Never tried to OC a CPU (no idea how to do that, so i think it might be better not to try it ^^)
So @SkipTalbot two more questions to you:
- Do you OC you CPU?
- What in Game settings do you have?
Thanks 
yeh, I’m betting it’s just a difference in settings.
The main takeaway here is that @charliebravo737 and I have the same hardware configuration and software yet are seeing significantly different results. Not even looking at framerate, the fact that I’m solidly GPU bound and he’s solidly CPU bound shows there’s something up somewhere. I’m in the middle of work stuff here, but when I get a chance I’ll send over some more screenshots matching your tests with a DV20 in a non-photogrammetry area. I suspect if I turn v-sync off I’ll be solidly above 60 fps, but we’ll see.
Here’s what I had to do to hit 60 fps in that A320 at Chicago Midway screenshot I sent:
-
Render scale at 80%. Gave me several FPS. But since you’re still CPU bound and see no difference at 1080 even this probably won’t help any.
-
Terrain LOD 135
-
Instrument refresh at Medium
-
The rest is Ultra at 4k
-
Clouds, AI air and ground traffic, and multiplayer were off so that I could keep my tests consistent.
-
That was two or three updates ago, which is probably super important to mention. I think I’m running 5 or so less fps now, not counting the stuttering intervals.
-
100 mhz GPU clock overclock and 150 mhz vram overclock, got me the little nudge needed to touch 60 fps. I was averaging 57-58ish for the most part, and had v-sync off for that test I believe so it was just a coincidence that it was at 60.0 fps. I do normally clamp it at 60 fps with the game’s v-sync, however. Otherwise the frame tearing is insufferable. I generally fly GA over rural so I’m usually at 60 fps, and I notice immediately when something causes a dip in my framerate as it’s super apparent with head tracking.
I don’t usually overclock the CPU. I’ve tinkered with some auto profiles, but didn’t see much effect, and didn’t expect to since I’m solidly GPU bound. I think overclocking the 5900x in my case would just generate more heat and instability. Might be worth tinkering with on your end though, or enabling precision boost overdrive.
I did see several FPS gain disabling SMT when I put the simulator into a CPU-bound state, however, and I leave the setting disabled for now.
You mentioned your XMP profile and I didn’t look at the numbers closely, but somebody on here was also posting about how they got terrible performance with their Ryzen if they didn’t have the right clock ratio between cpu and memory, 1:1 I think it was? Or it was 1800:3600? I haven’t tinkered with that, so I don’t know, and your settings may be fine already, but something to check out if you haven’t.
1 Like
Yeh but it’s not between CPU and memory, but between de infinity fabric clock and the memory, but this is normally handled automatically (infinity fabric clock should be half of the RAM clock (so 1800 for 3600MHz, or 1600 for 3200MHz).
1 Like
Hi @SkipTalbot ,
thanks for your very very detailled answer.
So you’re setting are higher than mine were ^^
That’s really strange, no idea what to do any more 
I really appreciate your help, thanks for that!
Regarding the RAM clock ratio / infinity fabric:
I’ve no idea how to change that, but I also think, that should (must?) be fine and be done automatically like @MortThe2nd stated.
well, increasing your graphics settings will likely lead to a scenario where you’re GPU limited as well, just like Skip
In the end, if you have an enjoyable experience, is it really worth breaking your head over? I mean, 52 or 59 FPS, it’s not that much of a difference is it?
You could test with SMT turned off, but then you’re basically downgrading your 5900X to a slightly higher clocked 5600X.
@MortThe2nd
No, I just decreased stuff, that’s CPU load heavy ^^
And that’s the problem. 45, 50, 60, whatever of those numbers, I’d be happy. I don’t need a constant 60 FPS in a SIM. But if I fly with an airliner at 30k ft, it’s fine, but that’s it.
Approach with the A320 or the new CRJ700 i got frames below 20, and that on a standard Airport.
VFR (DV20) NYC, I get frames between 20 and 30.
My goal is VR, but when I get those frames without, I don’t need to try VR
Actually I’m flying with SMT disabled, since I get ~5 FPS more.
What is quite strange. When I fly nowhere, and I get 50FPS,CPU is limiting and has a usage about 15%.
Then, when I fly NYC low level, frames go down to 20/30 still CPU is limiting, but CPU Usage is above 30%. So why does the sim doesn’t use 30% in the first scenario?
There must be sth limiting (but since Benchmarks are good, I don’t think it’s HW).
