Performance fix found!

[Jack_Spade]

I've spent days with idtech5tweaker trying to find the optimum settings to make The New Order and The Old Blood run at a playable framerate on my AMD hardware, and I got results I considered... acceptable. Not ideal. But there's a wrapper that translates D3D9/10/11 into Vulkan for Linux... and it also works for Windows!

It's called DVXK, and I went from turning down settings below their in-game minimum, to running at 60 FPS on max settings! Give it a try, see what you think.

[Cryiox]

DXVK does nothing for this game, given it's running on OpenGL, so DXVK doesn't have any DirectX API call to modify ...

[Jack_Spade]

Originally posted by Cryiox:

DXVK does nothing for this game, given it's running on OpenGL, so DXVK doesn't have any DirectX API call to modify ...

I don't know what to tell you. I had framerate issues, I put the dlls in the game folder, issues went away. I have no explanation for how it worked. I'm at a loss. But try it and see if it doesn't do the same thing for you.

[Pro cnr]

Like Cryiox said the game's running OGL, so adding the dll has literally no effect and actually these are not used, if you have better performance it's 100% placebo.
It can works for some games but the games have to run on Direct 3D 9/10/11 (and even 12 using VKD3D instead of DXVK)

[Cryiox]

If only VKD3D was on Windows as well.

I can see how DXVK is very useful in translating DX9/10/11 to Vulkan and getting some unexpected performance or frametime improvements, but DX12 to Vulkan? Aside from being a compatibility layer for Linux, not sure it does anything for performance.

[Jack_Spade]

Originally posted by Pro cnr:

Like Cryiox said the game's running OGL, so adding the dll has literally no effect and actually these are not used, if you have better performance it's 100% placebo.
It can works for some games but the games have to run on Direct 3D 9/10/11 (and even 12 using VKD3D instead of DXVK)

You're arguing philosophy and I'm arguing science. It shouldn't work. I know this. I'm as surprised as you are. But it's not a placebo, I've measured it. Framerates are a thing that can be quantified and recorded. If you want, you can try it. Get some hardware that struggles to run the game at 60 fps-- I believe this only happens on AMD/Radeon hardware-- and try the game both ways. Record your framerate. Don't change any of the graphics options between runs. Don't change graphics drivers or settings in the Radeon overlay. Don't change any variables except the presence or absence of DXVK 64-bit .dlls in the game install folder.

If you are unwilling to do this, then you are only speculating and it is dishonest to present your position as fact. I'm not saying this (just) because I'm touchy, I actually do welcome more data. Might be something peculiar to my machine. But opinions aren't data.

Can we try this? If you really want to shut me up, the fastest and easiest way to do this will be to prove me wrong. And if you do prove me wrong, I'll happily eat crow, provided your methodology is sound. For clarity, that would be using AMD/Radeon hardware and getting a slow framerate, then running DXVK with no appreciable improvement. The Steam overlay even lets you monitor FPS in real time, so it isn't subjective. There's also FRAPS if you want to log it.

[Cryiox]

To be fair, personally I did give it a shot regardless. Nothing changed on my hardware, 5700 XT / R5 3600 and I checked with RivaTunner.

[Jack_Spade]

Originally posted by Cryiox:

To be fair, personally I did give it a shot regardless. Nothing changed on my hardware, 5700 XT / R5 3600 and I checked with RivaTunner.

Ooh! My apologies. I'm running an RX 580 (8GB) and a Ryzen 5 2600. I was getting around 40 FPS at lowest settings, I ended up using idtech5tweaker to get a pretty consistent 50 fps by reducing VRAM-intensive settings below their in-game minimum. With DXVK, I get a consistent 60 FPS (cap) at in-game low to medium settings, but there's still some occasional chop as the game loads new areas. CPU usage seems to have increased, based on temps (didn't think to log CPU usage.) Also, two crashes when I tried to max out pages-per-frame, no crashes when set at 16.

What kind of frames were you getting beforehand? If it was around 60, you wouldn't see much improvement because the game caps at 60. And you did use the 64-bit version, I assume? Based on the results I got, I can't deny that DXVK was a dramatic improvement for me, but I'd be willing to accept that it was a happy accident due to some unforeseen interaction, like maybe breaking some process that was eating up VRAM, or that was taxing the processor.

Thank you for taking the time to try it. :)

[Cryiox]

The only thing that properly made performance decent was the dumb trick with renaming the exe. I don't think it matters what you rename it, as long as it's not the original name. After that, I mostly had a 60 fps locked experience. with drops to 50 here and there.

[Pro cnr]

Originally posted by Jack_Spade:

If you are unwilling to do this, then you are only speculating and it is dishonest to present your position as fact. I'm not saying this (just) because I'm touchy, I actually do welcome more data. Might be something peculiar to my machine. But opinions aren't data.

