Will 3080Ti maintain 30FPS at 4k Ultra setting?

[Republic3D 1,229]

Maybe this is a digression, but I just did some testing for reference on my particular system and decided to share it, since it likely applies to a 3080 TI (and maybe a 3080) as well:

R9 5900X | RTX 3090 | 3440x1440 90 Hz G-Sync | Ultra Settings | Terrain LOD 400 Object LOD 200 | Milviz Porter | Low complexity airport and scenery

V-Sync Off | Limiter Off | Res Scaling 100 = 65 FPS on the ground = GPU usage 85%
V-Sync On | Limiter 30 | Res Scaling 100 = 45 FPS on the ground = GPU usage 55-70%
V-Sync Off | Limiter Off | Res Scaling 130 = 55 FPS on the ground = GPU usage 100%
V-Sync On | Limiter 30 | Res Scaling 130 = 45 FPS on the ground = GPU usage 85%

Limiter at 60 = 90, Limiter at 30 = 45 due to my 90 Hz monitor refresh rate.

I've never really played with the Resolution Scaling settings before, but there's a really noticeable sharpness difference with 130 scaling vs 100 scaling. The placards and instruments in the cockpit are way sharper and easier to read at scaling 130. And the hills in the background became crystal sharp. 

The last setting with Limiter 30, Res Scaling 130 and 45 FPS seems to be a nice setting if you don't want the GPU to run at full blast and still enjoy the sharpness at res 130.

However, what's the point in having a 3090 (or 3080/3080 ti) if you're leaving the GPU at 85%? I decided to go for the No Limit and 130 Res Scale at 55 FPS, which feels just as fluid as 65 FPS. Now the GPU is running at 100% and stays at 85 C, but only fan speed at 71%. 

It surprises me how much clearer and sharper things get at 3440x1440 130 Res Scale (4472x1872) even though the monitor doesn't display all the pixels. It becomes the same as super sampling anti alias.


 

Edited October 8, 2021 by Republic3D

[Paul J 349]

-:Deleted:-

Edited October 8, 2021 by Paul J

[Maxis 1,830]

7 hours ago, Republic3D said:

Maybe this is a digression, but I just did some testing for reference on my particular system and decided to share it, since it likely applies to a 3080 TI (and maybe a 3080) as well:

R9 5900X | RTX 3090 | 3440x1440 90 Hz G-Sync | Ultra Settings | Terrain LOD 400 Object LOD 200 | Milviz Porter | Low complexity airport and scenery

V-Sync Off | Limiter Off | Res Scaling 100 = 65 FPS on the ground = GPU usage 85%
V-Sync On | Limiter 30 | Res Scaling 100 = 45 FPS on the ground = GPU usage 55-70%
V-Sync Off | Limiter Off | Res Scaling 130 = 55 FPS on the ground = GPU usage 100%
V-Sync On | Limiter 30 | Res Scaling 130 = 45 FPS on the ground = GPU usage 85%

Limiter at 60 = 90, Limiter at 30 = 45 due to my 90 Hz monitor refresh rate.

I've never really played with the Resolution Scaling settings before, but there's a really noticeable sharpness difference with 130 scaling vs 100 scaling. The placards and instruments in the cockpit are way sharper and easier to read at scaling 130. And the hills in the background became crystal sharp. 

The last setting with Limiter 30, Res Scaling 130 and 45 FPS seems to be a nice setting if you don't want the GPU to run at full blast and still enjoy the sharpness at res 130.

However, what's the point in having a 3090 (or 3080/3080 ti) if you're leaving the GPU at 85%? I decided to go for the No Limit and 130 Res Scale at 55 FPS, which feels just as fluid as 65 FPS. Now the GPU is running at 100% and stays at 85 C, but only fan speed at 71%. 

It surprises me how much clearer and sharper things get at 3440x1440 130 Res Scale (4472x1872) even though the monitor doesn't display all the pixels. It becomes the same as super sampling anti alias.


