Radeon DRM Linux 4.4 + Mesa 11.1 + DRI3 vs. AMD's Proprietary Driver

[dungeon]

Nice to see those comparions but for me is a game fast enough if it never drops below 60 Hz to enable vsync. Everything else you can benchmark but with no real world use.

In real gaming world these days everybody talks about FreeSync monitors... so higher and lower then default sync values does matter. Also PC is not XBONE where only 30 fps or 60 fps target matter

I liked other Steam games to compare much more, it is unlikely that so many play the tested games. OpenArena is a remake of Quake 3 from 1999, which was addicting at that time and a bit later when QuakeLive was introduced, but basically a Pi is enough to play it.

You can't play on PI version of openarena 0.8.8 which is benchmarked here and which has bloom reflection enabled... PI will probably be below 10 fps in that case.

Can you use CS Go, Metro Redux, Bioshock 3 and similar games next time?

Dunno if Metro and Bioshock work on r600 mesa drivers (of course by default)

Unigine looks nice but the only game I own with that engine is Oilrush and there you hardly can see a difference if tesselation is enabled or not.

Unigine benchmarks are GPU benchmarks (those should heat GPU at 100% all the time) so they are valuable to test how one GPU scale up to another, even if there is no apps which use it.

[duby229]

I can't help but disagree. The only kind of performance that matters is actual performance.

[dungeon]

I can't help but disagree. The only kind of performance that matters is actual performance.

Even if it render differently? Michael said for Valley radeonsi render differently then fglrx... if it is really obvious that it render differently, then benchmarking numbers should be taken as inproper too .

[duby229]

Even if it render differently? Michael said for Valley radeonsi render differently then fglrx... if it is really obvious that it render differently, then benchmarking numbers should be taken as inproper too .

It doesn't matter either way. What game uses any Unigine engine? If there are any why not just benchmark the games instead?

[eydee]

Dunno if Metro and Bioshock work on r600 mesa drivers (of course by default)

By default not, Steam itself does an opengl version check and displays an error message, and denies launching the games altogether. It is probably easy to bypass, but that's out of "default" territory.

[duby229]

By default not, Steam itself does an opengl version check and displays an error message, and denies launching the games altogether. It is probably easy to bypass, but that's out of "default" territory.

Yeah, about that.... I have an ironlake based laptop that uses the i915 driver on mesa. In wine it plays the windows version of Civ5 just fine, but I can't launch the linux version because ironlake is limited to GL2.1 for now. Any suggestions?

[dungeon]

It doesn't matter either way. What game uses any Unigine engine? If there are any why not just benchmark the games instead?

Because Unigine benchmarks are done in a way to be GPU benchmarks, games are not that. Let say in Xonotic, you benchmark your CPU also because it is capped to single thred, beside that you benchmark your sound card too, your networking too... your or anyone else network and sound setups so different sound & network drivers too, even your network ping or speed matters there...

Lowest common denominator for benchmarking and compare performance is to things render the same on both before you looking at the numbers... at least on eye, average untrained eye

[duby229]

So? What's the point in that? Why does it matter if Unigine thrashes your GPU? Do you benchmark Unigine for fun? Do you gather your friends around and compete on Unigine benchmarks?

[dungeon]

So? What's the point in that? Why does it matter if Unigine thrashes your GPU?

It matter because you can see what GPU is actually faster (that is why those are GPU benchmarks) or if it does not heat up GPU you can conclude your driver is buggy, etc..

Do you benchmark Unigine for fun? Do you gather your friends around and compete on Unigine benchmarks?

Nope of course, with GPU benchmarks GPUs scale up fine... because you don't benchmark mostly anything else but GPUs and GPU drivers.

[duby229]

It matter because you can see what GPU is actually faster (that is why those are GPU benchmarks) or if it does not heat up GPU you can conclude your driver is buggy, etc..

That's not true. Unigine was developed heavily on nVidia hardware and it took a long time for mesa to hack around enough to make it work. It's a hack job that no game uses for a reason.