Repository navigation
v0.5.0.1-test Minor Maintenance Update #246
Replies: 8 comments 25 replies
|
Will do a 2D flight and maybe a short VR. Still having trouble with my Quest3 and have to use the USB connection. The 787 is also having some problems since SU5 with that pesky autopilot. Been busy today. |
|
For comparison, I just did a NUQ run in VR in the 787 (handflown because I could not recall your autopilot instructions for this test), which is charted below. My GPU is working pretty hard too, with both VRAM and usage above 90% most of the time. You can see I am getting much smoother FPS around my fixed 40 FPS and significantly higher TLOD (1000 is on the 100 line). That dip right at the end is as I am pulling up to the terminal, which then quickly recovers. Even on landing I was getting over 500 TLOD whereas yours dropped to 25 on touchdown. As I said, something is definitely not right with your current setup. 
I note from your log file: that some of your VR settings are higher than mine: With the exception of cloud quality and parked aircraft quantity which I have manually set to high and off respectively in VR, I am otherwise using the high preset for all my other VR settings. I suggest you try that too, at least so we have a common MSFS settings baseline to compare results with. Your system really should be getting similar performance to mine. To summarise my recommended actions:
|
|
Will be doing an NUQ flight for comparison. Turned off the optimizer and will run a 2D then VR flight. These are the settings I was wondering about (DLSS etc). Not VD, which I cannot use with my sick Q3. Cannot use the Meta Link either except the cable. 
For the record here is the NUQ flight plan (simple) just climb to 3000' (put in AP). The AP is set to LCH and LNAV. There is one discontinuity. It does not transition smoothly to the ILS and you have to copy the RWAY28R line to the top of the LEGS and execute when you intercept and then also the APP button... this is very awkward and not yet perfect at all. The target speed used to be 140 at 25 DEG flaps but I seem to have to run faster now. This one I ran at 150 but the plane flew mostly at 145 and use 20 DEG flap. Something not right here. https://planner.flightsimulator.com/share/382363de-5986-4ede-bf70-5b486c86cc63 2D B-787 NUQ MSFS_AutoFPS20260504.log One more.. verified there has been a change in the flight dynamics... better not go under 150 KNOTS and 20 DEG flaps... fall off GS. I do not know why the FPS drops at landing as the GPS/ VRAM and CPU are not overloaded.... 2D B-787 NUQ2 MSFS_AutoFPS20260504.log Almost bedtime but will see if I can do a VR. Was pretty disappointing: VR B-787 NUQ MSFS_AutoFPS20260504.log VR settings: 
|
Yes, I said that I was using DLSS Balanced. I see that you are using DLAA which is a significantly higher load on the MSFS engine. Try Quality or Balanced for likely much better performance and negligible visual quality reduction if you are using Preset M DLSS 4.5, which I am pretty sure you are.
I already have that flight plan but the AP just did not want to play the game, trying to level me off at 1700 ft even though I had 3000 set, so I sacked it and hand flew it. The flight profile did not need to be super accurate to compare, and the vast difference in achievable TLOD was the main thing to note.
That is really bad. No more than TLOD Min for most of it, and half that when things got really bad. FPS on landing was half your cap, even with settings reduction active. Something is really off. Is it a similar experience when not using AutoFPS, or configuring it for monitor mode, and default TLOD set to 100? 
|
|
Yes, the VR. I think I have Meta almost convinced to take the Q3 back. I will run it again in a bit with the balanced setting. Here is the RNO 2D run with the DLSS set to balanced. I took a screen shot of the view of the clouds at High I believe now with your app and the DEV FPS. You can see why I wanted to do something with the CPU measures because it shows the main thread as full but your CPU measures (and GPU) show lots of headroom. It is running fine at altitude but you see the same thing when the FPS slow on landing. After landing I let it run on the runway since it seemed to be recovering. 2D B-787 RNO MSFS_AutoFPS20260505.log 
Here is VR not good at all. I tried to simplify and failed to get an easier method and like you flew it manually at the end of the day. There are new problems with the autopilot. |
As you say, apart from right on touch down, your 2D performance is actually pretty solid, and it recovers quite quickly thereafter. 
A similar result to before. Was this with DLSS set to balanced like you did for 2D? 
