Inconsistent FPS Fluctuation - High End New Rig - Any tips?

[steve310002 1,198]

I was recently lucky enough to get a new pre-built Lenovo Legion PC with an RTX 3080, 10700KF processor and 32gb RAM @ 3200mhz with MSFS installed on an NVme drive.

After more than a week of testing, I am worried something is not quite right as my fps are all over the place and I'd really appreciate the feedback of those with a similar setup.

I am running at 4K, ultra settings, vsync off and resolution scaling set to 100. I have all the latest drivers updated and have removed any bloatware. My PC is set to high performance both in Windows and the Nvidia control panel.

On Ultra, in the air above 20,000 feet I get 50 to 60fps which is great but on or near the ground, my fps are much lower than expected some locations significantly worse than others.

For example on the runway in Nice airport, France I get 45 fps in the Icon and between 45 to 50fps in the air at 2500 feet, while yesterday on the ground at KSFO I got only 22 to 30 fps on the runway with dips in fps as I pan the camera around, and after take off flying through very low cloud in live weather I got only 26fps until I broke through the cloud (in the TBM). It then increased to 40fps. Upon changing the weather to scattered clouds and flying low over downtown San Francisco I got between 30 and 45 fps (mostly 40 to 45). Again the frame rate was sporadically going up and down depending which way I pan the camera. Which really ruins any sense of smoothness.

At KJFK I get as low as 22 to 25 fps if I am looking towards the terminal in the direction of Manhatten (gets to 30fps if I am stationary for a moment) and looking elsewhere at KJFK (away from Manhatten direction) it goes up to about 35fps on the ground but fluctuates all over the place averageing 30fps. Flying low over Manhatten gives me about 40 to 45fps.

In EIDW (Dublin airport) on the ground I get 35 fps (again quite a bit of fluctuation) and in the air in the cessna 172 I get 38 fps flying low to the ground and 40 to 45 above 5000 feet

The general trend I am seeing is an average of 45fps over cities at about 2,500 feet, 50 to 60fps at 20,000 feet or above, but only 25 to 45 fps on the ground with more complex airports and addons airports giving me an average of 30fps (FSDreamteam's KORD for example) with quite a few stutters.

Being on or near the ground seems to quite dramatically lower fps which I guess is normal but to be honest I expected my new CPU and GPU to give me better fps on or near to the ground based on what others report.

I checked my CPU's performance and it is hitting 5.1ghz, however the CPU usage remains around 15 to 40%. My GPU usage in the air is 100% and developer mode says I am GPU limited whilst airborne. However, on or near the ground my GPU usage falls to 45% to 70% and it says in those scenarios I am limited by the main thread. However my CPU usage does not increase when on ground and I can see that no thread is close to maxed out (in task manager CPU performance). So it seems on the ground I get low fps and neither my GPU or CPU is being properly utilized. Not even one whole core. Is this normal? I have tried changing the scenery cache from Ultra to Medium but it makes no difference.

If anyone can advise if the fps I am getting are normal that would be great. So many people say they are getting 60fps on 4K Ultra but as every situation is different (ground versus sky, addon or no addon, city or countryside etc.), based on my experiene described here, it doesn't seem plausible that 60fps could always be achieved on an rtx 3080 and 10700k in Ultra 4K. Or is there something wrong with my machine?

What surprises me is the lack of CPU and GPU usage when on the ground. Do others experience this with a similar setup and subsequent significant lower fps? 

Also I get quite a lot of small stutters especially at lower altitudes and around airports probably due my massively fluctuating frame rates. 

I suppose this post is to alay my concerns something is wrong with my setup. Any feedback would be much appreciated. 

Edited September 28, 2021 by steve310002

[swiesma 625]

Well. I don’t read in detail as I am at work. But your FPS sound about right compared to mine (4K, Ultra, 3080ti, 5900x). 
 

