Overhead of Proton
skinmarquee Mar 31
Let's say Steam recommends an NVIDIA xyz. If I look on UserBenchmark, I could see that a abc has a certain % performance increase over xyz. What about is the magic number where I'll be able to run the games well? Like if it's 50% higher than Steam recommendations, or should it be 75% higher? Is there any way to make sense like this?
Pengling Mar 31
Proton translates Windows calls into Linux ones - the impact is negligible. You certainly don't need to be looking at 50% to 75% better hardware than the system-requirements to use it.
CatKiller Mar 31
There isn't a magic number.

Some games run a bit worse through Proton than on Windows, some run a bit better.

Trying to squeeze in at the minimum requirements is a bad time regardless of platform.

What you will get with Proton is slightly higher CPU load (something needs to be doing the translating), but Linux being lighter to run than Windows generally more than makes up for that.
skinmarquee Mar 31
I was thinking it would be significantly higher. Maybe this game I'm playing is more demanding than I thought.

I'm planning to upgrade my CPU from a i5 4570 to an i7 4790 soon. Maybe that'll offset some of the translation. I could try and build something better, but it'd take a year or two at least... I don't have the patience right now and it's like a classic car, might not have all the luxuries but there are thousands of games I can play.

It seems like anything DirectX 9 is a go for sure on this.
Linas Mar 31
No, seriously, the hardware requirements for running games via Proton and on Windows are more or less the same.

In my experience the games usually fall into 2 categories: they either work just as well as on Windows or don't work at all (or have serious issues). You need to consult the Steam Deck rating on Steam, and/or ProtonDB to see which one it is.

Proton is also still being worked on and more and more games are getting support, and existing ones are still getting performance improvements. E.g. when I first tried A Hat in Time, it was barely usable, barely getting 30 fps, and nowadays it is butter smooth on the same hardware.

You also want to keep your system (kernel, drivers) fairly up-to-date to get all the improvements.
Quoting: skinmarqueeI'm planning to upgrade my CPU from a i5 4570 to an i7 4790 soon. Maybe that'll offset some of the translation. I could try and build something better, but it'd take a year or two at least... I don't have the patience right now and it's like a classic car, might not have all the luxuries but there are thousands of games I can play.

Unless it's a very good deal I wouldn't waste the money. Even with 4 extra threads, you won't see a big difference.
skinmarquee Mar 31
I make games, so 15% improvement and 8 threads will save me a lot of time. I also plan to upgrade to 32 GB of DDR3 since it's cheap.
Maybe you're right though. Looking at the condition of used CPUs on eBay, I'm not so confident. Maybe I'll just upgrade the RAM. That's the last thing I wanted to do with it.

I got the SSD mounted properly with those fat HP screws, got a 1030 and put a new DVD drive (couldn't find anything but cheap looking blurays in the form factor) and then I'll upgrade the RAM and I think it'll be as good as it can get.

Anyhow this opens up many more games to me, knowing they are basically equivalent. Then I'll just have to buy some games. The other thread about Batman games is interesting.

Last edited by skinmarquee on 2 April 2024 at 12:28 am UTC
Even though your question has already been answered I just came to notice this thread just now and like to share some Benchmark I coincidently made today using the Superposition Unigine Benchmark. I ran it natively on Linux with OpenGL as well as using Wine and DXVK and Wine and OpenGL.

Out of all Benchmarks then Wine Benchmarks scored the best. Not by much but they did while the Wine OpenGL Benchmark is in the margin of error as it only differentiates from the native OpenGL Benchmarks by 1 - 2 fps.

Unigine Superposition Benchmark on Linux native and Wine

Even though this in one singular application and therefore not necessarily representative for any possible game or application out there it still shows that the overhead by Wine and Proton is often times in a margin of error and in fact can be ignored entirely in like 90% of all cases.

There are however new thing in the making (ntsync ) which will further improve game performance on Linux due to a kernel driver resembling the way a Windows operating system does handle threads and syncing processes.

Speaking of overhead I like to bring in native games not running via Proton or Wine into the discussion here too:
If a game is natively available on Linux it is not guaranteed that it will perform the best. There are numerous games out there featuring a not so well made Linux port. "Not so well made" means that the Windows version of the game ran better than the native version.

Last edited by Vortex_Acherontic on 2 April 2024 at 10:37 am UTC
Here's why I posted this in the first place. The Room 4: Old Sins lists only Generic DX11 on its Steam page, so I figured a GT 1030 should be more than enough since that usually means Intel in my experience. But even on Medium it runs sub-optimally. It's fine 99% of the time, but far from 60 FPS. I've only seen one puzzle that made it somewhat jank to play though.

It is a really good looking game though, probably more advanced PBR shaders and stuff than many games.

Last edited by skinmarquee on 2 April 2024 at 10:53 am UTC
Quoting: skinmarqueeHere's why I posted this in the first place. The Room 4: Old Sins lists only Generic DX11 on its Steam page, so I figured a GT 1030 should be more than enough since that usually means Intel in my experience. But even on Medium it runs sub-optimally. It's fine 99% of the time, but far from 60 FPS. I've only seen one puzzle that made it somewhat jank to play though.

It is a really good looking game though, probably more advanced PBR shaders and stuff than many games.

Looks like a Unity Engine game. According to PCGamingWiki Unity 2019. I recall some random performance issues with "older" Unity versions on Linux due to the mentioned threading. It might help to play around with some launch options:

PROTON_NO_FSYNC=1 PROTON_NO_ESYNC=1 %command%

Any combination of either or both set to 0 or 1. It might be possible to mitigate the issue to some extend.
Also I hat occurrences of games which I had to limit the amount of CPU cores visible to them using:

taskset -c 0,1,2,3 %command%
This will limit the game to run on your 1st (0), 2nd (1), 3rd (2) and 4th (3) thread / core. Don't really know how many threads / core your CPU has.

Maybe you can with any combination of these three get better results. But it is not guaranteed though.

Launch option with all 3:

PROTON_NO_FSYNC=1 PROTON_NO_ESYNC=1 taskset -c 0,1,2,3 %command%

Last edited by Vortex_Acherontic on 2 April 2024 at 11:27 am UTC
While you're here, please consider supporting GamingOnLinux on:

Reward Tiers: Patreon. Plain Donations: PayPal.

This ensures all of our main content remains totally free for everyone! Patreon supporters can also remove all adverts and sponsors! Supporting us helps bring good, fresh content. Without your continued support, we simply could not continue!

You can find even more ways to support us on this dedicated page any time. If you already are, thank you!
Login / Register


Or login with...
Sign in with Steam Sign in with Google
Social logins require cookies to stay logged in.