I have tested all possible ways to lock the fps on my 2080Ti, but nothing gives me better smoothness and overall performance than run it unlocked with G-Sync and V-Sync enabled in Nvidia Panel, I can not see any benefits at all by locking fps, especially at 30fps, it gave me the worst performance of all with horrible mouse and camera lag.

Edited October 17, 20205 yr by Ixoye

1 hour ago, Ixoye said:

I have tested all possible ways to lock the fps on my 2080Ti, but nothing gives me better smoothness and overall performance than run it unlocked with G-Sync and V-Sync enabled in Nvidia Panel, I can not see any benefits at all by locking fps, especially at 30fps, it gave me the worst performance of all with horrible mouse and camera lag.

I think one of the few things everyone agrees with is that locking the FPS in the sim leads to much lower performance than doing so using Nvidia Control Panel or other external applications. When I locked it at 30 in-game I was barely getting 25. Now I have it locked at 35 in NCP and I'm getting 35 even in cities (not counting NYC) on my 1080ti. 

That's interesting. I'll give NCP V-sync a try. I have to lock on 60fps in MSFS to get 30. If I lock at 30 often it's 15. Maybe MSFS is too aggressive on adaptive V-sync or sth.

The reason I decided to lock fps is that it appears to help eliminating stuttering.

Since few people have a 3090 and it seems some are curious about the performance, I will chime in here as I have done elsewhere. I have a 3090 MSI Ventus. So far I have had no trouble with it. I have it overclocked a little bit (+100 MHz on the core clock). I could probably go a bit higher but I'm not motivated to do so yet.  I am running an i9 9900K at 5 GHz across all cores with no AVX offset. The 3090 was purchased for machine learning work to replace the 2080 TI that I was using. 

In terms of MSFS performance, with the 172 steam gauge sitting on runway 4R at KEWR (typically a heavy load situation), scattered clouds, with 4K at (mostly) ultra settings (terrain LOD 200 with 100% render scaling) I get 47 fps. The sim is using 7.7 GB of VRAM and my total VRAM usage is 11.5 GB. Frame rates remain in around the 45-50 FPS range for a low flight (< 2000 ft) over NYC. In this situation, the sim is primarily main thread limited. I consider that to be a "practical" worst case - i.e. one that folks would typically encounter. It isn't an absolute worst case with something like render scaling at 200%. 

I could get the same performance in MSFS with the 2080 TI under the same conditions with the settings on high with a few on ultra and an terrain LOD of 100 instead of 200. So the sim looked very similar but obviously LOD was reduced in the distance. Under these conditions, however, the sim tended to be GPU limited (which actually made for somewhat smoother performance). 

So the 3090 is a bit of a step up from a 2080 TI but the difference isn't huge. The 2080 TI was expensive at the time and the 3090 is a few hundred more. I would argue that it isn't worth the upgrade unless you have an unlimited gaming budget. The 3080 is a much better bet except for the VRAM limit. Interestingly, the 2080 TI would have just about handled the aforementioned NYC scenario in VRAM but the 3080 would probably struggle. A 3080 with more VRAM (20 GB?) seems likely, but it will, of course, be more expensive. 

The bottom line (as has been mentioned many times by many people) is that the 3090 was not really intended for gaming. It is a good platform for a graphics workstation or for scientific work with CUDA. It is great that some folks on this forum are very happy with their 3090 but it is incorrect to think that the 3090 gets even close to 60 fps under all conditions at ultra settings at 4K which I have seen other people claim. To me, the sim looks great and performs very well with a 2080 TI by just dialing a couple of settings back a bit and the 3080 offers about 20 percent better performance. I have no problems with P3D or X-Plane visuals either, because I got into this hobby to learn about aviation, not to stare out the window admiring the scenery. In fact, I seem to recall having no problem with subLOGIC visuals either, way back in the mid 80's :) 

3 hours ago, Ixoye said:

I have tested all possible ways to lock the fps on my 2080Ti, but nothing gives me better smoothness and overall performance than run it unlocked with G-Sync and V-Sync enabled in Nvidia Panel, I can not see any benefits at all by locking fps, especially at 30fps, it gave me the worst performance of all with horrible mouse and camera lag.

 Ditto!!!