 

Thanks for this info .. Spot on and works nicely if you want to lower temps but not hard lock yourself to 30fps (as long as you have a monitor with decent Refresh rate mine is 100Hz). Still sticking to VRR but will keep this info in the back pocket if needed .. Oh and yeah if your just now discovering ultrawide 3440x1440 + 110%-130% Renderscale .welcome to the club.. Longtime member Some of us got the double benefit of increasing gpu utilization plus better AA and visuals pre SU5 .. never going back to 100% heh 🙂

Edited October 9, 2021 by Maxis

[Noel 4,483]

9 hours ago, Republic3D said:

However, what's the point in having a 3090 (or 3080/3080 ti) if you're leaving the GPU at 85%? I decided to go for the No Limit and 130 Res Scale at 55 FPS, which feels just as fluid as 65 FPS. Now the GPU is running at 100% and stays at 85 C, but only fan speed at 71%.

Seems to me the opposite is true:  the weaker your GPU is the more you need to push it.  

I had a 2070 Super previously.  This meant I was not able to run all sliders at ultra, needed to keep LOD at 150 or so, RS at 100, and most of all meant I was stopping the sim regularly when for example coming into diffuse clouds and lightening etc in order to dial back settings to offload the GPU, and that was w/ frames locked to 30 and vsync on.  Then when the demand cleared would stop and adjust back up to get the best visuals the 2070 Super could produce.  With the 3080Ti those days are over, I'm at LOD 300/400, RS 130, and never need to stop and adjust.  The GPU runs between 65% and 85%, temp around 70-75C.  I paid a lot of $$ for this GPU and I'll be using it for a good 6y.  For me, I don't have the need to run at over 30fps, and appreciate the headroom for upcoming improvements in weather, PMDGware, and the DX-12 port which is touted to be more demanding on the GPU.  So yes, for me this is the point in having a 3080Ti it's improved the experience immensely.   From running the 787-10 HD in the most complex scenarios I know it's not possible to sustain anywhere near 55FPS, not at these settings, and it would only be the same or worse for 3090.  

[Malcolm Street 74]

On 10/6/2021 at 9:19 PM, Bunchy said:

Hi all,

 

Just a quick shout out to all those with 3080Ti specifically
 

If you're going to fork out for a 3080Ti I'd stretch the extra for a 3090.

The reason is VRAM.  I went for a 3090 (65th birthday present to myself!) and find that in 4k around say Berlin or Vienna on Ultra settings or in VR (G2) I'm using as much as 16GB VRAM.  As graphics improve and DX12 comes in I'd expect this to rise further.  If you're going 4k it's a good piece of future-proofing.

[Ianrivaldosmith 5,317]

5 hours ago, Noel said:

For me, I don't have the need to run at over 30fps

We know. You tell the forum this several times per week. I am surprised the drum hasn’t worn out yet. 

[Ianrivaldosmith 5,317]

5 hours ago, Noel said:

and it would only be the same or worse for 3090.

Why would it be worse for the 3090? It’s a faster card than the 3080Ti. 

[westman 257]

18 minutes ago, Ianrivaldosmith said:

Why would it be worse for the 3090? It’s a faster card than the 3080Ti. 

its easy, check when the lovly 2070 super was the no 1 card think it gone be the 3080ti now if we follow Noel

[psolk 3,037]

50 minutes ago, Malcolm Street said:

If you're going to fork out for a 3080Ti I'd stretch the extra for a 3090.

The reason is VRAM.  I went for a 3090 (65th birthday present to myself!) and find that in 4k around say Berlin or Vienna on Ultra settings or in VR (G2) I'm using as much as 16GB VRAM.  As graphics improve and DX12 comes in I'd expect this to rise further.  If you're going 4k it's a good piece of future-proofing.

This is why I've been holding out for a 3080ti over the 3080 but can't find one at MSRP so stick with my 2080 for now.  By the time I can find a 3K card the new 3K ultra's and 4k's will be hitting LOL.  Only 3090's I can find are pre-built systems for around the $4,000 mark... Just not worth it when my 2080 with a 49 inch ultrawide G-Sync monitor and VRR is butter smooth all the way to 60FPS where I lock them.  

It's to the point where a true G-Sync monitor is as important as the GPU pushing it.   I'd wager a lesser card with VRR and G-Sync will actually provide very comparable if not equal or better results than a high powered card without G-Sync being locked with a static refresh rate or frame rate limiter the way we used to do it...  Hopefully the hardware manufacturers are giving us the resources where we don't need those performance limiting workarounds.  We've only ever needed them in FS anyway, I've never had to limit a refresh or frame rate in any other application or sim so fingers crossed with MSFS we are finally getting a sim capable of optimizing today's hardware not the hardware coming out 12 years from now.  

