Civilization VII Performance Benchmark

[Noreng]

There's definitely a driver issue here... the 4090 being only 22% faster than a 3080 Ti isn't something normal.

Here are some recent tests from GameGPU andit's pretty night and day

MSAA running slow is probably a memory bandwidth limitation.

 

[mb194dc]

Seems like a pointless game to benchmark, any benefit from more than 60fps in this title? More like just to show there is some benefit from 5090..? Even if it's driver or settings related and won't change your gameplay experience.

Cherry picking to suit an agenda ?

 

[kondamin]

Why they are asking €70 for the game as it sounds like it's an even more watered down experience since the last one.

 

[TheLostSwede]

Yeah, I use RTSS to keep it below 90 FPS. Radeon Chill kind of works but it's still buggy for some games.

No idea what that is and I have an Nvidia card...

Only sensible workaround appears to be to lower the screen refresh rate in the Windows display settings and using Vsync.

Also, note that there appears to be a UI scaling bug, where 100 and 150% scaling in Windows allows the game to run at full display resolution, but 125% forces some kind of lower resolution, which is really odd.

 

[Dragam1337]

I'd understand if it was between 5000-4000 series but 4090 is particularly bad. It only 18% faster than 4080S which is 10% lower than 28% average in W1zzard's latest reviews. I took time to check, this is THE lowest gap in performance between 4080S-4090 (again based on w1z's latest test). Interestingly 5090 is also exactly 60% faster than 4090 which is THE biggest gap in the reviews. So this the worst game for 4090 in against both 4080 Super and 5090. Which I think is really weird considering it does not look like simple problem like CPU bottleneck or something.
If if was just calculations, surely 60% more CUDA on 4090 would help it more against 4080S. It could explain 5000s' performance but not 4000s'.

I mean it's not really that important. GPU performance in this game is completely pointless anyway. Only metric that matters is turn times. I just found it interesting.

My initial thought was that perhaps the game more or less just scales with vram bandwidth, in which case the numbers for 5090, 5080 and 4090 makes sense... but then throw in all the other gpus (including the 4080) into the mix, and that theory kinda goes out the window.

 

[Onasi]

@TheLostSwede
Wait, does NVidia Control Panel frame cap just not work with Civ VII?

 

[dyonoctis]

No idea what that is and I have an Nvidia card...

Only sensible workaround appears to be to lower the screen refresh rate in the Windows display settings and using Vsync.

Also, note that there appears to be a UI scaling bug, where 100 and 150% scaling in Windows allows the game to run at full display resolution, but 125% forces some kind of lower resolution, which is really odd.

what about this ?

 

[Denver]

You can't. The only option right now is Vsync.

What do you mean? RivaTuner suddenly doesn't work anymore to lock the framerate?

 

[TheLostSwede]

@TheLostSwede
Wait, does NVidia Control Panel frame cap just not work with Civ VII?

what about this ?
View attachment 383874

I guess I have never had any need to use something like that in the past, so I wasn't even aware that there was such a thing.
It doesn't detect I have Civ VII installed though.

What do you mean? RivaTuner suddenly doesn't work anymore to lock the framerate?

That program is still around? Again, I have never had a need to limit the frame rate in any game before.

 

[btk2k2]

No turn time test makes this entire article pointless.

 

[Divide Overflow]

Still not going to touch this with Denuvo DRM.

 

[Onasi]

It doesn't detect I have Civ VII installed though.

Just add the .exe manually then? If for some reason the NVApp doesn’t have such arcane functionality you can always do so in the classic CP in 3D Settings - Program Settings.

 

[TheLostSwede]

Just add the .exe manually then? If for some reason the NVApp doesn’t have such arcane functionality you can always do so in the classic CP in 3D Settings - Program Settings.

Yeah, no such option in the new app.
The old control panel worked though, thanks.

 

[dgianstefani]

Yeah, no such option in the new app.
The old control panel worked though, thanks.

Like this.

 

[TheLostSwede]

Like this.

View attachment 383920

That doesn't work, or at least, it does nothing to limit the frame rate if you select the exe file of the game.

 

[ZenArcade]

No results for the 1080? Not everyone upgrades their video card every year.

 

[mate123]

No results for the 1080? Not everyone upgrades their video card every year.

It has been 84 9 years... Anyway, it should be between the 3050 and 3060

 

[Cheeseball]

No idea what that is and I have an Nvidia card...

Only sensible workaround appears to be to lower the screen refresh rate in the Windows display settings and using Vsync.

Also, note that there appears to be a UI scaling bug, where 100 and 150% scaling in Windows allows the game to run at full display resolution, but 125% forces some kind of lower resolution, which is really odd.

Ahh, I mean Riva Tuner Statistics Server (RTSS). While this is not its primary function, it has a built-in framerate limiter that somehow avoids most framepacing issues when enabled (compared to the NVIDIA Control Panel/Radeon Chill/in-game FPS limiters). Basically, this means I can limit the game to 90 FPS while still using my monitors optimal refresh rate (which is 240Hz for the 32GS95UE at 2160p) and minimize any input lag penalty.

Depending on the game, I either use RTSS or Radeon Chill (and even FRTC before AMD removed it). When I was gaming on an NVIDIA GPU, I used RTSS most of the time.

 

[mate123]

Huh?
View attachment 384157

He was asking for the gtx 1080

 

[W1zzard]

No results for the 1080? Not everyone upgrades their video card every year.

GTX 1080 should be roughly comparable to RTX 3050 performance in the results

 

[Contra]

and even FRTC before AMD removed it

Frame rate target control work perfectly on rdna2/3, also with FG. Where did AMD remove it?

 

[x4it3n]

MSAA running slow is probably a memory bandwidth limitation.

If the game was Memory Bandwidth bound then the 4090 should definitely be faster than the 5080 too, but for whatever reason on this game it is not...

Also the 4090 has 72MB of L2 Cache whereas the 3080 Ti only has 6MB (12x less!!) and the L2 Cache bandwidth on the 4090 is about ~5GB/s whereas the 3090 is only ~2.5GB/s. And the 5080 has 64MB of L2 (less than 4090 too).

So whether it's VRAM Bandwidth or L2 Bandwidth bound the 4090 should be a lot faster than the 3080 Ti and also faster than the 5080.
Imo there is definitely something else going on and I think it's a driver issue.

 

[Cheeseball]

Frame rate target control work perfectly on rdna2/3, also with FG. Where did AMD remove it?

My mistake. Its in the global graphics settings, not in the per game settings:

 

[Sweepi]

1. Thanks for the "Image Quality Comparison" - TPUs Tool is the best one i have seen so far
2. Regarding "High" vs "High +MSAA": The later looks so much better, so many details added (e.g. shadows around trees). Is there a mistake, or did i just forget how good MSAA is?

 

[Noreng]

If the game was Memory Bandwidth bound then the 4090 should definitely be faster than the 5080 too, but for whatever reason on this game it is not...

Also the 4090 has 72MB of L2 Cache whereas the 3080 Ti only has 6MB (12x less!!) and the L2 Cache bandwidth on the 4090 is about ~5GB/s whereas the 3090 is only ~2.5GB/s. And the 5080 has 64MB of L2 (less than 4090 too).

So whether it's VRAM Bandwidth or L2 Bandwidth bound the 4090 should be a lot faster than the 3080 Ti and also faster than the 5080.
Imo there is definitely something else going on and I think it's a driver issue.

W1zzard didn't test MSAA here, only GameGPU did that