Amazing. I thought I found a universal truth but I guess it just worked for me. The good news is I found something for me so I’m optimistic there’s something for you as well.
In my testing I did get a small fps improvement but remained CPU bound as you. I can move the TLOD slider from the min to the max value and my fps is not greatly affected. This only started for me in the SU10 beta. Before that TLOD had a huge impact on performance. Prior to SU10 I ran with TLOD of 100. Now I run at 250 without issue.
@tykey6 suggested looking into the settings on the other tabs, specifically traffic. Have you tried those? Maybe AI traffic versus real time? There must be some setting with a great impact on the CPU.
Question: did you cap your framerate using the OXRTK? Note that when I cap framerate using the AMD software for my GPU it doesn’t work. It only works when I use the framerate throttle setting in OXRTK.
I just did a test of mine and it’s saying GPU Limited although it flashes between Main Thread and GPU on the developer measurements.
But comparing the values there is a clear difference in what OXRTK is reporting versus the developer reporting. @mbucchia has stated that he’s measuring differently because there’s not clear guidance to how the developer measurements are taken. I can state that foveated rendering in OXRTK works when the TK measurements show I’m GPU limited.
I tried all the settings in both the traffic and data sections and none of them switched me to GPU limited. Only capping the framerates using OXRTK has made a difference for me.
This table is over year old now, so it’s somewhat out of date. However it does a complete breakdown of all settings and shows their impact on both CPU and GPU in MSFS. Worth a read anyway.
Looking into this some more, it looks like for some users app CPU is always equal to the inverse of your frame rate. I haven’t figured out yet why this only happens to certain users (for example @Mayhem6633 doesn’t seem to have this problem). As you’ve observed @ResetXPDR this also confuses the OpenXR Tools for WMR overlay.
Are you both using the same version of the sim? SU10 or SU11?
SU10 giving me the same. In both cases Dev Mode overlay is 100% unusable for me with things flickering all over the place and switching between Limited by RdrThread/GPU all the time.
With OXR Toolkit version 1.2 my appCPU showed around 6 to 18 ms depending on aircraft/on the ground/airborne/PG etc. while my appGPU was between 28 to 40 ms. So clearly GPU bound. Checking in MSFS developer mode always ”limited by GPU” with the occasional CPU spike (like once every 30 sec or so).
With version 1.2.1 I’m always CPU bound. Now CPUapp is always 3 to 7ms over appGPU. No matter the scenario - on the ground at KJFK in the A32NX or at C152 over the ocean.
I had that happen on both SU10 and SU11 when I was messing around with a frame rate lock, even after I disabled it. It took a reboot of my system to restore DevMode to its usual behaviour of telling me I’m mostly GPU bound.
Frame lock introduces wait on the CPU to lower your frame rate, this always translates into getting CPU frame time to be equal to the inverse of FPS. The OpenXR Toolkit frame throttling however is eliminated in appCPU so at least we know this isn’t your issue.
FYI what rdr CPU reads in 1.2.1 is what app CPU was in 1.2.0. It doesn’t account for the CPU pre-render start work.
There is obviously something wrong with the new app CPU metric with MSFS, though it’s troubling that it doesnt happen for everyone. Also, the fact that MSFS is also fooling the OpenXR Tools for WMR overlay is not a good sign.
Hi. I’m really not technically knowledgeable enough to contribute much in this debate, but my layman suggestion to everyone who is cpu bound is to simply raise your render scale to put as much work on GPU as possible (because being cpu bound in my experience introduces stutters).
For example. My usual appcpu is now around 10-14 ms, apprndr is 4-8ms and appgpu is 20-28ms. This is with a cpu that can quite easily bottleneck my gpu. I have 9700k and 3080ti and G2. But, i’m running msfs at OXR 150%. I haven’t tried it in quite a while, but i’m pretty sure that at 100% render scale i’d be much more cpu bound. I am very confident the game runs much smoother at higher render scale than lower (since forever, even when i had rtx 2070). I am running the game using MR but contrary to recent suggestions i am also capping my framerate in nvidia control panel to 30 fps. I just find that this positively affects the game in terms of preventing stutters. The explanation behind it is somewhere along the lines that it saves what the cpu is doing in msfs outside of VR.
So TLDR…i suggest everyone to try using high resolutions. On a G2 i suggest around 3800x3800 per eye which is roughly what 150% oxr tools scale gives you.
Sorry to everyone who has deeper techical understanding of these things, if i’m saying some dumb things that i’m not aware of. Running things this way i have a completely smooth experience (with dlss at quality) and MR is constantly at 30 (logically, since i limit the fps in CP to 30). It wouldn’t go to 45 anyway, or just very rarely, but it doesn’t drop to 22.5 either. I don’t really fly over large cities though. But from what i tested it will keep MR to 30 there as well, i may have to lower the DLSS to balanced though.
Do we know which circumstances cause CTD when enabling Turbo mode?
I’m running DX12 with HAGS on and Game mode off at the minute.
I’m also at 200% render scale in OXR tools for WMR using a G2 on a 3090 in SU11.
I don’t use MR as I’ve always suffered with artifacts.
I’m also at 200% in OpenXR WMR as I read somewhere long ago that it helps with glass cockpit legibility with the G2 (and it does). Couldn’t do it until I upgraded to a 3090.
For me it looks like with TAA 100% everything is as clear as with DLSS >100% but DLSS has a horrible ghosting on everything that’s moving - planes, cars, changing digits on AP/radios. TAA100% has none of that. So I’m sticking at TAA. Perofrmance difference is negligible qith DLSS at 100% and would only suffer at anything more than DLSS 100%.
