 3x OS-boot sections, that do all have different nVidia versions from the 399.24 until one of the last. Easy to check right click on the drivers .dll files and select properties.  Also when hoovering with the mouse over the .dll I noticed it is also labeled with an different version number. 

Edited March 17, 2019 by Dutch2

[SCG_Fenris_Wolf]

Add entries and optimization maybe for some new games. If they juggle around everything all the time we'd have all hell breaking loose in regards to compatibility issues.

 

Anyway, just do some test runs, 10 times on each driver the same track on repeat. Out of these 10 take the average min and median fps from the 10 tries each. Then compare both, and we got very convincing results.

 

P.S. Just seen your Thread, very interesting Will look at it now

P.P.S. You need to use the same test track for each try to get comparable results.

Edited March 18, 2019 by SCG_Fenris_Wolf

[Dutch2]

19 hours ago, SCG_Fenris_Wolf said:

Add entries and optimization maybe for some new games. If they juggle around everything all the time we'd have all hell breaking loose in regards to compatibility issues.

 

Anyway, just do some test runs, 10 times on each driver the same track on repeat. Out of these 10 take the average min and median fps from the 10 tries each. Then compare both, and we got very convincing results.

 

P.S. Just seen your Thread, very interesting Will look at it now

P.P.S. You need to use the same test track for each try to get comparable results.

 

I also hope the @Arthur-A   will do the test  only 10 would be a bit high I suppose. Think he would see this very clear if this is the case. 

[Rei-sen]

Sorry for the delay. I did some quick tests in QMB on Kuban Summer map (4 FW-190 D9 + 4 BF-109 G2s VS 4 PE-2s + 4 P-39s) with in-game FPS counter. I actually was getting higher FPS with 418.91 driver. The FPS was between 45 and 54 at low altitude. With 399.24 the FPS was more at 45 mark and rarely rose to 51. It was quite smooth, with WMR's re-projection turned on.  

I know it's not the thorough test, but it was quite convincing for me. So I'll stick to newest NVIDIA drivers.

[Guest deleted@134347]

I'm running the latest nvidia drivers so I can use Motion Smoothing with SteamVR... which far outweighs any possible fps gains with older drivers..

[dburne]

No problem with latest drivers here.