EXPERIMENTAL: Resizable BAR in DCS for NVIDIA RTX 30 Series GPUs

[ShaunX]

I'm posting this in VR because, let's face it, we need it the most ! but it's certainly not intended to be VR specific.

 

With the recent release of BIOS updates for some MBs to enable Resizable BAR (ReBAR), NVIDIA have started to test and enable its use on RTX series 30 GPUs for selected games via game profile updates in its game ready driver releases.

 

https://www.nvidia.com/en-us/geforce/news/geforce-rtx-30-series-resizable-bar-support/

 

On 30th March it released a driver enabling ReBAR for 17 games of which, as you might expect, DCS is NOT one of them.

 

However, a forum member @FoxTwo has identified what is thought to be the 3 "flags" in an NVIDIA Game Profile which are believed to enable ReBAR for a game. The flags can only be set for DCS via a utility called NVIDIA Profile Inspector (https://github.com/Orbmu2k/nvidiaProfileInspector)

 

It should be stressed that this is currently experimental. We have no way to technically verify that the change actually enables ReBAR in DCS and can only go by comparing experience with and without the change. Some of us believe we are seeing a performanace boost but equally this could just be random or placebo effect.

 

If you want to try this there are a number of pre-requisites:

 

1. Your MB must support ReBAR and have it Enabled. Consult your MB website and update BIOS and enable as required.

 

2. You must have an RTX series 30 GPU. Install the game ready driver released on 30th March.

 

3. ALL RTX 30 GPUs require a Firmware update. Again, consult your manufacturer's website (for my ASUS card is was named "NVIDIA Resizable BAR flashing tool")

Then go into the NVIDA Control Panel and bring up the System Information window. You should see "Resizable BAR    YES".

 

4. This is the experimental part, so be very careful and backup first. This is very much "DO AT YOUR OWN RISK"
    
    Launch NVIDIA Profile Inspector. Select "Digital Combat Simulator: Black Shark" from the dropdown.
    Click "Show unknown settings from NVIDIA Predefined Profiles" - second button from right on the toolstrip.
    Use the export button to "export the current profile including predefined settings" THIS IS YOUR BACKUP!
    Scroll down to the "Unknown" Section. You should see the following..
    

    
    Note the 3 entries which list 17 profiles.
    
    On the first 2, click on them and amend 0x00000000 to 0x00000001
    On the 3rd, click on its dropdown and select the only item.
    You should now see this..

    

    Click on "Apply Changes" and you're good to go.
    
    Once applied You can revert those changes by clicking on the icon that appears to the far right of each entry when selected "Restore this value to NVIDIA Defaults" and then "Apply Changes" again.
    
Please report your experience to this thread, whether possitive or negative.

 

If you find any way to technically confirm whether this method is actually enabling ReBAR for DCS, or indeed if we are missing something, please post.

 

AMD GPUS: I know little about AMD GPUs but I believe their series 5000 and 6000 cards already support ReBAR (or SAM as they call it) so you may only require enabling it on your MB - see this youtuber's claim that it is improving performance in DCS  - https://www.youtube.com/watch?v=xIU1dHA5JGg

 

[imacken]

Useful.

I'm still not sure if it's doing anything in DCS or not.

[coldViPer]

Interesting, I have enabled resize bar some days ago but didn't do any tests assuming DCS is not supported. Maybe I'll try this..

[ShaunX]

1 minute ago, imacken said:

Useful.

I'm still not sure if it's doing anything in DCS or not.

 

Me neither, but I was bouyed by the AMD GPU user's Youtube post. As AMD cards seem to enable support regardless of game he "appears" to be seeing a performance increase which is encouraging. Maybe we still need more work to force NVIDIA cards to enable for DCS. Hoping an NVIDIA guru gets on this thread to confirm.

[zildac]

13 minutes ago, imacken said:

Useful.

I'm still not sure if it's doing anything in DCS or not.

Likewise, I've left it on for the minute anyway.

[some1]

I've tried with and without these entries in Profile Inspector yesterday and did not see any significant difference, it was 1-2 FPS in either direction. Rebar confirmed active in nVidia control panel. 

 

That was in 2D, RTX3080 with Ryzen 5900x, DCS on High settings preset 3840x1600 res. 

[ShaunX]

2 minutes ago, some1 said:

I've tried with and without these entries in Profile Inspector yesterday and did not see any significant difference, it was 1-2 FPS in either direction. Rebar confirmed active in nVidia control panel. 

