Only one man's experiences, and not sure if it's worth posting, but -

I have now DDU'd the new NV driver off and reverted back to the latest game-ready driver. Hasn't actually made that much difference but possibly a little less stuttery. DX12 still a bit pointless on my system, no benefits and actually feels more "fidgety". Been playing around all morning, HAGS on/off, TLODs between 100 and 300, main graphics settings up and down between Ultra and High, changed my scaling to 90% and back.

Net result? The only actual benefit I can see with SU10 is that I don't get the CTDs any more, and I left a flight going for an hour or so and it didn't get any more flakey at the end than it was at the beginning. Performance is mostly around SU9 levels, which I in turn felt was a step down over previous releases, and I still get fairly lousy frame rates around places like EGLL or over London generally. EGLL (standard 'enhanced' MS version) won't get up much above 20fps regardless of my settings, and I similarly get stutters and micro-pauses on a flight across London. Edited to add - I don't obsess about frame rates, in that I normally would run at 30 locked, but I wanted to see what was possible since putting the 3090 in a few months back. Smooth is what I want, and it ain't there!

I am main thread limited, and looking at the CPU utilisation, yes, I can see CP0 is running between maybe 85 and 100% most of the time and the other 31 processors/threads are doing not a lot. Which is pretty pathetic for a 2022 model sim.

At least I don't crash out all the time now. So I'll just restrict myself to flying in low density areas for now while I wait and see if the proper update for DX12 improves things. This sim still looks great, and I'm happy for all you folks getting great performance out of it, but as far as I/m concerned, it took a big hit in performance and visual quality back around SU5 and it hasn't gotten back to what it was yet.

Just for comparison, I fired up my P3Dv5 system, with Heathrow max and True Earth GB, EGLC etc and did the same sort of flight from EGLL east over London. Visuals all maxed. Looked like cr*p now by comparison, and was only getting maybe 20 fps, but it was smooth with hardly any stuttering, and there was significant distribution of activity across all 32 processor threads. Asobo/MS needs to get DX12 working the way P3D has done.

Edited September 22, 20223 yr by andy1252
text

EDIT: posted before reading the rest of the thread.

37 minutes ago, Noel said:

And that's a purely artificial reason for not using Vsync--if you want actual SMOOTHEST performance, it's Vsync every time, be it MSFS or P3D.

I use the fast v-sync option in NCP as I get visual glitches with G-Sync on my OLED. Works great although there’s a tendency towards tearing whereas G-Sync was 100% smooth.

Edited September 22, 20223 yr by Cpt_Piett

1 hour ago, Roy Warren said:

What am I missing here.  From all I've read DX12 and DLSS is dependent on the new nVidia Game Ready driver that hasn't been released.  The Studio Driver has nothing to do with MSFS.

So it seems some are getting the cart before the horse.  I seems to me that the new driver will eliminate all the problems listed in the pages of comments in this thread.

I'm staying with DX11 until it comes out. DX12 is still Beta.

Roy

Quite wrong Roy, but I can understand the confusion with everything going on. 

Both DLSS and DX12 have been working in SU10 beta on the current game ready driver, but the Nvidia drivers needed updating to stop artefacts (like giant multi coloured trees ) from appearing in DX12.

This fix has now appeared first in the latest studio driver.  The game ready driver will follow at some stage with exactly the same fix included, but it doesn't matter, the studio driver works fine in MSFS anyway, so no point in swapping.  The DX12 artefacts are now gone!

As for DLSS, this has nothing to do with the drivers.  It is working as intended except it can cause slight blurring on the flat screens, which is up to Asobo to correct by masking them from DLSS.  They have acknowledged this.

You are quite right about DX10 being not being fully developed yet in MSFS though, and people need to remember this. There will be more to come.

Last night I loaded a flight with the PMDG NG into EBBR one of the best paywear airports out there, I gain 0 FPS withing DX11 or DX12. But as it happened to be the golden hour and paning around a lot in the VC and out, there was no question to me the lighting with DX12 was far better than DX11, DX11 was/is brighter but lacks the depth DX12 brings. I also "think" DX12 is smoother. I did a full flight from EICK to EGGP in the NG all as the sun was setting and yes the VC was darker than within DX11 but imo the lighting is so much richer and better. Its also nice to use the VC lighting.

Tonight is Fenix EGAA to EGGP (EZY) .

I also saw no CPU changers when using the NG, unlike the Fenix when within the new DX12 all my main cores where for the first time used. One addon users the old way and one MSFS way! I wonder?

Edited September 22, 20223 yr by Nyxx

8 minutes ago, bobcat999 said:

This fix has now appeared first in the latest studio driver.  The game ready driver will follow at some stage with exactly the same fix included, but it doesn't matter, the studio driver works fine in MSFS anyway, so no point in swapping.  The DX12 artefacts are now gone!

I think this is not the consensus yet, as there seems to be a lot of speculation whether the game ready driver is needed to “unlock” the full potential of DX12. As for my own experience, it seems to be working, so I suspect the GRD will not bring anything new to the party.

I’m bit confused though that I’m not seeing “proper” multithreading with my 12900. HWiNFO reports 1-2 (of 24) threads in use, task manager is misleading as threads seems to be shifting around i.e. not consistent usage. And the same with Process Lasso.

31 minutes ago, andy1252 said:

I have now DDU'd the new NV driver off and reverted back to the latest game-ready driver.