Maybe Asobo just don’t like my hardware ^^
maybe you’re just more affected by the current performance bug then.
I would expect slowdowns on approach with AI traffic / airport static planes etc, but I believe you’ve already turned those down already for your tests.
Did you try selecting ‘use generic models’ for both player aircraft and AI craft?
1 Like
I would refrain from doin performance testing at the current state of affairs with the performance issues affecting the sim!
Ok, had some time to do some fps gawking. What’s changed since my 60 fps screenshot:
- Updated Nvidia drivers from 457.51 to 461.92
- Sim Update 3 (60 fps screen shot was dated 31 December 2020, so before UK World Update)
- Latest 20H2 Windows updates
Same settings as 60 fps screenshot after updates:
So right off the bat I’ve lost 5-7 fps, which I expected and have seen since a couple updates ago. What’s interesting, however, is that I’m MainThread limited in this situation whereas I was universally GPU limited at 4k before the updates. But the difference between the MainThread and GPU times is incredibly close, it seems like this is the sweet spot balance between between CPU and GPU loading. Maybe Asobo was actually shooting for that, or it’s just a coincidence?
If I up the Terrain LOD from 135 to 200, Willis Tower (Sears Tower) is recognizable at this distance, and I’ve also upped the render scale from 80 to 100:
I’m back to being GPU bound, whereas you’re still MainThread limited, which is interesting. Again, the difference is really close between CPU and GPU loading though.
If I increase the instrument refresh from Medium to High, I lose another 3 fps:
Everything is pretty much maxed out now. I’m in a photogrammetry area with an airliner and all the sliders pushed to the max.
Couple quick tests with the DV20, which I actually haven’t flown at all in this game, but should because I’ve actually taken lessons in it and remember it being a fun and easy plane to fly compared to the ratty buckets of bolts Cessnas I did PPL in.
Loaded up at 0C8 Cushing, a non-photogrammetry area out in the country with Terrain LOD at 200 still and 100 render scale, 4k ultra everything else:
What’s interesting about this is I’m getting the same framerate as the A320 in Chicago. I thought it would be substantially higher. Solidly GPU bound now, whereas again you’re always CPU bound.
At altitude (really high for the DV20) I can finally breach 60 fps with vsync off yet everything still maxed out settings wise:
So there’s a big difference with settings and updates here, but again I think the difference between your performance and mine is still quite substantial. More than it should be given the exact same hardware and software. Unless there are additional settings we’re overlooking, there seems to be an issue holding your performance back. And again, that you’re always CPU bound, where I’m still primarily GPU bound in most situations should be a clue.
Don’t know if this helps any, but I like to nerd out with settings and specs sometimes.
1 Like
Thank you, yes.
There are at least a couple performance issues rattling around in the sim, but the numbers I posted above are pretty much what I was getting prior to World Update 3: UK. I’ve been hit with nasty periodic stuttering (really low FPS) intervals now too, but they don’t manifest right away, and I pulled the above screen shots right after loading a flight, and after the sim settled down from loading scenery. So I think you can still do some baseline testing if you avoid the current issues, unless another bug is constantly present.
I’ve been doing most of my flying lately in the Spitfire, over flat non photogrammetry areas with the Terrain LOD set to nothing in order to avoid the stuttering. Still looks pretty in that situation, but yeah, I’m hoping for a fix real soon.
1 Like
Yeah, I use generic models now for both…
@SkipTalbot
Wow, thanks for your very detailed tests and answers.
I’ll try todo the same testing(A320, Chicago, same Settings) (I hope I’ll get the chance to do it soon) and post my results.
Now it should be really the same (I’ll also do an empty community folder).
Did you turn off all the traffic (also ground etc?).
Today I did a “CPU Load test”. What was very interesting is, that MSFS can use a lot of my CPU, if it wants:
So you can see CPU usage above 66%, btw of course Limited by MainThread, 18 FPS.
But this was Low Level in the City of Seattle with ground and air traffic and LOD’s set to 200.
But that may show, that the problem isn’t the CPU itself. Because in the other tests, CPU usage was about 30% (still limited by MainThread).
As you can see here, Limited by Mainthread, but compared with the other one, CPU is bored here:
Less CPU usage (33%), but still only 30FPS in an GA Aircraft, no Photography area and limited by mainthread.
@SkipTalbot
So I now did some extensive Testing, interesting results, when I’m limited by MainThread, perfomance is about 10 FPS less then yours. So what is interesting, I get more Frames in Chicago den in Munich (Germany), but espacially in the Airliner ist’s limited ba Main Thread ALL the time.