Can we try this? If you really want to shut me up, the fastest and easiest way to do this will be to prove me wrong. And if you do prove me wrong, I'll happily eat crow, provided your methodology is sound. For clarity, that would be using AMD/Radeon hardware and getting a slow framerate, then running DXVK with no appreciable improvement. The Steam overlay even lets you monitor FPS in real time, so it isn't subjective. There's also FRAPS if you want to log it.

You definitely don't know what you are talking about, and if you want proof you actually have them on you computer, or in fact this is the abscence of these that is the proof

Go check your game files you'll see no log files prooving the dll were used
There has to be some log files at least for the initialization of the wrapper. It generate a log file for any dll used as follow [exe name]_[dll name].log, even if the lib is unused in the game and it's just loaded then it must have created these log files.
For exemple, if you play Yakuza 4 Remastered with dxvk you will have 2 log file generated because 2 dll are used
Yakuza4_d3d11.log and Yakuza4_dxgi.log
Another way to check if the dll are used or not is by deleting these while playing, this is not possible if the process use the library
Another trickier solution but if you know this tools it's easy, by using process explorer to see the dependencies if they're used or not, (spoiler alert they're not)

And can you tell me what version of dxvk you're using ? It actually matters

[Jack_Spade]

Originally posted by Cryiox:

The only thing that properly made performance decent was the dumb trick with renaming the exe. I don't think it matters what you rename it, as long as it's not the original name. After that, I mostly had a 60 fps locked experience. with drops to 50 here and there.

See, I tried that and I got nothing. Tried it with x32, tried it with x86, no change. Idtech5tweaker was a workaround, it allowed me to limit texture size and pages-per-frame, but that still only got me to ~50 fps.

Originally posted by Pro cnr:

You definitely don't know what you are talking about, and if you want proof you actually have them on you computer, or in fact this is the abscence of these that is the proof

Go check your game files you'll see no log files prooving the dll were used
There has to be some log files at least for the initialization of the wrapper. It generate a log file for any dll used as follow [exe name]_[dll name].log, even if the lib is unused in the game and it's just loaded then it must have created these log files.
For exemple, if you play Yakuza 4 Remastered with dxvk you will have 2 log file generated because 2 dll are used
Yakuza4_d3d11.log and Yakuza4_dxgi.log
Another way to check if the dll are used or not is by deleting these while playing, this is not possible if the process use the library
Another trickier solution but if you know this tools it's easy, by using process explorer to see the dependencies if they're used or not, (spoiler alert they're not)

And can you tell me what version of dxvk you're using ? It actually matters

See, that's the thing: I'm not claiming to know what I'm talking about. I know that when I do a certain thing, I get a certain result. The process between that input and that output is a black box to me-- maybe somebody smarter than me can explain why that is. I was even hoping it would be you, since you seem knowledgeable. What I won't do, is deny what I've seen because it doesn't jive with how I understand things to work. You change the model to fit the data, not the other way around.

Since you're interested, I'm using dxvk 1.9, 64 bit. I see no log files, which must mean that it's not working as designed. However, it is clearly doing something because it's the only variable. As I said earlier, it's possible that it's just breaking something, and that something was causing problems. I don't know. Even if that's the case, it's still a desirable outcome. There is one other thing I can check, if you'll give me half an hour.

[Pro cnr]

You obviously did some other thing that did the trick.
If you did A, B and C actions at the same time that does not imply that All 3 action were actually useful, that's my point from the beginning. It can't be that, there is absolutely no way because that's not how it works, you don't just put some random libs for them to be magically used from no where.

When I asked the dxvk version I had another thing in mind, like Did you use the dxvk from the original github ? (this is the one https ://github .com/doitsujin/dxvk)

[Jack_Spade]

Originally posted by Pro cnr:

You obviously did some other thing that did the trick.
If you did A, B and C actions at the same time that does not imply that All 3 action were actually useful. That's my point from the beginning, it can't be that, there is absolutely no way because that's not how it works, you don't just put some random libs for them to be magically used from no where.

When I asked the dxvk version I had another thing in mind, like Did you use the dxvk from the original github ? (this is the one https ://github .com/doitsujin/dxvk)

I get what you're saying, what I'm saying is I can't think of anything else. Except for Id5tweaker, obviously, but I got rid of that to test. I'm not sure what's happening. I suspect you're right, but I can't figure out what happened. The only other thing I can think of is that maybe my display driver updated automatically... there was a Windows update around that time, also. Those are two things I haven't accounted for.

An hour later...
Just checked, it runs about the same either way. Maybe it was a driver update? I know I wasn't imagining a ~15 fps jump. False alarm, everyone.