Did you check your BIOS if XMP is activated? Ryzen likes fast RAM and OEMs Rend to forget to make the BIOS settings. 
Also you could run 3D Mark Timespy and post your result and we could check it. 
 

Also check the temps of the GPU. The 3000 series runs pretty hot.

 

edit: if you have a standard scenario with less fps - please name it and I will try to replicate it and show you my fps. 
 

edit2: 60 fps in 4K Ultra is in no way achievable in all situations. 

Edited September 28, 2021 by swiesma

[united001 126]

It would be nice to know how long you've been flight simming Steve, but seems like for years. I would wonder what your prior FPS experiences have been with less robust hardware. In my experience on my machine (stats prema-posted below) I've never ever seen anything above 48 FPS. But presently consistently 40-45 at altitude and on the ground. I find that my main reductions in FPS depend almost entirely on cloud cover and density. Nothing else seems to matter. If I fly clear skies, there's no airport that reduces that. Only clouds. And, importantly, I am using almost all ultra settings except a few. So FPS "all over the place," I am unsure my set-up does that other than what I mentioned. I strongly suspect your system is just fine given that no two PCs  are alike. Good flying!

United001

[steve310002 1,198]

 

20 minutes ago, swiesma said:

Did you check your BIOS if XMP is activated?

Unfortunately (and rather annoyingly) Lenovo do not allow XMP profiles to be altered in their bios. I can overclock the CPU which I have done to 5.1ghz (it appears to stay on 5ghz when running MSFS boosting to 5.1ghz on one or two cores sometimes)

I ran Cinebench on my CPU and it causes my CPU to throttle under max load on all cores due to current limit throttling (it throttles back to 3.4ghz). I saw this by stress testing on Intels Enhanced Tuning Software. My cinebench score was really low in the 8000s as a consequence of the current limit throttling.

I read that Lenovo limit the power the motherboard can receive to prevent overstressing the CPU. So basically CPU stress tests don't work properly but it does not reflect real world usage stress if you see my point. Using CPU monitoring software I can see that all cores are hitting at least 5ghz pretty much constantly when running MSFS. So it seems current limit throttling is not the cause here of my seemingly weird fps (but annoying that Lenovo don't allow me to decide to give my motherboard more current to test it). 

My PC runs very cool. The CPU is watercooled and rarely gets warmer than 65c with MSFS running and GPU stays at 70c under full load. So there has been no thermal throttling occuring.

I also tried Red Dead Redemption 2 on Ultra 4K and get 60 to 70fps which makes me think that at least the GPU is working fine. This seems in line with other benchmarks for that game.

Sorry for the long post above, I thought it best to give precise details and fps differences in various scenarios. 

Edited September 28, 2021 by steve310002

[fogboundturtle 2,708]

XMP profile can make a huge difference in MSFS. I've tested 2400mhz versus 3600mhz  and it was night and day.

[steve310002 1,198]

3 minutes ago, united001 said:

It would be nice to know how long you've been flight simming Steve, but seems like for years. I would wonder what your prior FPS experiences have been with less robust hardware

Yup, been simming 20 years. My old rig was a gtx 1080 and 6700k which I ran in 4K on a mix of high and ultra settings but at 80 render scaling. I would get about 22 fps on the ground at the likes of KSFO and 22 to 30 in the air depending on location.

Overall though there were less stutters and more consistent FPS when compared with my new rig.

What I get in the air on my new rig is fine (40 to 50ps) but on or near the ground I expected more than 25fps for example at KJFK, especially with such a powerful and pricey new rig. I am more worried that something I am doing is wrong or something could be wrong with my new PC.

If this is what everyone is getting at 4K with a similar rig then I can happily live with it. 

I couldn't see your rig specs. Are you running a similar setup and at 4K?

[steve310002 1,198]

7 minutes ago, fogboundturtle said:

XMP profile can make a huge difference in MSFS. I've tested 2400mhz versus 3600mhz  and it was night and day.

