DXVK 0.72 is out, bringing more configuration options and game fixes

I thought games are more compatible with Proton rather than Wine+DXVK. But it turns out the opposite is true. I just tried Batman: AK with Wine Staging 3.15 and DXVK 0.72, and it works! The performance isn't too good though, about 30-50 fps with drops. I had 60-80 on Windows with max graphics settings, twice better. Well, at least I don't have to deal with Windows anymore to play my favourite games, GTA V and B:AK. Thank you YoRHa-2B for this.

Performance is actually pretty good on my machine, even with using the latest official short term driver (396.54), though I don't know what I would get on Windows. Using 396.54.05 (https://developer.nvidia.com/vulkan-driver) should further improve performance. Please also make sure that you CPU governor is set to "Performance".

What happens if i buy a game today on windows, play with proton and next month a linux native port is made? who will take the money?

My understanding was that Feral gets their share as soon as they have officially announced a port, not before. I also guess that this applies only when buying on Linux and not playing on Windows in the first two weeks.

Would love to know too how this would work with Steam Play now.


Last edited by jens on 15 Sep 2018 at 2:25 pm UTC

it seems that shadow of the tomb raider works very well with Proton. Oh man! I really feel sad for Feral now the have zero incentive for a real port.

Well, performance wise, not paying attention to the graphical glitches, according to this video:
https://www.youtube.com/watch?v=0rFZh85TTmI
performance are about 25%...35% lower than on windows, which means on linux it runs between 1/4 and 1/3 less of the original speed.
This is not what i'd call working well, and feral has been bashed for way less performance drop on their ports.

Now, considering the latest tomb rider from feral suffers of an average of **just** 10% performance drop, i can't see them stopping to port it just because of that.

Is your experience with proton different, maybe?

I think Proton and DXVK are nice projects over the original plain wine, and they sure brings more speed and friendlyness to the final user; but i can't consider buying a game made for windows, knowing it will run at 3/4 of the original speed.
I could change my mind if/when it will run at least at 85..90%, not less.

What happens if i buy a game today on windows, play with proton and next month a linux native port is made? who will take the money?

With the latest Vulkan driver from Nvidia there is a huge boost of performance across the board. Most games seem to run between 80 and 90% of the windows performance, i.e. about the same as a good Linux port. Shadow's seems to be at 15-20%. This is more or less the performance hit of the ROTR port compared to Windows.

I did not tried shadow of the tomb rider, but the driver used in the video i linked is Nvidia-396.54 and performance are not that good at all.
(EDIT: (does 396.54.05 really improves it?))
...but still, if i've to run on a lower performance, i still prefer to give my money to the porter, even if the proton wrapped version would perform the same.

My understanding was that Feral gets their share as soon as they have officially announced a port, not before. I also guess that this applies only when buying on Linux and not playing on Windows in the first two weeks.

Would love to know too how this would work with Steam Play now.

Since playing on proton is counted as playing on linux, it would be great if Feral (or anyone) could be rewarded by Valve for their ports with (part of?) the earnings made thanks to linux people playing with proton on that particular game. Given that:
1) they port not too late
2) the quality of the port is good

That would make proton to help new native ports to come rather than the opposite.
What do you think?


Last edited by kokoko3k on 15 Sep 2018 at 2:32 pm UTC

according to this video:
https://www.youtube.com/watch?v=0rFZh85TTmI

... which was made before anyone even had a chance to look at any of the issues. Someone made an updated benchmark comparison with fixed DXVK + new driver, and the visual issues should hopefully be gone with the next Nvidia beta driver update as well (I recorded a video on my AMD system as well, no issues there).

I understand if you prefer a native port (we all do), but let's at least be up-to-date with the comparisons.


Last edited by YoRHa-2B on 15 Sep 2018 at 3:39 pm UTC

according to this video:
https://www.youtube.com/watch?v=0rFZh85TTmI

... which was made before anyone even had a chance to look at any of the issues. Someone made an updated benchmark comparison with fixed DXVK + new driver, and the visual issues should hopefully be gone with the next Nvidia beta driver update as well (I recorded a video on my AMD system as well, no issues there).

I understand if you prefer a native port (we all do), but let's at least be up-to-date with the comparisons.

