nVidia Driver Experiment in FPS

[The_Nephilim]

Hey Guys,

 

I was installing and reinstalling various nVidia drivers to see which would work best for me and I thought I would share my findings. This was just a quick test I did mostly just wanted to check the FPS in 2 different scenarios with 2 differnt maps and 3 sets of nVidia drivers. this is no way a thorough test I am just a guy with a desktop LOL!!

 

So here are the results I will first list the drivers I used and I did a DDU uninstall and reinstall of the nVidia drivers. so that was the most extensive bit I did in my test just to make sure they where as clean as I could get.. Plus this is a 1 day old Install of Windows and DCS.

nVidia driver 456.71:

 

Syria map:  Instant Action Ramp Hot Start     FPS = 71-72

Caucasus:  Instant Action Runway Hot Start =  79-81

 

-------------------------------------------------------------------

 

nVidia driver 512.15

 

Syria:  IARHS    FPS    = 70-72

Caucasus:   IARHS      FPS   =   75-77

 

---------------------------------------------------------------------

 

nVidia driver  452.06

 

Syria:   IARHS      FPS  =   66-67

Caucasus:     IARHS    FPS  =  74-75

 

----------------------------------------------------------------------

nVidia driver 452.06 ----  I changed 1 settings in nVidia Control Panel:

 

1. Power managemnt mode = Prefer max Performance.

 

Syria:  IARHS     FPS   =  61-62

Caucasus:   IARHS      FPS  =  70-71

 

 

I also changed the performance level in the 512.15 drivers and as with the 452.06 I had worse performance then if I left the settings at their default..

Being not too sure what to conclude here but from the numbers for myself it was best with an older set of drivers and Defauly settings in the nVidia Control Panel..

Maybe we  could check and see when the 2xxx series came out and which driver it was. and maybe try the last set of 10series drivers to try out. but I really just did this as a quick test to see what would happen..

 

Wierd, how I thought setting to max performance really would have helped but in my experiment it was the opposite.. I do know it makes the card run Full out all the time and I found optimal is the same but when I am not using the video card in a game or whatnot the clock speeds declock which is good on the electric bill LOL!!

This was just a quick experiment and by no means an exhaustive one so please remember YMMV and I only have a 1070GTX so newer drivers would be negligible for me, but if you have a 3xxx series card newer could be better in your case.

 

I would say maybe experiment with the max performance setting as it really made the FPS worse in my instance..

 

You can try a few yourself depending on which card you have I would not just default to the latest and greatest is not always the case..

 

 

 

 

 

 

 

[FR4GGL3]

I want to say thank you. Because you tried something (performance level) and got the idea that it won't help you on your machine.

We all should do such things from time to time. There are no such things like "best settings" as there are too many variables on a whole system. You have a 10xx Series Card. My 30xx Series Card may behave different. I even don't know your CPU Energy State Settings or your RAM Performance as you have a 10xxx Series Intel CPU VS my 9xxx Series Intel CPU.

 

Most of the time the standard Settings in the control panel of your graphics driver are working best. These are tested and meant to work. There are some weird programs that need some tweaking, but yeah, it is not that common (e.g. if you use Cinebench for testing your CPU test it with and without max performance mode in nvcp. It doesn't make much sense, but I get better scores with max performance enabled - on a CPU Benchmark. Weird, I know.). Then there are games with weird VSync Implementations. Some Systems work better if you don't use the game VSync and force VSync via NV control panel. Some Systems don't need VSync at all as long as the FPS are capped and well withing the monitors Freesync / GSync Range (works for me most of the time - but there are Games like Street Fighter IV Ultra or really old Games that I need to run via DGVoodoo. These need VSync via nvcp.)

 

I think you get the idea. This makes it so hard to tell people what settings to choose. People are forced to try by themselfes.

[BitMaster]

After fixing systems for many years, I tell ya, it only takes one wrong checkbox and your performance takes a big hit.

Finding that one box is an endavour for itself. 

Now throw in "Gaming Performance" and the ice gets even thinner.

I understand people, like my son, who went out and bought a PS5, just becaus it's simpler.

[Bossco82]

Yep what he said, all it can take is one box not ticked or un-ticked properly. It can all go out the window. When I first installed my 3090 I thought "how the hell can it be more stuttery" It does make you see the advantage of having a PS5 or Xbox but then I would have nothing to fix!

[Hiob]

4 hours ago, The_Nephilim said:

I also changed the performance level in the 512.15 drivers and as with the 452.06 I had worse performance then if I left the settings at their default..