I see my RAM is supposed to be 3200mhz (though I read bizzarely that Lenovo have it running at less than that for some reason... at around 2900mhz though I might be wrong). 

I would have thought 32gb of 3200mhz RAM would work quite well. Do you think upgrading my RAM to 3600ghz would make such a difference to be worth it? It'd be quite costly if not.

[fogboundturtle 2,708]

1 minute ago, steve310002 said:

I see my RAM is supposed to be 3200mhz (though I read bizzarely that Lenovo have it running at less than that for some reason... at around 2900mhz though I might be wrong). 

I would have thought 32gb of 3200mhz RAM would work quite well. Do you think upgrading my RAM to 3600ghz would make such a difference to be worth it? It'd be quite costly if not.

On Intel is doesn't make as much of a difference. The maximum speed is 3200mhz. On AMD Ryzem, it's a totally different thing. 

[swiesma 625]

4 minutes ago, fogboundturtle said:

On Intel is doesn't make as much of a difference. The maximum speed is 3200mhz. On AMD Ryzem, it's a totally different thing. 

That's absolutely true.

@steve310002 My CPU is also EDC limited in Cinebench, but as you said: all core full load does not represent reality.

Can you run a 3D Mark Timespy? It will also give you an idea how you performe compared to same systems that were benched.

I really wonder how your 3000 series GPU stays so cool. Is it water or air cooled?

 

[Noel 4,504]

27 minutes ago, steve310002 said:

Also I get quite a lot of small stutters especially at lower altitudes and around airports probably due my massively fluctuating frame rates. 

Steve, I can't speak to the large variability in frame rate you're seeing, though it's completely plausible since demand varies greatly situation dependent, so it would be surprising to expect stable frame rate in this context unless the sim modulates detail and other demand factors in real time and I don't think that happens to a large degree currently in MSFS.  I can relate to seeing frame rate be lower despite seemingly having CPU/GPU/VRAM headroom, and I have never had a good explanation for this--usually attributing to server issues or something outside of local hardware.

What I can speak to is this:  I have only very rare random stutters anywhere--including the most complex areas on the planet.  As well, I have a rock-solid steady frame rate of 30, again everywhere on the planet.   Some here will state they can't tolerate a rock solid frame rate of 30, and instead are willing to modify settings depending on conditions, in order to aim for higher frame rates.   For me, I detest stopping a flight and ever needing to adjust settings, and I never have to--the 3080Ti gave me this over the former GPU.  As you must know, no one here can adjust all settings to the absolute maximum settings and at the same time maintain a frame rate of 60 everywhere, in every plane currently available.   Of note, 'Ultra' in sim equates to an LOD of 2.00000 for both Objects and Terrain detail.  I'm running LOD-O at 4.0000, and LOD-T at 3.00000, by editing these settings in UserOptions.cfg I think it's called and this ends up consuming more VRAM, which I have the luxury of having available.  Plus, at 3440x1440 I'm also running a render scale of 120, upscaling from the native resolution of my display.  I have crisp textures well out in to the distance everywhere.  I use the vsync option in-sim, set to 30, with a screen refresh of 60Hz.  This keeps the CPU and GPU running as low as required to maintain that 30, and not a bit more.  Hence both the CPU and GPU run very cool, even in a warm room of 75F.  At 5.0Ghz all cores my CPU runs around 45-52C, and my GPU around 68C.  In fact, I've turned down my CPU overclock as it's just not necessary.  I fly the 787-10 HD mod, Citation L, TBM930 mostly.