So all tests wer done with the Preset Ultra (nothning changed except LOD and render Scale) and clear skies. I started the PC Shut Off, waited 5 mins, start MSFS, start the flight, wait one Minute until stable frames and take a Screenshot.
So here my results:
Test 1 - KMDW:
- LOD: ALL 200
- Render Scale: 100
- Traffic : OFF
- Live Player: OFF
- Ground Traffic: All 0 (Land, Sea, Airportlife)
Test 2 - KMDW:
- LOD: ALL 135
- Render Scale: 80
- Traffic : OFF
- Live Player: OFF
- Ground Traffic: All 0 (Land, Sea, Airportlife)
Test 3 - KMDW:
- LOD: ALL 200
- Render Scale: 80
- Traffic : OFF
- Live Player: OFF
- Ground Traffic: All 0 (Land, Sea, Airportlife)
Test 4 - KMDW:
- LOD: ALL 200
- Render Scale: 100
- Traffic : OFF
- Live Player: OFF
- Ground Traffic: All 40 (Land, Sea, Airportlife)
Test 5 - KMDW:
- LOD: ALL 135
- Render Scale: 80
- Traffic : OFF
- Live Player: OFF
- Ground Traffic: All 40 (Land, Sea, Airportlife)
Test 6 - KMDW:
- LOD: ALL 200
- Render Scale: 80
- Traffic : OFF
- Live Player: OFF
- Ground Traffic: All 40 (Land, Sea, Airportlife)
Test 7 - 0C8:
- LOD: ALL 200
- Render Scale: 100
- Traffic : OFF
- Live Player: OFF
- Ground Traffic: All 0 (Land, Sea, Airportlife)
Test 8 - 0C8:
- LOD: ALL 135
- Render Scale: 80
- Traffic : OFF
- Live Player: OFF
- Ground Traffic: All 0 (Land, Sea, Airportlife)
@SkipTalbot Did another test, that really shows the problem in a very good way:
EDDM, CRJ700, standing on the RWY:
4K, Ultra Preset, 130% Rendering (so more than 5K):
1080P, Ultra Preset, 100% Rendering:
1080P, Ultra Preset, 100% Rendering:
So 1080P low has the same frames as more than 5K Ultra 
That CAN’T be normal, right?
1 Like
Well, it’s perfectly normal, since you’re main thread limited in that scenario. Lowering resolution will only give you more FPS if you’re limited by the GPU.
While I still would advise against doing too much testing with the current performance bug still around, a direction to look for is settings that specifically hammer the main thread, like the glass panel refresh rate, and maybe terrain/object distance sliders. The CRJ is also a pretty demanding plane on the CPU. I’m sure if you would pick the C152 you’d see different results.
2 Likes
@MortThe2nd
I also change the setting from Ultra to low in the Screens (like discribed above).
That means terrain/object sliders from 200 to 25. Both.
Glass panel refresh is low.
Of course it is different with the C152, but it isn’t normal that nothing changes, independent of all settings, and that I’m limited by Mainthread all the time. I don’t say it’s the CPU, but is has to be sth.
Btw, I’m struggling with that stuff since January 14th, so before the update, without the CRJ.
I don’t know what else to tell you.
When you’re main thread limited, most graphical settings will have little to no effect on the FPS.
So it’s either something in your hardware setup / driver settings / chipset firmware / BIOS settings / BIOS version / Windows power/performance settings that limits the single core performance, or something in your game settings that puts more load on the CPU.
If you run Benchmark programs (like Cinebench R20) are you getting the single core performance numbers you would expect? I can run that on my Ryzen 5800X so we can maybe compare that?
edit: Here’s my single core score on Cinebench R20: 624
1 Like
Haha, i know, i still have no more clue, that’s why I’m hoping in general (not espacially you), that there might be somebody with THE hint in will help.
I know, that graphics setting will not have much influence on FPS if I’m limited by Mainthread.
That’s why I’m desperate, to find out, what limits me…
I tried the same settings as @SkipTalbot , same CPU, same RAM configuration, same graphic chip, and I’m limited by MT, and he’s limited by GPU and get much higher frames.
Even if I set ALL settings down (turn off Traffic, Online Services, 3D, Ground stuff etc.) I’m still limited by MT.
Windows Power Settings is all on high performance.
Benchmarks are fine I think, i don’t have R20, but R23
Let me download R23 and check mine.
1 Like