You had the recent studio driver but went back to the GRD? Because of CTDs? And with the studio driver you didn’t get proper multithreading in DX12? Sorry but I didn’t want to quote your whole post as it was quite long.

11 minutes ago, Cpt_Piett said:

I think this is not the consensus yet, as there seems to be a lot of speculation whether the game ready driver is needed to “unlock” the full potential of DX12. As for my own experience, it seems to be working, so I suspect the GRD will not bring anything new to the party.

True!  Although I never expect full consensus with flight simmers, or people in general!  

I have not seen anything about unlocking the full DX10 potential, and if anything, Asobo will need to do that at their end. 
The new driver just stopped the artefacts, as was quoted in the driver release notes, and that has certainly worked.

The point that is a bit vague, and the one I think you allude to, is that Asobo said MSFS needs the new Nvidia driver to trigger the new memory management code.  So the question is, is this in the studio driver?  Or will it actually be in the next game ready driver?  I wish Asobo could be a bit more clear sometimes. 

What makes it even more foggier, is that some people on the new studio driver are getting better CPU thread balancing in DX12, and some aren't - the plot thickens! 

PS. I just realised the above sentence has nothing to do with memory management, but I believe the artefacts were caused by DX12 GPU memory management problems, so if the notes say they are fixed in the studio driver (as in the release notes), then that must be the memory management problem fixed I suppose.

8 hours ago, abrams_tank said:

Nope, nobody on the MSFS team ever said this. It’s just bad information passed around from people who made assumptions.

The fix from NVidia that Asobo needed help from NVidia from was for corrupted images in DX 12, and that fix has been released in the Sfudio driver from NVidia (but the Game Ready driver with this same fix has not been released yet).

Edit: Martial did acknowledge they are getting help from NVidia but the solution to blurry glass panels with DLSS is likely the adding of a mask to the glass panels, not an NVidia driver update.

I'm using the most recent studio driver. I'm still using DX11, however.

Is there a major difference in performance in using the studio vs the gamer edition?. Thanks! 

10 minutes ago, bobcat999 said:

The point that is a bit vague, and the one I think you allude to, is that Asobo said MSFS needs the new Nvidia driver to trigger the new memory management code.  So the question is, is this in the studio driver?  Or will it actually be in the next game ready driver?  I wish Asobo could be a bit more clear sometimes. 

Exactly. Without confirming which driver is needed, there’s going to be speculations. Here’s a thread on the topic with 76 replies. Not sure if reading it will lead to less confusion though 😆

I’ll report back in a few hours with my verdict (unless I get stuck in the rabbit hole).

https://forums.flightsimulator.com/t/post-your-dx12-with-studio-driver-results-here/543823

Edited September 22, 20223 yr by Cpt_Piett

3 minutes ago, evaamo said:

I'm using the most recent studio driver. I'm still using DX11, however.

Is there a major difference in performance in using the studio vs the gamer edition?. Thanks! 

No.  And in fact, if you are staying with DX11, you don't need to update your driver at all, it is purely for DX12 problems.

15 minutes ago, bobcat999 said:

What makes it even more foggier, is that some people on the new studio driver are getting better CPU thread balancing in DX12, and some aren't - the plot thickens! 

I saw this last night, I loaded the PMDG NG and within DX12 saw nothing new, i was frankly very disapointed. But when I got up today and installed the Fenix update, it was a completly diffrent story. My only thought is the NG is writed in old code and hence a first load time of 10 mins, the Fenix is in MSFS code and loads as fast as any default MSFS aircraft. I have no idea why the Fenix show the CPU working as I hoped and the NG does not. But as you have seen in my other post Rob, the Fenix within DX12 is workings as I hoped with the new drivers. Not tried any other aircraft at all.

Edited September 22, 20223 yr by Nyxx

7 minutes ago, evaamo said:

I'm using the most recent studio driver. I'm still using DX11, however.

Is there a major difference in performance in using the studio vs the gamer edition?. Thanks! 

I think that’s what everyone would like to know at this point, lol. 

EDIT: With DX11 there should be no difference.

Edited September 22, 20223 yr by Cpt_Piett

5 minutes ago, bobcat999 said:

No.  And in fact, if you are staying with DX11, you don't need to update your driver at all, it is purely for DX12 problems.

Thanks, I updated it thinking I'd try DX12, but I'm getting such a good DX11 performance in SU10 compared to SU9 that my motivation to try DX12 has faded.

Edited September 22, 20223 yr by evaamo

1 hour ago, Ianrivaldosmith said:

Quite a lot by all accounts. 

Your response is true.  I am missing a lot and by reading all these posts I am trying to learn what to do.  I have to make a decision on which way to go, the same as you.

So, while your response was a fact, it does nothing to help me in deciding what to do. As a lot of simmers do, I come to Avsim to learn and I didn't learn anything from your post.

Roy

I don't notice any FPS improvement with DLSS on, exact same FPS, so my issues must be CPU bound.

5 hours ago, Cpt_Piett said:

Initially in the beta, DX12 gave outstanding performance with solid multithreading. 

Then they removed the memory manager to avoid visual glitches, and DX12 performed significantly worse than DX11 (stutters especially when panning). 

I've now had a few flights with the final version of SU10 and the studio driver, and DX12 performs as good as DX11, if not a bit better. 

What’s your theory on what we’re missing from the initial beta? Is it something from Asobo, Nvidia, or both?

Edited September 22, 20223 yr by Virtual-Chris