Running Benchmarks On Newer, Popular AAA Linux Games***

[Michael]

On a somewhat related note - can you run these tests often in what you report? I know you said you have a CS:GO test profile, some others for some Valve games, and you recently did one for Civ5, but you never really seem to run any of those tests. At most you seem to occasionally have a "special" article where you only include those tests and not your normal ones.

They aren't used too often due to Steam... With regard to issues caching, Steam breaking the test profiles routinely, and other issues expressed in the past. At least the new Metro Redux test profiles are in good shape now so will be using those frequently for OGL4 drivers. Traces also aren't prone due to the headaches relating to Steam.

[baffledmollusc]

So with all of that said, do OpenGL game trace benchmarks interest you?

Very much so.

I mainly care about comparing the performance of various cards/drivers and tracking regressions/improvements, so this would be great!

[OneTimeShot]

I'd be interested in just a grid of various games running on various cards/drivers and a quick "working/not working" report. Performance testing would be a bonus, but I'd like to know if there are particular cards to avoid.

[chimpy]

Maybe I'm wrong, but did'nt The Talos Principle have linux benchmarking.

[Michael]

Maybe I'm wrong, but did'nt The Talos Principle have linux benchmarking.

I've worked with the developers on that.... Right now there's a bug on their end and was supposed to be fixed in the next update. That was a few weeks ago but haven't heard / checked to see if they pushed any new update recently.

[darkbasic]

I quote: "It's far from perfect, but better than what we have now."
Much more important, you will need to actually *USE* the new benchmarks: currently there are plenty of them (like CS:GO) which you used just once.

[Michael]

I quote: "It's far from perfect, but better than what we have now."
Much more important, you will need to actually *USE* the new benchmarks: currently there are plenty of them (like CS:GO) which you used just once.

There's reasons for that.... i.e. right now CS:GO with one of their updates they pushed down turned the "time demo quit" process into "time demo infinite loop".... So until they fix that (since I can't pull down older revisions with Steam), I can't use it since it's broken and just endlessly runs the demo over and over. With some of the other test profiles, they break too from time to time due to Valve either changing some library paths, etc. At least with traces don't have to worry about running benchmarks through this big mess caused by Steam.

[[Knuckles]]

It may be a bit of work, but for those games already supported by the pts, it would be interesting if you did an analysis of the behavior of the trace vs the real game.

Comparing the cpu and fps of the real demo vs the trace. This way at least we could get some insights on how similar (or dissimilar) it behaves. For instance, I would be expecting a bigger cpu usage on the original game vs the trance. But what about the gpu? Does it get similar fps? Does it also slow down in graphics-intensive parts? I'd expect a top-of-the-line cpu to present similar performance for traces and the real game, but a lower spec cpu to favor the traces.

[darkbasic]

There's reasons for that.... i.e. right now CS:GO with one of their updates they pushed down turned the "time demo quit" process into "time demo infinite loop".... So until they fix that (since I can't pull down older revisions with Steam), I can't use it since it's broken and just endlessly runs the demo over and over. With some of the other test profiles, they break too from time to time due to Valve either changing some library paths, etc. At least with traces don't have to worry about running benchmarks through this big mess caused by Steam.

Steam sucks so much :-(

[Kano]

I recorded myself apitrace from Killing Floor to compare gfx drivers, the size for just a few seconds is extreme. Do you really want to use traces?