 

That was in 2D, RTX3080 with Ryzen 5900x, DCS on High settings preset 3840x1600 res. 

 

Thanks. However, I don't believe "Rebar confirmed active in nVidia control panel", ie. "Resizable BAR   YES" means that it's enabled in DCS. It may be that our attempt at forcing in on in the profile is unsuccessful, the change to the 3 entries is still guesswork. We could really use an NVIDIA expert.

[bandit648]

I just tested this after enabling rebar support and from my results I don't think it's making any difference in DCS. I used a track file I created for benchmarks that features a heavy combat scene to stress the GPU. I re-ran this several times with the rebar settings on vs off. Highlighted in green below is rebar=on in Nvidia Inspector.

 

Running the latest open beta @ 4k resolution. System Specs:

 

i9-10900K
64 GB G.Skill DDR4 CL16 3600Mhz
ASUS Z590 ROG MAXIMUS XIII HERO (BIOS 0605)
NVIDIA RTX 3090 Founders Edition
48-inch LG OLED48CX
 

 

05-04-2021, 11:53:12 DCS.exe benchmark completed, 7332 frames rendered in 106.671 s
                     Average framerate  :   68.7 FPS
                     Minimum framerate  :   41.8 FPS
                     Maximum framerate  :   89.6 FPS
                     1% low framerate   :   37.3 FPS
                     0.1% low framerate :    4.1 FPS
05-04-2021, 11:58:54 DCS.exe benchmark completed, 7120 frames rendered in 104.750 s
                     Average framerate  :   67.9 FPS
                     Minimum framerate  :   49.9 FPS
                     Maximum framerate  :   86.7 FPS
                     1% low framerate   :   38.4 FPS
                     0.1% low framerate :   30.7 FPS
05-04-2021, 12:10:56 DCS.exe benchmark completed, 7036 frames rendered in 106.250 s
                     Average framerate  :   66.2 FPS
                     Minimum framerate  :   51.3 FPS
                     Maximum framerate  :   82.7 FPS
                     1% low framerate   :   51.0 FPS
                     0.1% low framerate :   33.9 FPS
05-04-2021, 12:15:33 DCS.exe benchmark completed, 7281 frames rendered in 105.797 s
                     Average framerate  :   68.8 FPS
                     Minimum framerate  :   52.9 FPS
                     Maximum framerate  :   86.8 FPS
                     1% low framerate   :   44.8 FPS
                     0.1% low framerate :   34.3 FPS

[speed-of-heat]

Where did you get the BIOS to support your Maximus X Hero, last i could see ASUS have not yet released either their X  or XI MB, which you would require in order to get any different perf 

[PicklePicklePickle]

bandit648, when you did your test did you enable rebar support in the asus bios? I just noticed that they have a rebar support switch in the latest beta Asus BIOS 0704. I noticed it this weekend while benchmarking the 11900K but I was too benchmarked out to try it.

Edited April 5, 2021 by PicklePicklePickle

[bandit648]

5 hours ago, speed-of-heat said:

Where did you get the BIOS to support your Maximus X Hero, last i could see ASUS have not yet released either their X  or XI MB, which you would require in order to get any different perf 

 

I have the Z590 Maximus XIII Hero. Rebar support was available at launch.

 

12 minutes ago, PicklePicklePickle said:

bandit648, when you did your test did you enable rebar support in the asus bios? I just noticed that they have a rebar support switch in the latest beta Asus BIOS 0704. I noticed it this weekend while benchmarking the 11900K but I was too benchmarked out to try it.

 

 

Yes, I enabled it the BIOS and then you can verify its working in the NVCP system information. There's an option at the top of the menu bar in the BIOS to enable it. I got the Maximus XIII Hero right when it launched and the option was there even with original BIOS.

 

[speed-of-heat]

Ah, your sig still says Maximus hero X

Edited April 6, 2021 by speed-of-heat

[bandit648]

11 hours ago, speed-of-heat said:

Paola, your sig still says Maximus hero X

Sorry, need to update that. Just built a new system recently 

 

[Woona]

Hi guys, really excellent work. Blown away at how quick someone found out about the profile flags.

Here's my performance measurement on a single set of benchmarks on my regular Batumi run. While I don't see increases in maximum FPS, I see a pull in averages (94 -> 99) and minimums (60 -> 64) which are luckily both the metrics I want to see lifted with any performance benchmarks. The 99th percentile numbers are well within margin of error so no recorded differences there so far, unfortunately. 

