User Info RequiredLow fps during battles

The_Bad_Wolf

Hello!

I have a Win 10, i7-2600K (2x 3,4 GHz), GTX 970 4GB. I can run Witcher 3 on max with no problem, AC Unity as well and Crysis 3 at least on very high (that was on my older GPU.. I guess now it can only be better)

But when I play Warhammer or Rome 2 I get about 60-70 fps in benchmark, before the battle I get also around 50, but when I close up or the battle begins I get from 5-20, at times 30 fps max on Ultra setting (it was recommended for me). In Rome 2 it's slightly better on low settings, but in Warhammer it's all the same no matter on what settings. Even in Shougn 2 in massive battles (like 5-10k) I have 5 fps close to the battlefield. About 20-30 from above.

Total war games are really the only games I have problem with. Is it the problem with my aging CPU, since I've heard that these games are CPU heavy and I should buy a better one, or is it something on the games' side?

Bsham

Unit detail, shadows, AA, grass and trees are usually the cause. Unit detail and shadows especially in any TW game.

If you put everything to max performance but shadows to performance and terrain detail to quality coupled with SSAO, ultra unit size and msaa, it plays well and still looks good.

JinShepard01

Biggest culprit is that the recommended settings/gaming pc(From CA) is very off in Warhammer. The best thing is to toy around with settings.

HuntingDog

The_Bad_Wolf said:

Total war games are really the only games I have problem with. Is it the problem with my aging CPU, since I've heard that these games are CPU heavy and I should buy a better one, or is it something on the games' side?

Single CPU core speed is almost everything in TW games. Your CPU isn't the latest but there wasn't a lot of progress in CPU speed per clock cycle. The latest Skylake CPU will be only about 20% faster with the same clock speed as your
sandy bridge CPU.

Fortunately for u, u have one of the best CPUs for overclocking. If u haven't overclocked your CPU yet, I would strongly advise to explore that route first. A 2600K can often overclock to 4.5Ghz or more. Even a mild overclock of 4.2 Ghz would be noticeable improvement over stock speed.

There are some preparations to be made (checking motherboard for OC capabilities, getting better cooling for middle to high overclocks, ...) but could be really worth it.

Another possible worthy upgrade for TW games is the RAM. TW games are one of the few games which can actually benefit from higher RAM speed. Should u still have 1333Mhz RAM, upgrading to sth like 1866RAM (not sure about max supported speed for your mobo) with low latencies can actually give a noticeable boost to
minimum framerates.

And as bsham said, there are some settings which have a bigger influence on CPU performance. These are:
- unit size (obvious since with fewer units there is less to calculate)
- shadows (especially in older TW games it has a big influence when zooming in)
- unit detail (units are rendered with less detailed animations/polys/shadows)

JinShepard01

Unit size being CPU I get, but shadows and unit detail, isn't that GPU? While the size increase, ofc does the demand on the GPU then, but asaik, they aren't CPU bound.

HuntingDog

Trust me, I tested the shadow setting a lot in former TW games (haven't tested it in Warhammer yet). It always had a big influence on CPU performance when zoomed in to a lot of units.

Maybe that's also the reason why CA did add an extra warning sign to the 'extreme shadow' setting in TW-WH

As to why, I am not sure. That would be a question for a graphics engine programmer. But unit detail is related to shadows, because with lower unit details u will have fewer complex unit models and therefore fewer complex shadows. And maybe with lower unit detail u also get fewer units with their own shadow.

JinShepard01

i7-4790K @4.4 ghz, gtx [email protected], 16 GB Ram, Asus MAXIMUS VII RANGER

For reference, on Ultra settings ( shadow extreme, AA and SSOA on, on a 980ti (sadly just died =( ) I had 44 fps on average )

Now iam playing on a 1440p monitor and these are the tests I made (toke a while to write..)

