I decided to do a bit of a learning exercise today and do my first ever video comparison between two platforms. I picked Tomb Raider since that's what's hot right now. The performance difference between Windows 10 and Linux is quite large in Tomb Raider.
This is my first ever attempt, but I think it turned out reasonably well. The problem I've found is getting a smooth video of the benchmark on Linux, as all recording software made it slightly stutter (with no effect on actual FPS, tested against my previous results). This happens in some games, others are fine.
The second issue is the Tomb Raider benchmark is actually a little different each time with the animations that happen and their timings. Even in this video the timing of the animations at one point is different, but the rest is fine.
To note: This was on the exact same settings, double checked. High preset, motion blur off.
Tested on
Linux Mint 17.3
Intel i7-5960X
Nvidia 980ti
16GB RAM
Also, the game is actually heavier than the benchmark as I noted in my port report.
The performance difference is quite striking, and quite sad to see such a major difference in performance when testing it myself. Considering the power of my test rig, I do expect better.
I have no doubt Feral Interactive will push out future patches to improve things. They do support their titles very well, but it's still sad to see such a massive difference. I didn't even realize the difference was so large until I performed these tests.
This is my first ever attempt, but I think it turned out reasonably well. The problem I've found is getting a smooth video of the benchmark on Linux, as all recording software made it slightly stutter (with no effect on actual FPS, tested against my previous results). This happens in some games, others are fine.
The second issue is the Tomb Raider benchmark is actually a little different each time with the animations that happen and their timings. Even in this video the timing of the animations at one point is different, but the rest is fine.
To note: This was on the exact same settings, double checked. High preset, motion blur off.
Tested on
Linux Mint 17.3
Intel i7-5960X
Nvidia 980ti
16GB RAM
Also, the game is actually heavier than the benchmark as I noted in my port report.
YouTube videos require cookies, you must accept their cookies to view. View cookie preferences.
Direct Link
Direct Link
The performance difference is quite striking, and quite sad to see such a major difference in performance when testing it myself. Considering the power of my test rig, I do expect better.
I have no doubt Feral Interactive will push out future patches to improve things. They do support their titles very well, but it's still sad to see such a massive difference. I didn't even realize the difference was so large until I performed these tests.
Some you may have missed, popular articles from the last month:
With a GTX 970 and the 361.42 driver on Ultra I'm finding the FPS to be disappointing at times. Frequently sub 60 fps, down into the mid 30s for extended periods which really isn't what I'd expect for a three year old game with this GPU (nevermind the i7-4790K and 16GB Ram).
It's playable, and I've never really placed a great deal of emphasis on fps as long as it doesn't interfere with playing the game but ...
Last edited by GBee on 2 May 2016 at 7:19 am UTC
It's playable, and I've never really placed a great deal of emphasis on fps as long as it doesn't interfere with playing the game but ...
Last edited by GBee on 2 May 2016 at 7:19 am UTC
0 Likes
Quoting: omer666Well don't forget nVidia's part in this.
Lately I remember playing Total War Attila (not a Feral port by any mean) and getting awful performance with the drivers it was released for, then everybody was blaming Creative Assembly for not optimising their game engine. Quite recently I updated to 361 series and gave it a try, to my surprise the game was running quite smoothly.
To be fair, I think Feral always do their best, but don't forget, nVidia need to get their drivers updated on time for game releases. Also the game's been out on Windows for more than two years, and it is very often used for benchmarks in the Windows world. Drivers must have been continually optimised in two years time... Hence that big difference between the two OS.
As usual, we have to wait. Not so long ago, I've had been waiting for nVidia to correct a display glitch happening on GNOME for a year and a half.
I think this company really needs to get their things sorted out way more quickly...
I feel like I need to clarify one thing regarding my opinion - I know that Feral may not be the one to blame for performance. But Feral is the one who announces release, and Feral is the one who makes game available for us. And there was no information about known performance issues or they did not say that the game is released, but there is performance update that we should expect. Instead, we have a product, that seems to be final according to all the advertisements, but is not according to performance.
And I just expect those issues to be solved, and Feral needs to take part in fixing process. Even if nVidia is the one to blame.
If this is fixed and performance is comparable to Windows, then I will be happy to give them my money for the full price.
But sadly, currently I have an impression that we have a half product.
1 Likes, Who?
It would also help if Feral published some details about their porting rigs & what hardware they themselves tested the game on.
1 Likes, Who?
QuoteIf this is fixed and performance is comparable to Windows, then I will be happy to give them my money for the full price.
Only you're not paying the full price, the full price on launch was far higher. I know it might seem like a petty point, I know that Windows users can buy it at the same price as we are paying now but the majority bought the game when it was launched at more than twice the current asking price. Or to put it another way, the Linux launch price is less than half the Windows launch price which makes it a bargin even given the performance issues, which as I've already said do not really impact the playability of the game - at no point does it turn into a slideshow or lag noticeably.
If it wasn't possible to monitor the FPS in the game, would people even be raising the framerate as an issue? I doubt it.
2 Likes, Who?
Fun fact: nVidia drops support for the non-recent GPUs much faster than AMD does. AMD supports cards from the same gcn version. The nVidia 780ti which should only be a bit slower then the 780ti performs very badly in modern games (up to - 50% while it is supposed to be 10-20% slower at max).
I wonder if the game runs better with recent AMDGPU Pro drivers if you compare Windows vs Linux.
