Total War: Warhammer III Benchmark Test & Performance Analysis

[W1zzard]

That would be totally ok with me if the idea of the article is to see how the game runs on the most used / common cards on the market. if they idea is to show how does it rum against different cards vendors, I think it should be or try to be like 50/50 until intel bring the new GPUS and in that case it should try to be 33.33% each brand. But is just my opinion and I know is hard to please everyone.

intel will release 3 cards only, so 3 intel cards, 3 amd cards and 3 nvidia cards, got it

 

[droopyRO]

That would be totally ok with me if the idea of the article is to see how the game runs on the most used / common cards on the market. if they idea is to show how does it rum against different cards vendors, I think it should be or try to be like 50/50 until intel bring the new GPUS and in that case it should try to be 33.33% each brand. But is just my opinion and I know is hard to please everyone.

AMD cards perform worse in this game than Nvidia ones for the same price category. It's that simple. HUB came to the same conclusion.

 

[pathstrider]

Only note of caution is on the 1% lows - they are the same despite the higher average on nvidia. That’s hiding some big frame time hikes.

kitguru go in to detail and while I’m not sure about their numbers (they seem to be 10% higher than anyone else or what I can duplicate) I can see the frame time hikes on my nvidia cards.

 

[Noreng]

Why does the comparison is 10 Nvidia cards vs 8 AMD cards? would it make more sense to remove the 2080 as there is already a 2080 ti on the chart and add the vanilla 6800. it would make a more fair comparison because for nvidia there is the 3070 and 3080 on the chart.

If any cards should be removed from this list, it's the 3090 and 6900 XT, as they perform very similarly to the 3080 and 6800 XT (for obvious reasons). There's no point in adding the RX 6800 to the list, it's based on the same Navi 21 die as the 6800 XT and 6900 XT, meaning it'll perform slightly slower. I'd say the list paints a pretty decent picture.

If any cards should be added at this point, I'd say it should be the RX 580 and GTX 1060, purely because of their popularity.

Only note of caution is on the 1% lows - they are the same despite the higher average on nvidia. That’s hiding some big frame time hikes.

kitguru go in to detail and while I’m not sure about their numbers (they seem to be 10% higher than anyone else or what I can duplicate) I can see the frame time hikes on my nvidia cards.

I strongly suspect that's because KitGuru tested with the "unlimited video memory" option, it was mostly the 8GB cards and below that struggled. Or possibly first run shader caching issues.

 

[pathstrider]

I strongly suspect that's because KitGuru tested with the "unlimited video memory" option, it was mostly the 8GB cards and below that struggled. Or possibly first run shader caching issues.

Didn't make any difference for me when I tried on or off (which is right given unlimited video memory should reduce performance if anything)- and the KitGuru's figures are almost across the board better.

I may be wrong on this, but aren't shader caching issues more a Dx12 thing?

 

[W1zzard]

and the KitGuru's figures are almost across the board better.

Are they using the integrated benchmark? That's laughably light, gotta test with an actual realistic benchmark scene

 

[pathstrider]

Are they using the integrated benchmark? That's laughably light, gotta test with an actual realistic benchmark scene

They are - I do appreciate that you've gone with a harder test, it was the right choice.

But their results from thier battle benchmark are still better than anyone elses, as far as I can tell. The only difference from your set up (and Hardware Unboxed) seems to be DDR5 memory, but the difference seems a lot for just that.

 

[droopyRO]

Nvidia GPUs also have a clear issue with frame pacing.

They are right about that. Look at how smooth the AMD line is compared to the NV one, that is all over the place.

Are they using the integrated benchmark? That's laughably light, gotta test with an actual realistic benchmark scene

Yes, but their results are repeatable. While only you know how you tested it.

If you get good at this game. Large 2v2 armies battles are not an issue, since you will rarely fight them, except in the Vortex campaign on WH2 when Skaven attack.
And you could always make it worse A siege map with 2 armies defending and 2-4 armies attacking. That will get you sub 30 fps in WH2 regardless of your hardware. Since i think that there are limits to the game engine.