 

Just a quick question, because I think, that this setting in particular shouldn't affect performance at all (unless you are temp limited). Have you closed DCS, changed the setting and restarted? Or have you just alt-tab'd out of dcs, changed, and alt-tab'd back in? In case of the latter, have you made sure to put DCS back to fullscreen? Because, it isn't by itself, and that will cost you that amount of FPS.

[The_Nephilim]

Well I completely shut off DCS and changed the setting and restarted DCS I did not ALT + TAB out at anytime changing settings..

 

I was running on my 1080P Monitor and not in VR I can recheck the 452.06 with alt enter to run full screen and see if there is a change? but I have full screen checked in options in DCS but I will recheck..

[Hiob]

11 minutes ago, The_Nephilim said:

Well I completely shut off DCS and changed the setting and restarted DCS I did not ALT + TAB out at anytime changing settings..

Ok, than there is no issue here.... I will check with my own system again if power management settings make a difference.

11 minutes ago, The_Nephilim said:

 but I have full screen checked in options in DCS but I will recheck..

That doesn't help. If you have any other application running in the background and just switch between dcs with alt+tab, you won't be in real fullscreen (Just a borderless "fullscreen" window).

 

Another little nitpick - if you want to measure performance, just sitting at the start of a hotstart misson isn't very reliable. Some missions have scripted events running in the background or load stuff while you are already "there". That's not inevitable a problem, but it can be. A better way is to save a track and use the replay function for reproducable conditions.

Don't get me wrong, I don't want to debunk your results. Those are just some experiences I stumbled across whilst doing thorough benchmarking for my own purposes.

Edited April 20, 2022 by Hiob

[The_Nephilim]

No I don't get you wrong and if I made an error I don't mind it being pointed out. the trk replay sounds like a good idea I will try that.

 

I reran the test with the 452.06 drivers with performance mode and optimal they where real close in performance so much. in Syria maximum performance I got 71-74fps. with Optimal I got  72-73 but it dipped to 69fps for a milisecond a few times so maybe that is where Maximum is a slight bit better in my case..

[Hiob]

power settings: prefer maximum performance

will basically just stop the card from throttling all the way to idle. A little bit as if you would keep the rpm high in a manual geared car to have the maximum power available. But when under load it will behave the same regardless of the setting. And since you will demand 99% constantly while flying, it willnever throttle down.

That - in theory - may lead to a more temperature saturated card, since it doesn’t cool down as much. But that won‘t matter past a few seconds. If you do the „jump into an instant action mission“ approach, at least sit there for a minute until everything has settled.

But as I said, I will test again on my own system. I like surprises! … and by no means am I an authority on the matter!

[FR4GGL3]

25 minutes ago, Hiob said:

power settings: prefer maximum performance

will basically just stop the card from throttling all the way to idle. A little bit as if you would keep the rpm high in a manual geared car to have the maximum power available. But when under load it will behave the same regardless of the setting. And since you will demand 99% constantly while flying, it willnever throttle down.

Pretty much this! It will only throttle if it gets bored because the CPU can't provide the needed data fast enough.

[Hiob]

Ok, did the testing with my default benchmark. (2 min flight low altitude over Syria with pretty high settings).

Prefer Maximum Power: 91.1 FPS on average

Normal Power: 91.7 FPS on average.
 

So the difference is only .6 fps - admitted, in favor of the „normal“ setting.

(side story: undervolting my 3080 to 0.9V @ 2010MHz, gives +1 fps, but reduces temps from 44 to 40 C and power consumption from 350 W to <300 W…., at stock it runs around 1950@1.037V)

 

Oh, btw… with „prefer maximum power“ the gpu is idling on the desktop with 1800MHz and 90W(!), whereas in „normal“ it idles on desktop with ~300MHz and 20 Watts…. Consider this. A lot of wasted energy and heat.

 

Another finding - you need to restart your computer for this setting to take effect reliably.

Edited April 20, 2022 by Hiob

[BitMaster]

vor 14 Stunden schrieb Hiob:

 

 

Oh, btw… with „prefer maximum power“ the gpu is idling on the desktop with 1800MHz and 90W(!), whereas in „normal“ it idles on desktop with ~300MHz and 20 Watts…. Consider this. A lot of wasted energy and heat.

Yep, I have said this many times before in other threads. What a waste of energy.

You can power 3 Mac M1's with those 90w and do serious work, no...Nvidia wants to be a heater