I wonder if the game runs better with recent AMDGPU Pro drivers if you compare Windows vs Linux.
0 Likes
I say give Feral a break here - it could be something as small as the Steam overlay affecting the results :D
Linux has IMHO never been about high-end hardware so will be nice to see a comparison with lower-end GPU's -besides anything over 80fps most screens are limited by their refresh rate and wont even render those "frames" - worse they actually cause texture tearing and other irritations...
Last edited by Xelancer on 2 May 2016 at 9:20 am UTC
Linux has IMHO never been about high-end hardware so will be nice to see a comparison with lower-end GPU's -besides anything over 80fps most screens are limited by their refresh rate and wont even render those "frames" - worse they actually cause texture tearing and other irritations...
Last edited by Xelancer on 2 May 2016 at 9:20 am UTC
0 Likes
Quoting: XelancerOn the Linux version there is a flock of birds that fly past - YET those same birds don't seem to appear on the Windoze version?? Is there a decal limit or something on the Win10 version, could something like that explain the discrepancy in the results? I know some drivers even hard-limit to polygon counts and reduce polygons considered outside of visible range?Even dedicated benchmarks are rarely identical to the last details. Run the benchmark enough times and you'll notice a bunch of differences between runs (on the same system).
2 Likes, Who?
Quoting: dubigrasuEven dedicated benchmarks are rarely identical to the last details. Run the benchmark enough times and you'll notice a bunch of differences between runs (on the same system).Yes you are right; and when you are talking two distinct technologies (like DX game converted to OpenGL) I suppose those discrepancies will only grow. Sure makes a strong case for Vulkan going forward...
0 Likes
Ultimate quality preset at 1920x1200 on a GTX Titan X with driver v364.19.
Benchmark Statistics:
* Min FPS: 49.2
* Max FPS: 85.1
* Average FPS: 68.0
Benchmark Statistics:
* Min FPS: 49.2
* Max FPS: 85.1
* Average FPS: 68.0
0 Likes
that's really sad. i'm gonna try that on my ancient AMD/AMD-hardware soon, expecting 15-30fps x'D
0 Likes
Quoting: wojtek88I feel like I need to clarify one thing regarding my opinion - I know that Feral may not be the one to blame for performance. But Feral is the one who announces release, and Feral is the one who makes game available for us. And there was no information about known performance issues or they did not say that the game is released, but there is performance update that we should expect. Instead, we have a product, that seems to be final according to all the advertisements, but is not according to performance.
And I just expect those issues to be solved, and Feral needs to take part in fixing process. Even if nVidia is the one to blame.
If this is fixed and performance is comparable to Windows, then I will be happy to give them my money for the full price.
But sadly, currently I have an impression that we have a half product.
No worries, I was not pointing at you by any mean, I just wanted to clarify who's doing what as far as pure performance is concerned, and nVidia is a factor we should not overlook. Their monopoly doesn't encourage them to push things forward and it's a real shame at the time. As their hardware is included in every Steam hardware produced as of now, their Linux drivers should be top priority... But obviously, they're not.
Last edited by omer666 on 2 May 2016 at 7:35 pm UTC
0 Likes
Quoting: Mountain ManActually, the Linux version of XCOM 2 ran comparable to and even slightly better than the Windows version. It was Firaxis who "downgraded" some of the visuals in the performance patch. Feral just ported those changes into the Linux version.Wow, that's quite surprising! Then I guess some hope's restored, though XCOM 2 was released recently and costs much more than Tomb Raider after three years. So it's much more likely to be optimized. The driver issue still stands as I see and unfortunately 35x.xx branch just isn't for me, NS2 crashes with it frequently and Steam crashes as well when changing game category.
View video on youtube.com
0 Likes
Quoting: Guest355.11 has worked very well for me (GTX 660). It’s the 358 series that made Steam crash. But I don’t know about "NS2".Natural Selection 2. I tried 352.79 today, Steam indeed doesn't crash but NS2 does. I suspect that the crash was fixed in 36x.xx on the driver's side as it was present for pretty long time, random crash just during the game. NVIDIA even investigated it and pointed to the game binary but I suspect they made a game-specific fix in the driver eventually like they do for lots of Windows games. And I'm grateful for that.
The performance isn't very different in Tomb Raider between 352.79 and 361.19 actually. Maybe 1-2 FPS at most while in the heavy zone of the map. I had about 15-18 FPS there, the very beginning of the Roth arc. BTW, setting __GL_THREADED_OPTIMIZATIONS=1 helps to get another 3 FPS while loading the CPU for 50% more. Normally I have 200% (it's actually using threading, yay!) and after enabling the optimization I've got 250-280%. To easily switch it on I've created a script like this:
#!/bin/sh
env __GL_THREADED_OPTIMIZATIONS=1 "$@"
and saved it under the name /usr/local/bin/glto, then I put "glto %command%" to the launch parameters of the game and voila.
Too bad Feral still hasn't sorted out the issue with XCOM 2 as it has such a massive performance hit on 361.xx branch. It doesn't really look like a driver-only issue (the new GLVND architecture and such) since other games aren't affected.
1 Likes, Who?
Quoting: rkfgI've created a script like this:Nice idea, thanks.
#!/bin/sh
env __GL_THREADED_OPTIMIZATIONS=1 "$@"
and saved it under the name /usr/local/bin/glto, then I put "glto %command%" to the launch parameters of the game and voila.
0 Likes
See more from me