I haven't given it the empirical test as I'm heading out the door soon, but I'll sit down tonight and add empirical data. Microstutters, dropped frames thru FrameView, yada yada.

Edited April 11, 2021 by Woona

[Brainfreeze]

So is that just the 3 flags mentioned above (sorry bit confused).  Does not feel like it is doing much

[Woona]

One more run, this time across McCarran in Nevada where I usually test ground performance and anisotropic filtering. Definite preference to the theoretically applied ReBAR. Note the flipped order as compared to last post.

 

 

  

9 minutes ago, Brainfreeze said:

So is that just the 3 flags mentioned above (sorry bit confused).  Does not feel like it is doing much

ReBAR is a relatively new tech and we're only starting to learn about it, but we already know it's system-dependent. Since you only mention the appropriate flags, did you check that your hardware is compatible? - Install relevant BIOS and VBIOS updates, then update your GPU driver and enable ReBAR in your UEFI? Also note that with VR headsets we're often locked to a reprojection framerate or full framerate, so a bump from 70-80 in an average situation means you'll still only see 45fps/whichever rate is your reprojection.

Edited April 11, 2021 by Woona

[imacken]

I'm intrigued by the fact that image shows D3D12 against DCS when it's 11.

[Woona]

1 minute ago, imacken said:

I'm intrigued by the fact that image shows D3D12 against DCS when it's 11.

Oh sorry lmao, I dragged up the FPS values on top of a completely different benchmark (COD Modern Warfare) so you had the overview of categories. If only.

[imacken]

12 minutes ago, Brainfreeze said:

Does not feel like it is doing much

The thing is, these things never do, individually.  I mean can one actually tell the difference between 120 and 127 fps? It all only shows in benchmarks.

However, the cumulative effect of several of these little steps forward is a good thing.

[Brainfreeze]

9 minutes ago, Woona said:

Since you only mention the appropriate flags, did you check that your hardware is compatible? - Install relevant BIOS and VBIOS updates, then update your GPU driver and enable ReBAR in your UEFI? Also note that with VR headsets we're often locked to a reprojection framerate or full framerate, so a bump from 70-80 in an average situation means you'll still only see 45fps/whichever rate is your reprojection.

 

Yep I went through the all Bios /VBIOS/Driver etc.   Indeed I locked mine to 45fps which is plenty but hoping to get rid of these spots where buildings are flying past in a stutter or micro glitches or Syria map fps getting pretty low from time to time.  For now checking your experiments see if it will give something meaningful (was hoping for 10% improvement area).  

[imacken]

4 minutes ago, Brainfreeze said:

Yep I went through the all Bios /VBIOS/Driver etc.   Indeed I locked mine to 45fps which is plenty but hoping to get rid of these spots where buildings are flying past in a stutter or micro glitches or Syria map fps getting pretty low from time to time.  For now checking your experiments see if it will give something meaningful (was hoping for 10% improvement area).  

10% is unrealistic. The general consensus is around 5% at best. 

[speed-of-heat]

Yep, that’s what my reading shows me, any improvement is headroom fro reducing edge cases...

[Woona]

From general run-to-run testing, it seems like I'm getting the biggest benefit on the ground. Minimums are about 5%+ in the air, anywhere between 10%-13% on the ground at larger airfields. So the "10%" hoping is a little hopeful overall, but on the ground with my system and settings, it's actually not at all far off. The ground is where my performance hitches just below reprojection the most and where I'd be happiest to see change. Gonna cook some dinner and hop in DCS, and just check if the ground feels different with or without ReBAR. If it feels different, I'll try to measure the track files afterward.

Edited April 11, 2021 by Woona

[imacken]

2 minutes ago, Woona said:

From general run-to-run testing, it seems like I'm getting the biggest benefit on the ground. Minimums are about 5%+ in the air, anywhere between 10%-13% on the ground at larger airfields. So the "10%" hoping is a little hopeful, but on the ground with my system and settings, it's actually not at all far off. The ground is where my performance hitches the most. Gonna cook some dinner and hop in DCS and just get an empirical feel for the ground and see if I can get some metrics in MP tracks.

That is going well above anything anyone else is seeing!

Do a Google search. Values between 1 and 4% are typical, ignoring Nvidia’s site of course!

[Brainfreeze]

1 hour ago, imacken said:

10% is unrealistic. The general consensus is around 5% at best. 

I know but it has such a good name...