I also note that while you have reduced some VR settings, you have not changed your VR settings to the actual default High preset. I strongly suggest you try that, at the very least so we can do an apples to apples comparision. |
|
Long Aussi 2D flight. Could not get VR to work with wire. Praying for return. |
The MSFS reading in the app is just the average load across all CPU cores. As you saw when I was coding this, I was not able to get more detailed stats than that unfortunately.
I remember it was pretty easy but for some reason the day I tested in the AP was not having any of it. Thanks for the refresher in any case, which I will refer to when I next try this.
Apart from your usual FPS and TLOD dump on touchdown, this is really solid 
|
|
Will be doing a long Aussi Flight today and have a new cable supposedly will attach my USB 3.2 port to the Q3. The port is supposed to run at some astronomical speed (20 GBs). Tests in Q3 at 2.7 GBs. It is all USB C straight through. We will see. I think I will get anew Q3... they seem to be accepting the RMA and preparing a label. In the interim I was looking in detail at the Crystal Light and seems like a difficult row to hoe.... I do have a good DP port GPU etc... but still people seem to have trouble. A lot of trouble with set up and determining the intrapupilary distance... I just had a good exam and asked for an exact reading (67mm). This is also cranked into the Q3. My old roomate is an opthalmologist and says that the horizontal centering is relatively easy but vertical is so hard they just guess and hope for the best. It is usually not as crucial for small/normal lenses but can be a problem with some of the larger lens that are popular nowadays. On the programming, is there any way to deconstruct the compiled code for the MSFS DEV FPS code to see if there is a tag or whatever you might access through your current means? Irritating for me to watch that main thread die when I have a very fast processor etc... |
I may have already mentioned that I received a full refund on my 2 year old Quest 3 two months ago, after it would not get past the logo screen on start up. I got all excited about the possibilities of a new, higher resolution VR headset but eventually realised that not only would they cost a lot more than the Quest 3 but they either had major design flaws (eg. glare with the BSB2e) or the much higher resolution without reliable dynamic foveated rendering would likely mean I would feel compelled to update to the stupidly priced 5090 and still not really be satisfied. I ended up going with another Quest 3 and I am very happy with that choice.
I tried this very thing when I was toying with getting perf data direct from MSFS. While I could find some tags, none of them were mainthread and even when I did I could not locate their associated data as they were changing too rapidly to get a good memory search lock onto them. |
|
Well was able to run this flight with the new cable in VR but it was blurry. Good news is they have accepted to replace. Just hope the new one works and was not me and my system. Too bad on the tag... I know you spent a lot of time on it... This was a very challenging flight and never could have done all these back country airports without VR. |
From a TLOD automation vs FPS accuracy perspective, that is actually quite a pleasing result: 
If I contrast that with the automation in 0.5.0.0 from one of the flights you did just before release, it is significantly better: 
It's a pity your headset was blurry, which I hope will be resolved with the replacement you are getting. Was it otherwise relatively smooth? |
|
One more change for the day with Test5 now released. I have come to the realisation that tuning the Fixed Target FPS code for FreeTLOD will never result in an optimal solution as the two features have conflicting objectives ie. Fixed Target FPS wants to tightly manage FPS but in doing so TLOD can wander quite a bit and FreeTLOD wants to tightly manage TLOD but that causes FPS to vary. As such, I believe that where I am at now with the app is that there is a suitable compromise between the two ie. you will get reasonable, but not great, FPS and TLOD tolerance. FreeTLOD is only a late entrant to the AutoFPS feature set and it is really best used with a manual, non-fixed FPS. With the help of Copilot, I have come up with some words for the readme revision coming up when this test version is formally released that describes my recommended usage of with app with a Fixed Target FPS, as follows: Using Fixed Target FPS: How It Affects TLODRecommended: Fixed / Extra TLODWhen you choose Fixed Target FPS, the controller’s priority is to keep your frame rate steady. To make that possible, TLOD needs to follow a predictable pattern. Fixed and Extra TLOD modes provide this by using a stable altitude‑based schedule:
Because this schedule is fixed and conservative, the controller can keep your FPS locked while still giving you higher detail at higher altitudes. When set up correctly such that the TLODs at the various altitudes are easily achievable most of the time, this combination delivers both a stable FPS and consistent, realistic TLOD behaviour. Optional: Free TLOD (not ideal with Fixed FPS)Free TLOD lets TLOD rise and fall dynamically based on performance. This works well when FPS is allowed to vary, but with a fixed FPS target it creates a tug‑of‑war: Free TLOD tries to increase detail when performance looks good, while Fixed FPS tries to reduce detail to protect the frame rate. You can still use Free TLOD with a fixed FPS target, but it will always involve compromises—either FPS won’t stay perfectly locked, or TLOD won’t behave as freely as expected. In Short
|
|
The long one I uploaded above is with Test5. I had to use wired VR. Hope my new Q3 will be OK. On that CPU measure. Did you try various arithmetic variations of the variables you were able to get? Maybe they are reporting a difference or sum value? Or some adjusted value? Too bad we cannot log the DEV values because then we could do some math to relate them to the values you can get. |
I tried all different combinations of summation, averaging and peak monitoring to no avail. The issue is that the data Windows reports is per process, not per process per thread, and even if the latter was possible that still would not translate to actual MSFS internal frame times. The best I could come up with is what you see in the app now. |
|
RC1. Quite a reduction on landing on the Reno flight. Waited around and did not spring back as fast as previously on the runway. |
Yes, that's from a recent change where if you get into settings reduction territory, which you did on this flight, then it lowers its recorded peak TLOD to your current TLOD which means that when it recovers you do not get the rapid increase up to the last peak TLOD because you are pretty much there. This is intentional, because if you have just had reduction settings happen it is IMO a reasonable assumption to take it easy coming out of them so as not to immediately trigger them again. You can actually see on your recovery that this is valid because you experience another 7 minor FPS breaches on the way up meaning your system was still close to the edge performance-wise. 
|
|
Fingers crossed but maybe the VR is fixed. Here is the log from Reno this AM in VR. Running at 500 MBs with fingers crossed on more time. I noticed the the performance overlay (option for the bottom) for VD has an app speed measure that looked a lot like the DEV FTS Main Thread numbers. VR B-787 RNO MSFS_AutoFPS20260510.log Longer VR flight with the Aussi Group in Tasmania. Still looks good. |
Apart from your usual FPS and TLOD tanking on landing, this is much better otherwise. 
Even better. What did you do to fix your VR, as that is quite the turnaround? |
|
It was complicated. After a week I threw in the towel and begged for help from an IT guy who is in the Aussi Group and he had me "Ping -t" the port assigned to the Q3 and that seemed to open it up. For some reason it needed this to lock on. Neither he not I understand it. Have been running the ping in the background costs nothing. In a couple days will stop and see what happens. Then, he had me look at the VD flash screen at the bottom of the window and noticed I was running at only 200 Mbs. I then changed that setting and a couple other things based on a YouTube video. Like the Codec and I was shocked to see how good it looked and the numbers I was seeing on your app. I do not want to breath on it. It is running at 500, which seems to be the max in VD. It theory could run faster with the Meta app wired but looks terrible. This looks very very good almost 2D. In that VD screen/app I could see the reading that was running about 20ms +/-, which is what I see on the DEV FPS. I was wondering if that could be a proxy. |
|
Well, whatever it really was, I am glad you are back in the VR skies with good performance and visuals again. Re VD performance data, I'd say that is coming from the VD app in the Quest 3 and that would be even harder to obtain than MSFS internal performance metrics. I think it is best to let sleeping dogs lay. |
|
Long Aussi Group flight in slow Ford Trimotor. In and out of VR to multitask. Seemed to work fine. This was after a reboot so was happy to be able to get VR back. |
|
Your log didn't upload correctly. You might have clicked on reply before it finished uploading. In any case, I am very pleased for you that your VR issue seems to be sorted. I would hate to not be able to fly properly in VR. |
|
Humm, one more try. I hope the VR is sorted out. Feels on the edge. The Meta app does not work.... Here is the file in VR and a later 2D RC2 file. |
Some wobbles at the start but otherwise completely fine. 
The usual big dip on landing. That'll probably do for logs now as I am nearing release and it all seems to be going well. 
|
Uh oh!
There was an error while loading. Please reload this page.
MSFS Compatibility: 1.39.12.0, 1.7.27.0.
Release notes for 0.5.0.1-test
Test1:
This discussion was created from the release v0.5.0.1-test Minor Maintenance Update.
All reactions