 

[Republic3D 1,229]

7 hours ago, Noel said:

Seems to me the opposite is true:  the weaker your GPU is the more you need to push it.  

I had a 2070 Super previously.  This meant I was not able to run all sliders at ultra, needed to keep LOD at 150 or so, RS at 100, and most of all meant I was stopping the sim regularly when for example coming into diffuse clouds and lightening etc in order to dial back settings to offload the GPU, and that was w/ frames locked to 30 and vsync on.  Then when the demand cleared would stop and adjust back up to get the best visuals the 2070 Super could produce.  With the 3080Ti those days are over, I'm at LOD 300/400, RS 130, and never need to stop and adjust.  The GPU runs between 65% and 85%, temp around 70-75C.  I paid a lot of $$ for this GPU and I'll be using it for a good 6y.  For me, I don't have the need to run at over 30fps, and appreciate the headroom for upcoming improvements in weather, PMDGware, and the DX-12 port which is touted to be more demanding on the GPU.  So yes, for me this is the point in having a 3080Ti it's improved the experience immensely.   From running the 787-10 HD in the most complex scenarios I know it's not possible to sustain anywhere near 55FPS, not at these settings, and it would only be the same or worse for 3090.  

I think it's quite clear that a slower cards has to work harder. 

My point was that if you've spent €1500-2000 for a graphics card, why not put it to work and make the experience the best it can possibly be? That's my reasoning. So when I run it at Unlimited FPS, No V-Sync, G-Sync On and 130 Scaling I get 55 fps on the ground at this fairly non complex scenery. 55 FPS on the ground is usually 65-75 FPS in the air. Since my FPS is unlocked, it will fluctuate slightly depending on the weather, aircraft and surrounding terrain. But it likely won't sink below 45 FPS at any point. With G-Sync on, the frame times are really good, and the refresh rate of the monitor is locked to the FPS output of the card. Meaning a very smooth experience. 

I understand your reasoning that you don't need more than 30 fps, that's your choice. I also get that you want the headroom available for when the weather and complex scenery hits so you get a stable experience. The way I see it, that's not much of an issue since it will stay at a good FPS rate regardless, and even at the low end it will still feel smooth due to G-Sync. Your system is really powerful, so to me it would seem wasteful to limit yourself to 30 fps. Either way, just two different approaches I guess. 

Edited October 9, 2021 by Republic3D

[steve310002 1,198]

Just a warning relating to a thread I started a couple of weeks back concerning OEM pre-built systems (as these are the easiest ways to get 3080s right now). I have a 3080 and a 10700kf.

Some posters say they get super high fps on the ground and I believe them but I think it's more complex than that and can inadvertently misleading to those considering parting with huge amounts of cash for a new GPU.

On my machine at a small city (e.g. Nice) 'non complex airport' in true 4K on ultra settings I get 45 to 48 fps with dips (not 65fps like the 3080ti owner but also less than I expect with my 3080 based on the reports of others). Meanwhile in Seattle I get only 30fps (maybe due to AI traffic and multiplayer but still a far cry from what I had expected)!  It really depends on the scenery near the aiport as well. Is it a city, is there heavy forestry, is it a desert etc.? This all impacts fps obviously so I think what would be super useful would be for everyone to agree on several airports depicting different scenarios as benchmarks. For example Area 51 is a good non complex fps test. Or KLAX or KJFK as a complex one. And then maybe Nice as an in between. Then it would be easier for us to compare scenarios with different GPUs and CPUs and see what the reality is with fps differences. Otherwise one persons non complex example is different to anothers. 

I just think a lot of people are misled thinking they will hit high fps in all senarios with a 3080 or maybe even a 3080ti. It really depends on the location with most setups as to what fps you will get and the type of system you have. Sounds obvious but it's really not.

Edited October 9, 2021 by steve310002

[steve310002 1,198]

Double post...please ignore

Edited October 9, 2021 by steve310002

[Republic3D 1,229]

15 minutes ago, steve310002 said:

Just a warning relating to a thread I started a couple of weeks back concerning OEM pre-built systems (as these are the easiest ways to get 3080s right now). I have a 3080 and a 10700kf.

