Hello, I've been running TSW4 for some 1-2 months and while it's running quite fine on my main PC, as for my laptop I can clearly see the performance of TSW4 is definitely lower than that of TSW3 using the same/similar settings. Or I should say the general FPS values are comparable, but TSW4 has way more stuttering than TSW3 does which ruins the experience. If you say that the reason is TSW4 same level settings provide higher details I am not sure of that, I do not observe any visual difference, it's just that where TSW3 is basically butter smooth TSW4 has noticeable stutters (same route, same date, same weather). Has anyone else noticed that? Is there any setting (in the ini files perhaps) that I can change to make TSW4 closer to TSW3 performance wise? I'd like to remove TSW3 from my laptop for good, but this is preventing me from doing that... Any help is greatly appreciated.
Have you used any custom .ini settings in TSW3? Specifically "create shaders on load"? This would explain it. Having this parm in both, TSW3 and TSW4, I experience about the same and comparable level of stuttering between the two.
Not to my knowledge. Could you elaborate on what this setting is? It may be something I did in the past and just don't remember.
If you added it on purpose, then you would have probably known about it. But it is possible that you added it together with a bunch of other engine.ini parameters in a pack like this for example https://forums.dovetailgames.com/threads/definitive-stutter-fix-via-engine-ini-settings.70559/ : this was one of the first ones to use the: [SystemSettings] r.CreateShadersOnLoad=1 parameter, which reused the previously compiled shaders from cache, instead of compiling them every single run. A game changer imho. If you do not use it yet, then I suggest to try it, however not just on its own, but as part of the package of other parm enhancements, like for example: https://forums.dovetailgames.com/threads/enhancement-lighting-graphic-enhancement-v4-2.74685/
So after more testing with vanilla TSW 3 and 4 installations I can say that: - on a powerful PC the performance difference is rather small or hard to spot. It may also be that this PC is on 60 fps most of the time so there is no space to make the difference visible... - on a weaker laptop there is a significant difference both in terms of FPS and especially overall stutter using settings that are as close as possible between the two games. Certainly there is no apparent visual difference that would justify the performance hit So unfortunately not a good direction with TSW4...
Interesting - is this on like for like routes and trains etc? I'll point the team to this thread see if they can explain what the cause is. Matt.
Hi Matt, yes, it is of course on the same route, same scenario, same place, same date and same weather (and the same graphics settings). It seems to be affecting all routes I tried (Niddertalbahn being one of them). I am getting slightly lower FPS on average, but more importantly there are frequent stutters and dips in the framerate in TSW4 that do not exist (or are much better in) TSW3. Both are running with DX11 as far as I know, but I'll doublecheck this.
So my understanding is that both TSW3 and TSW4 are running on DX11 by default and DX12 must be specifically turned on via a command line argument -DX12, is that correct? If so then I am running on DX11 on both. I can try to see what will change with DX12, but I would expect that DX11 - as the more stable of the two options - should provide comparable results across TSW release on any hardware...?
For a brief moment, TSW4 was switched to use DX12 as the default... and I'm not getting 100% clarity from the team as to whether that change stuck or not. So first question really was - are you using the dx12 launch parameter or not because if not, then I can simply chase down the answer to what the default is and then we can proceed from there. Yes DX11 ought to be smoother at least for the shader compilation issue, we found DX12 hitches a lot more even if the FPS can be higher overall in some cases. Matt.
Understood, so no, I've never used the -DX12 parameter so far. Is there a similar parameter to force DX11 to make sure one is running that?
Without a parameter, the game uses DirectX 11. 12 is as stable as 11 but performing much better, it can only lead to flashing shadow artefacts on some locations when the sun is at a low angle. (Which is seldomly happening, so no big deal for me, as the Pros largely outweigh this small glitch.)
Are you sure of that? Please see Matt's comment above. TSW4 did use DX12 as the default a while back. The question is if it still does or not.
I can't make anything out of Matt's statement. DX11 has stuttering and hitches whilst DX12 is much smoother on my system and many others reported here.
Everyone's system is different. We have feedback that it makes some systems run it worse, no difference on some, and significantly better on others. Matt.
I've checked for you. By default, without any parameters, TSW4 is using DirectX 11 (using -dx12, the D3D12.dll and D3D12Core.dll get loaded additionally to the DX9 and 11 dlls.)
Ok cool thank you - that means the change of default did indeed get completely reverted which is what I thought. So yes you're good with your tests jstange as it means yoru tsw3 and 4 examples are both on dx11. I'll feed that to the team as well and try and get some results in our tests. Thanks! Matt.
I think the advantage of the enhanced and optimized DX12 command set highly depends on your GPU, as workload is significantly shifted from CPU to GPU under 12.
Ok with all tests running on DX11, I got one of the beta team to do a test starting with a wiped nvidia cache, then run TSW3 twice, then wipe cache, then run TSW4 twice. Showing pre and post cache hitching, and equivalents between TSW3 and TSW4. I won't go into specific numbers as its largely irrelevant - the hitching analysis tool only works on development builds and they'll always hitch a lot more than a shipping build, but in theory, the relationships between them should still be relevant. We are just starting to analyse this so as I get more data and such we can dig further into it, but wanted to share some early feedback. All tests were done on the exact same content, with the exact same DLC installed, on the exact same PC. We also picked content that has NOT seen any uplifts / changes between TSW3 and TSW4 to try and keep the tests as like-for-like as possible. So - first run to second (empty shader cache to filled shader cache), TSW3 is about 27% less hitched from first run to second. TSW4 is about 37% less hitched, first run to second. First run, they are broadly equivalent - there's literally 6 hitch difference between TSW3 and 4 in the first runs here, but second runs, TSW4 comes in with 14% less hitches than TSW3. Bottom line - these tests show TSW4 coming in quite a lot better on hitches on second run, broadly equivalent on first and benefit massively from the cache on subsequent runs. As I say, we are still looking into it further, this is one test result. Matt.