Self-quoting, sorry i've to:

according to this video:
https://www.youtube.com/watch?v=0rFZh85TTmI
performance are about 25%...35% lower than on windows,

...at 2:47 of the benchmark YOU linked, i'm reading the following:
min: 62.75% (37.25% drop)
max: 83.46% (16.54% drop)
avg: 78.87% (21.13% drop)
So, instead of 25% to 35% lower, is 16.5% to 37% lower?
I don't know what is your way of reading... well, numbers.
I read:
The higher the native framerate, the lower the performance drop, not very useful.
...and when the windows framerate is at his lower value, the linux framerate gap is at his maximum value, URRAH!

I hope you understand by yourself how not ideal this is, but If you're ok with those numbers and still think the wrapped version "works well (tm)", well, good for you!

I really, really hope Feral does not.


Last edited by kokoko3k on 15 Sep 2018 at 5:00 pm UTC

I think the problem that is causing performance loose in Shadow of the Tomb Raider is DENUVO..

Any news when this version of DXVK will be integrated into Proton?

I think the problem that is causing performance loose in Shadow of the Tomb Raider is DENUVO..

I don't think so.
Even here: https://www.youtube.com/watch?v=TYmSoEhJL18
the average drop is still over 20%, and to be honest,when the framerate is higher on windows, sometimes wine preforms even better.
However you can see that when the framerate is low on windows, the gap vs wine/proton is higher (fps are even cutted by half,or more, if you look the first column in the video description).


Last edited by kokoko3k on 15 Sep 2018 at 5:12 pm UTC

according to this video:
https://www.youtube.com/watch?v=0rFZh85TTmI

... which was made before anyone even had a chance to look at any of the issues. Someone made an updated benchmark comparison with fixed DXVK + new driver, and the visual issues should hopefully be gone with the next Nvidia beta driver update as well (I recorded a video on my AMD system as well, no issues there).

I understand if you prefer a native port (we all do), but let's at least be up-to-date with the comparisons.

Self-quoting, sorry i've to:

according to this video:
https://www.youtube.com/watch?v=0rFZh85TTmI
performance are about 25%...35% lower than on windows,

...at 2:47 of the benchmark YOU linked, i'm reading the following:
min: 62.75% (37.25% drop)
max: 83.46% (16.54% drop)
avg: 78.87% (21.13% drop)
So, instead of 25% to 35% lower, is 16.5% to 37% lower?
I don't know what is your way of reading... well, numbers.
I read:
The higher the native framerate, the lower the performance drop, not very useful.
...and when the windows framerate is at his lower value, the linux framerate gap is at his maximum value, URRAH!

I hope you understand by yourself how not ideal this is, but If you're ok with those numbers and still think the wrapped version "works well (tm)", well, good for you!

I really, really hope Feral does not.

Looking at both links @YoRHa-2B posted I see a AAA game that seems perfectly playable on Linux at day1 after release (and would even count as a Linux purchase if I would bought it now via Steam on Linux). This looks very close to ideal to me.

Certainly I would prefer a version from Feral. They would have a completely different approach and possibilities with the source at hand and could provide a super smooth and fast experience. Though I very much doubt that we will see a Linux port from them. Even if they would release their version within half a year from now I doubt that there are much people left that haven't bought the game for Steam Play already. They simply wouldn't earn any money with it.

As I stated in the LiS:BtS thread, I very much hope that I'm wrong, but I think we wont see much from Feral for Linux in the future. Their focus will shift to Mac an iOS. May be, may be, their Linux skills will be asked again when Linux has reached such a market share that developers and publishers take Linux into account from the beginning on. It is a long road to that, lets see how it turns out. I guess Steam Play is our best bet for that to happen.


Last edited by jens on 16 Sep 2018 at 8:40 am UTC

Just liking the fact more games are playable on Linux, especially ones we know will never get ported.
All I can safely assume these days is if a developer hasn't said anything about Linux then it wont be.
If the game works with proton, great! it will count as a Linux purchase otherwise oh well wine it is.
At least we have more options now!.