1. Shadow: Extre Unit: Ultra AA: 8x SSOA: On FPS: 26.2
   
2. Shadow: Ultra Unit: Ultra AA: 8x SSOA: On FPS: 26.5
   
3. Shadow: Extre Unit: Large AA: 8x SSOA: On FPS: 26.6
   
4. Shadow: Ultra Unit: Large AA: 8x SSOA: On FPS: 26.8
   
5. Shadow: Extre Unit: Ultra AA: Off SSOA: On FPS: 45.2
   
6. Shadow: Ultra Unit: Large AA: Off SSOA: On FPS: 46.7
   
7. Shadow: Extre Unit: Ultra AA: Off SSOA: Off FPS: 49.4
   
8. Shadow: Ultra Unit: Large AA: Off SSOA: Off FPS: 51.3
   
9. Shadow: High Unit: Large AA: Off SSOA: Off FPS: 51.6
   
10. Shadow: Ultra Unit: Ultra AA: Off SSOA: Off FPS: 50.1
    

What do you make of this?

HuntingDog

Hi, nice test with GPU heavy (MSAA, SSAO) and CPU heavy settings (unit size, shadows).

Did u use the built-in benchmark ? The benchmark is not the best CPU test because there are not really a lot of units around and the camera essentially never zooms in on a lot of units. A big 2v2 battle in the campaign will have a much bigger CPU load, especially zoomed in.

All your results with 8xMSAA and SSAO enabled on a 970 @ 1440p will be clearly GPU limited (maybe even the VRAM will already start to fill up). So essentially don't see a difference between unit sizes and shadow quality there.

In the tests with MSAA but with SSAO there is already a small FPS difference of 1.5fps or about 3% between extra shadows + ultra size and ultra shadows + large unit size.

And with SSAO off, there is a 2.2 or 4% fps difference between shadow high + unit larg and shadow extreme + unit ultra.

Due to the GPU heavy nature of the benchmark (I assume u used it), most sequences of the benchmark will still be GPU limited (mostly because of 970 @ 1440p) and that's why we only see a small difference in performance which can probably be mostly attributed to the CPU.

If u want to test a more CPU heavy scene, try something similar as shown here (u can see that GPU utilization is only 68%, pointing to a CPU limitation):

It's from this nice performance review of TW-WH:
dsogaming.com/pc-performance-analyses/total-war-warhammer-pc-performance-analysis/

JinShepard01

Aye the ingame benchmark is what i used, in reality it is much more demanding, that I am sadly aware of, wasn't my choice btw to use a 970 for 1440p lol, but alas that my 980 ti died.

I did some more testing... rofl.. (totally hijacking the thread but it gives OP also a good idea what to do or not :P)

Settings were Ultra and then I turned off AA and SSOA and Ultra shadow with Ultra unit. What i did was turn something lower or off, and the next test I turned it back on and try the next one on/off or High instead of Ultra..

FPS: 50.1 Unlimited Memory off
FPS: 51.1 Distort off
FPS: 51.3 Texture Quality High
FPS: 52.2 VFX High
FPS: 52.2 Depth of Field off
FPS: 59.6 Screen Space Reflection off
FPS: 53.0 Texture filtering Trilinear
FPS: 50.2 Water detail High
FPS: 50.4 Grass Detail High
FPS: 50.1 Proximity Fading off
FPS: 50.2 Buildings High
FPS: 50.2 Terrain High

So for benchmark test wise best performance while getting the most eye candy would be set it on ultra, then turn off AA, SSOA and Screen Space Reflection and I would get 59 fps.

Then a actual real life pic ( horrible quality since its a steam pic, afterburner didnt pick it up when screenshotting. ) 23 fps WITH High shadows and large unit count, meaning I even went further down then the benchmark settings that gave 59 fps ( so you see how useless that benchmark actually is xD )

The_Bad_Wolf

Hey everyone!

Thanks for all the responses, I appreciate it. Sorry for the late post, but I had no connection to the internet for the past week. I will try different options and see if I can find something satisfying. I will surely look into the 'overlocking' HuntingDog mentioned. I will have to do some research about it first though
Also some informative posts on your side JinShepard01, cheers

Thanks again!