 

[W1zzard]

Large

Yeah, I'm not testing large armies, just a full battle setup with large armies off

 

[droopyRO]

English is not my first language. I meant "large" as in the scope of the battle, not an ingame option.

 

[pathstrider]

They are right about that. Look at how smooth the AMD line is compared to the NV one, that is all over the place.

I just remembered that Gamer's Nexus identified something similar in three kingdoms - at the 15 min mark in this video:

 

[nguyen]

in-game benchmark are not representative of gameplay in WH3 regarding frame pacing, I have flawless frame pacing on my 3090 in actual big battle but the in-game benchmark does show frametime spikes

46FPS avg and 39FPS 1% Low is still very playable, have no problem even at highest battle difficulty for me
Edit: just got owned right after this battle and lost the campaign

 

[pathstrider]

in-game benchmark are not representative of gameplay in WH3 regarding frame pacing, I have flawless frame pacing on my 3090 in actual big battle but the in-game benchmark does show frametime spikes

46FPS avg and 39FPS 1% Low is still very playable, have no problem even at highest battle difficulty for me
Edit: just got owned right after this battle and lost the campaign

It could be you didn’t zoom in during your battle - that’ll minimise the detail so could explain why you don’t get the dips.

Now for me that’s fine too as I rarely zoom in during a fight but some people do.

tbh it’s the campaign map performance that is my major problem- it’s significantly worse than battle and has even more stutters. When it snows in Kislev my 3090 loses about 20fps!

and bad luck with the game, my sympathies

 

[droopyRO]

I hope there are some software optimizations they could do. Both at the driver and game level. But i think this is it for now.
For Ultra details and 60 fps we will have to wait for next gen GPUs. That will give us time to ready our wallets and bank accounts

 

[pathstrider]

I hope there are some software optimizations they could do. Both at the driver and game level. But i think this is it for now.
For Ultra details and 60 fps we will have to wait for next gen GPUs. That will give us time to ready our wallets and bank accounts

Yeah, it's ok - but it's a disappointment given it seems a step backwards in several ways from the last Total War (Troy), technically at least.

And a minor thing, but I'd love to know what's so graphically intensive that the menu takes up 90% of my 3090!

 

[nguyen]

It could be you didn’t zoom in during your battle - that’ll minimise the detail so could explain why you don’t get the dips.

Now for me that’s fine too as I rarely zoom in during a fight but some people do.

tbh it’s the campaign map performance that is my major problem- it’s significantly worse than battle and has even more stutters. When it snows in Kislev my 3090 loses about 20fps!

and bad luck with the game, my sympathies

try force enabling Resizeable Bar for WH3 using NVInspector, it does seem to smooth out stutters in the campaign map for me.

 

[pathstrider]

try force enabling Resizeable Bar for WH3 using NVInspector, it does seem to smooth out stutters in the campaign map for me.

Thanks, unfortunately I got my card before Rebar came out and it only has a single BIOS - given how things are right now I'm not going to risk bricking it so I'll have to leave it for now.

I have a 3060 in another computer, might see if it makes a difference with that....

Similarly, some people are reporting GPU hardware scheduling is helping, but it seems 50/50

 

[W1zzard]

but I'd love to know what's so graphically intensive that the menu takes up 90% of my 3090!

"Intense" depends on the FPS, too, not only GPU load. What's probably happening is that the devs don't limit the FPS in the menus, so it runs super high FPS, while other game limit the FPS to 30, 60 or v-sync

 

[pathstrider]

"Intense" depends on the FPS, too, not only GPU load. What's probably happening is that the devs don't limit the FPS in the menus, so it runs super high FPS, while other game limit the FPS to 30, 60 or v-sync

I’ve got my fps capped at 140.

The flames in the background are good, but not *that* good

 

[W1zzard]

Indeed, 140 FPS and 90% GPU load for that menu is unexpected

 

[arni-gx]

i have TW warhammer 1-2 in my Pc steam, but i have never playing the campaign story mode..... only the custom battles.... its very confusing.... i have thought that this TW warhammer 3 its still confusing too about tall units in the custom battles....