Some posters say they get super high fps on the ground and I believe them but I think it's more complex than that and can inadvertently misleading to those considering parting with huge amounts of cash for a new GPU.

On my machine at a small city (e.g. Nice) 'non complex airport' in true 4K on ultra settings I get 45 to 48 fps with dips (not 65fps like the 3080ti owner but also less than I expect with my 3080 based on the reports of others). Meanwhile in Seattle I get only 30fps (maybe due to AI traffic and multiplayer but still a far cry from what I had expected)!  It really depends on the scenery near the aiport as well. Is it a city, is there heavy forestry, is it a desert etc.? This all impacts fps obviously so I think what would be super useful would be for everyone to agree on several airports depicting different scenarios as benchmarks. For example Area 51 is a good non complex fps test. Or KLAX or KJFK as a complex one. And then maybe Nice as an in between. Then it would be easier for us to compare scenarios with different GPUs and CPUs and see what the reality is with fps differences. Otherwise one persons non complex example is different to anothers. 

I just think a lot of people are misled thinking they will hit high fps in all senarios with a 3080 or maybe even a 3080ti. It really depends on the location with most setups as to what fps you will get and the type of system you have. Sounds obvious but it's really not.

This is all true. 

45 FPS on the ground at Nice seems like very good performance to me. 

It varies a lot depending on where you are, and how optimized the scenery is. Urban or desert, weather, a lot of different factors. One of the biggest FPS detractors are certain payware airports that have not been optimized with proper LOD use. 

At the same time, in urban or complex sceneries the CPU plays a very big part in how fluid the simulator seems to run. When I upgraded from an R9 3900X to a R9 5900X (pre SU5), the biggest difference in performance was urban areas. 

It could be a good idea using a fixed airport for the various type of tests so it would be standardized. 

[Noel 4,483]

50 minutes ago, Republic3D said:

I think it's quite clear that a slower cards has to work harder. 

My point was that if you've spent €1500-2000 for a graphics card, why not put it to work and make the experience the best it can possibly be? That's my reasoning. So when I run it at Unlimited FPS, No V-Sync, G-Sync On and 130 Scaling I get 55 fps on the ground at this fairly non complex scenery. 55 FPS on the ground is usually 65-75 FPS in the air. Since my FPS is unlocked, it will fluctuate slightly depending on the weather, aircraft and surrounding terrain. But it likely won't sink below 45 FPS at any point. With G-Sync on, the frame times are really good, and the refresh rate of the monitor is locked to the FPS output of the card. Meaning a very smooth experience. 

I just TO at EKCH in the 787-10 HD and at the gate frame rate is 42.  CPU load went to 87%, GPU 90%, and w/o a G-sync capable display there is no such thing as smooth, therefore I'm really limited to 30 since the only other option is 60.   I've used NCP's limiter to higher frame rates, say 40, and it does not compare favorably in terms of smooth, stutter-free animation to what happens w/ Vsync and presumably Gsync.  I'm not sure how R9 5900x compares w/ 9900K, but I know 3080Ti and 3090 are virtually identical--a while back I read one benchmark that showed 3080Ti outperformed 3090--but it was academic, meaningless in terms of real world performance difference, and most benchmarks since should make this clear.  So not sure where you're going to see 55 on the ground consistently***.  Once again, if smooth stutter-free is the goal then for me at the moment I'm limited to 30.  I've scrutinized the difference between vsync at 30 and 60, and agree there is a difference.  But for me, it's subtle, and in the course of doing what I do in MSFS, it's just not relevant, fortunately.  And therefore much less the case at say a Gsync'd frame rate of 40.

***I'm talking about how I'm set up, not changing sliders/.opt edits to be able to maintain 55, or 45, etc:  LOD-O 300, LOD-T 400,. RS 130, all sliders at Ultra, 8xSSAA, 16xAF.

Edited October 9, 2021 by Noel

[Noel 4,483]

27 minutes ago, steve310002 said:

I just think a lot of people are misled thinking they will hit high fps in all senarios with a 3080 or maybe even a 3080ti. It really depends on the location with most setups as to what fps you will get and the type of system you have. Sounds obvious but it's really not.

Absolutely, positively the case, and of course depends exactly where all sliders and edits are.