...at 2:47 of the benchmark YOU linked, i'm reading the following:
min: 62.75% (37.25% drop)
max: 83.46% (16.54% drop)
avg: 78.87% (21.13% drop)
So, instead of 25% to 35% lower, is 16.5% to 37% lower?
I don't know what is your way of reading... well, numbers.
I read:
The higher the native framerate, the lower the performance drop, not very useful.
...and when the windows framerate is at his lower value, the linux framerate gap is at his maximum value, URRAH!

You're forgetting the performance drop from recording, also counting the min fps can give you the wrong impression. I can get 10 fps because of shader caching at the beginning of a benchmark and then have 70fps for the rest.
40 fps average on Ultra is pretty good IMO.(I would probably disable Motion Blur)
I hope proton doesn't affect Feral in any negative way, but unless Linux usage/sales increase on steam it might just do that.
EDIT: Also, those games and NVIDIA's driver are optimized for Windows, not for Linux. So Feral have the advantage of working with driver developers.


Last edited by BOYSSSSS on 16 Sep 2018 at 8:32 am UTC

...at 2:47 of the benchmark YOU linked, i'm reading the following:
min: 62.75% (37.25% drop)
max: 83.46% (16.54% drop)
avg: 78.87% (21.13% drop)
So, instead of 25% to 35% lower, is 16.5% to 37% lower?
I don't know what is your way of reading... well, numbers.
I read:
The higher the native framerate, the lower the performance drop, not very useful.
...and when the windows framerate is at his lower value, the linux framerate gap is at his maximum value, URRAH!

You're forgetting the performance drop from recording, also counting the min fps can give you the wrong impression. I can get 10 fps because of shader caching at the beginning of a benchmark and then have 70fps for the rest.

Well, the performance drop due to the recording should affect both systems, not just Linux.
Also, DXVK has an on disk cache, and it is very unlikely that the benchmark has been done for the first time and recorded it, so the shader cache should be already warmed up.

Well, the performance drop due to the recording should affect both systems, not just Linux.
Also, DXVK has an on disk cache, and it is very unlikely that the benchmark has been done for the first time and recorded it, so the shader cache should be already warmed up.

He didn't record windows, he just said it's 86% of what windows gets.
About shader cache, my point still stands. If fps dips for a second for some reason on a 2 hour gameplay that wouldn't matter to me.

Well, the performance drop due to the recording should affect both systems, not just Linux.
Also, DXVK has an on disk cache, and it is very unlikely that the benchmark has been done for the first time and recorded it, so the shader cache should be already warmed up.

He didn't record windows, he just said it's 86% of what windows gets.
About shader cache, my point still stands. If fps dips for a second for some reason on a 2 hour gameplay that wouldn't matter to me.

!?
"They", all did:
https://www.youtube.com/watch?v=qubt4vrrh_E
https://www.youtube.com/watch?v=TYmSoEhJL18
https://www.youtube.com/watch?v=0rFZh85TTmI
What are yyou referring to?

"If fps dips for a second", it does not matter to me too, the point is that is very unlikely one makes a benchmark and record it with a cold shader cache, so the minimum fps is NOT due to that, unfortunately.

!?
"They", all did:
https://www.youtube.com/watch?v=qubt4vrrh_E
https://www.youtube.com/watch?v=TYmSoEhJL18
https://www.youtube.com/watch?v=0rFZh85TTmI
What are yyou referring to?

I was referring to the video of YoRHa-2B's comment you were responding to https://youtu.be/5Kd7hAshw-4

"If fps dips for a second", it does not matter to me too, the point is that is very unlikely one makes a benchmark and record it with a cold shader cache, so the minimum fps is NOT due to that, unfortunately.

Maybe no, but it is due to recording performance drop. Like I said "You're forgetting the performance drop from recording, also counting the min fps can give you the wrong impression." and "He didn't record windows, he just said it's 86% of what windows gets."


Last edited by BOYSSSSS on 17 Sep 2018 at 1:31 pm UTC

YoRHa-2B also left a comment in the video
"Just ran some tests with the settings shown in this video:
- Windows DX11: 49 FPS
- Windows DX12: 55 FPS
- Linux DXVK: 42 FPS (without recording)"
And this is at Ultra preset. It's good enough IMO.

Great results.
Maybe on nvidia (mine card, and the one used in the videos i linked above) the gap is a bit higher.