G-Sync is critical. Investing in better monitor technology has been a game changer (pun intended) for me. 

5 hours ago, Ianrivaldosmith said:

What do you lock it with? Your advice is pretty poor, no offence intended. Merely because without stating what you are actually locking it with, that can make a whole world of difference. As can the type of monitor you are using....... 

Locking it in MSFS. When I lock it in MSFS it severely lowers my frame rates much more than what I locked it too.The sim also runs a lot worse. I've seen many with different cards that have the same behavior.  Perhaps locking it outside of MSFS (say with nVidia or RivaTuner) is different, but doing it in MSFS seems to make things worse.

The render scale in MSFS is the SSAA setting? I'm confused about what Rob is saying.

28 minutes ago, captain420 said:

The render scale in MSFS is the SSAA setting? I'm confused about what Rob is saying.

I was literally just about to ask the same question! Is there a way to turn on SSAA without touching the render scale? Because things do look better at 110 render scale setting on my rig, and I wonder if that's because SSAA is kicking in? 

I'd guess that rendering scale is effectively SSAA. Would be interesting to see TAA + scale 100 vs  scale 200 without TAA.

The resolution scaling is essentially OGSSAA. Increase it to 140 for the equivalent to 2x SSAA, and 200 for 4x SSAA. For 8x SSAA, 280 scaling would be needed.

Personally I think that, at least on a 1080p monitor, anything from 120 to 150 is the sweet spot. As with other games that employ TAA and a resolution scaling option, at above 150 you get diminishing returns, and the downsampling algorithm outputs a very sharp image that brings out the aliasing even further.

Just now, FlyIce said:

I'd guess that rendering scale is effectively SSAA. Would be interesting to see TAA + scale 100 vs  scale 200 without TAA.

There is so much temporal aliasing and the simulator depends on the TAA to denoise shadows and reflections, so the latter would look much worse.

46 minutes ago, captain420 said:

The render scale in MSFS is the SSAA setting? I'm confused about what Rob is saying.

Am confused also, I have read a few times that it’s scales you resolution and read that 200 on a 4K = 8k, so we need too find out as if it’s just an SSAA setting that’s new to me.

3 minutes ago, ChaoticBeauty said:

The resolution scaling is essentially OGSSAA. Increase it to 140 for the equivalent to 2x SSAA, and 200 for 4x SSAA. For 8x SSAA, 280 scaling would be needed.

Personally I think that, at least on a 1080p monitor, anything from 120 to 150 is the sweet spot. As with other games that employ TAA and a resolution scaling option, at above 150 you get diminishing returns, and the downsampling algorithm outputs a very sharp image that brings out the aliasing even further.

There is so much temporal aliasing and the simulator depends on the TAA to denoise shadows and reflections, so the latter would look much worse.

Well I learned something new. Thank you 

On 10/16/2020 at 11:53 AM, Ianrivaldosmith said:

So, few people wanted some reports when I had this card in. 
I have set everything to ultra. I have locked my frame rate using rivatuner to 60fps. 
I am maintaining the 60fps pretty much everywhere no matter where I am except major cities where it fluctuates between 45-60. However one thing to notice is the frame time. It’s substantially better. There is odd micro stutter here and there every 5 mins, but I believe that’s down to the program coding itself. 
 

Overall would I purchase the 3090 again for MSFS? Yes! Would I purchase it over a 3080? Yes! The 15% difference may cost a lot, however every 1% or more in fight simulation counts..... especially when we start adding planes and scenery etc..... it has given me headroom for the future, plus when they start making the code more efficient and switch to DX12, there is no Vram worry either. 

Resolution on 1st monitor is 3840x1600. And second monitor is bog standard HD. CPU is 9900Ks watercooled at 5GhZ. 32Gb DDR4 3200mhz ram. 3090 gpu. 
 

PS my 2080Ti is now for sale. Just pm me. 

Interesting. That's basically, exactly my hardware setup. I'm waiting for delivery of my 3090, so I can't wait! 

Rob is right, the render scaling when set to more than 100% will apply SSAA. I just did a test and I do notice a big difference between 100 and 130. The jaggies are pretty much gone. Seems like I lost about 10fps by doing this. Can I regain that fps back by lowering texture supersampling?

 

Edited October 17, 20205 yr by captain420