Once you have the basic frame rate issue solved and have rock-steady frame rates everywhere, the remaining sources of microstutters that I have discovered which do have work arounds, can be explored here:

https://www.avsim.com/forums/topic/607809-random-brief-stutters-at-complex-airports-in-787-10/?do=findComment&comment=4632883

To contest the claims by a few outspoken people here that a rock-steady vsync'd locked frame rate of 30 is not tolerable, I posted this poll recently to show the opposite is true--many people find this option perfectly desirable--especially in the context of the other option:  needing to adjust settings per scenario in order to maintain high frame rates.  There is also a very reasonable middle ground for some, who are able to vsync to something higher than 30 but less than 60.  I feel vsync always is preferred over other forms of frame lock in terms of smooth animation.  True 'G-sync' monitors can offer this, and there are other methods as well.  At some point down the road I will move to vsync to 60Hz w/ a frame lock of 60, but as I say that really isn't doable everywhere quite yet, even at 3440x1440.  In fact, if I turn off vsync flying out of KJFK fully maxed out frame rate will likely be around 34 or so, so we're nowhere near 60 given maxed out graphic detail.

https://www.avsim.com/forums/topic/607016-if-you-run-msfs-locked-at-30fps-w-vsync-on-please-respond/?do=findComment&comment=4620271

 

[swiesma 625]

11 minutes ago, Noel said:

 

What I can speak to is this:  I have only very rare random stutters anywhere--including the most complex areas on the planet.  As well, I have a rock-solid steady frame rate of 30, again everywhere on the planet.

 

Ah... here we go again.

 

Hint: Limit to 10 FPS, it's even more stable and the human eye can't detect a difference especially in flightsims. It's like buying a Ferrari, going around Nurburging, in first gear, and saying: "Looking, a steady, rock stable 40 km/h!"

Edit: Btw: Your strange "voting" - there is a saying in Germany: "Traue niemals einer Statisitik die du nicht selbst gefälscht hast" - never trust a statistic you din't manipulate yourself. 

Oh and Edit2: Op is asking if his performance numbers are ok, and you feed him your fairytale of 30 FPS.

Edited September 28, 2021 by swiesma

[CarlosF 1,241]

1 hour ago, steve310002 said:

I am worried something is not quite right

Do you have live traffic on?  this one is a fps killer, try it with it off and re-test. 

[hanhamreds 592]

Regarding being CPU but not seeing your CPU usage rise - that's essentially right as you are seeing a score for the overall CPU overhead - and not the single thread which will be pegged at 100% and limiting frames.

 

With regards to fps, yours seem about right imo, one thing you could do is limit FPS to minimise frame time spikes - MSFS does suffer from this somewhat

[steve310002 1,198]

41 minutes ago, swiesma said:

Can you run a 3D Mark Timespy? It will also give you an idea how you performe compared to same systems that were benched.

I really wonder how your 3000 series GPU stays so cool. Is it water or air cooled?

I don't own 3D Mark unfortunately. I see it's value so might consider buying it. 

My GPU temps surprise me too. The Lenovo case has great cooling so probably why.

41 minutes ago, Noel said:

Steve, I can't speak to the large variability in frame rate you're seeing, though it's completely plausible since demand varies greatly situation dependent, so it would be surprising to expect stable frame rate in this context unless the sim modulates detail and other demand factors in real time and I don't think that happens to a large degree currently in MSFS.

Thanks Noel for the tips and explanation. I will consider locking at 30fps but for the moment want to be sure things are working correctly and to check it's not a hardware or windows issue causing my fps fluctations.

 

30 minutes ago, CarlosF said:

Do you have live traffic on?  this one is a fps killer, try it with it off and re-test. 

I tried that last week and it made no difference unfortunately. Will try again though at a busier airport.

Edited September 28, 2021 by steve310002

[steve310002 1,198]

4 minutes ago, hanhamreds said:

With regards to fps, yours seem about right imo,

That's good to know. I have seen quite a few people claim to get 50 to 60fps on a similar rig all the time so it had me worried something was maybe not right especially since SU5 was supposed to give an fps boost. I thought that boost would give more fps with my new setup than what I am getting on or near the ground especially when compared to my old rig which did a good job for what it was.