AMD's Radeon Gallium3D Starts Posing A Threat To Catalyst

[bridgman]

IIRC everything after rv710/730 is supported but rv770 and earlier (including all the IGPs) are not.

Fixed typo -- missing a "7"...

And finally, you need to get 3.12-next to get the cpufreq ondemand governor fixed, or you need to manually find out and set its values.

Good point -- I forgot about the CPU governor.

[brosis]

Just noticed 6950 gets major slowdown everywhere. Can anyone explain this? Is it still because Michael uses Ubuntu (that has Intel pstate=off) with ondemand??? Even if 3.12 is patched, this can be the corner case - if such a high-end gpu is not feed with data adequately, it will definitely IOWait.

Michael, as nobody has a possibility to retest under exactly same condition, could you please rerun 6950 test, but with governor set to performance to clear things off?
Because, according to my tests, with manually tuned ondemand from 3.10 one reaches ~93% of glxgears fps compared to performance. (with stock ondemand -> 25%)

If you do this test, and it confirms, this would also mean that with cpu governor set to performance (or pstate for intel), there (might) would be about 7% performance boost, which should reach or even surpass catalyst..!

This is because opensource radeon driver is known to impact very light CPU overhead (compared to others), thus not triggering CPU out of sleep state.

[halfmanhalfamazing]

but rv770 and earlier (including all the IGPs) are not. The older UVD internals were sufficiently different in the areas which affect our ability to release open source support that we pretty much have to go through the whole review exercise a second time.

Bridgman,

Do you know if AMD is committed to doing that review exercise a second time for the earlier UVDs?

Thanks.

[GreatEmerald]

UVD is supported in the open drivers for more recent chips -- IIRC everything after rv710/730 is supported but rv770 and earlier (including all the IGPs) are not. The older UVD internals were sufficiently different in the areas which affect our ability to release open source support that we pretty much have to go through the whole review exercise a second time.

The ATI naming scheme sure is wacky. How come rv710 and rv730 come after rv770 and rv790...

[rohcQaH]

Just noticed 6950 gets major slowdown everywhere. Can anyone explain this?

According to Wikipedia, the 69xx are the first to use the VLIW4 ISA instead of the older VLIW5, so they'd use a different shader compiler. Is the 6950 still driven by r600g?

But then again, isn't Xonotic way more shader intensive than OpenArena? If it was an issue with the shader compiler, you'd see the difference with Xonotic, too.

So no, I can't explain it.

[markg85]

I think the last test (GpuTest v0.5.0) is the most interesting one from a technological point of view. It shows that the OSS driver still has WAYS to go to be on par with the catalyst driver on windows. And that is very promising because most game benchmarks are close to catalyst! In other terms, if the improvements keep flowing in i can see the OSS driver be quite a bit faster then the catalyst driver on windows or linux.

Interesting times ahead, that's for sure!

[Gps4l]

And now do the same with a HD7970 or a R9 290X using games like Serious Sam 3, DOTA 2 or Left 4 Dead 2.

I think everyone knows that the driver for the 4000 to 6000 series cards is wonderful and that there's no real reason to not use it over Catalyst. The situation for newer cards (and thus for gamers who bought their AMD card after 2011 and want to switch to Linux) is still dire if you don't use Catalyst (and Catalyst is dire itself).

I'm so happy that I was able to replace Catalyst with the open source driver on my laptop which runs the latest Mesa git snapshots and 3.12-rc7 right now, but it also has a Mobility Radeon HD5650 - I don't think I would've done that if I had an actual up-to-date GPU.

So yeah ... keep up the good work AMD, but please, start focusing on newer GPUs as well. I don't think recommending a 6870 for playing games on Linux sounds that nice, especially if everyone else already shouts "use NVIDIA on Linux only".

I definitely would like to see those games in the benchmarks.

Serious sam 3 on catalyst and opensource on Linux compared to catalyst on windows and openGl vs Direct X on windows.

The developers of Croteam stated that there is no reason in the game engine, why it should not run a tiny bit faster on linux, compared to direct X on windows.
So it might be a good game to see, how good the drivers are at the moment.
With this game, I experienced a big performance improvement, about half a year ago, with the catalsyt driver on Linux.


Left for dead 2 on Linux still isn't very smooth. (opnSUSE 12.3 HD5750 )
And I still remember the faster zombies ( with nvidia though)

http://blogs.valvesoftware.com/linux/faster-zombies/

[bridgman]

Bridgman,

Do you know if AMD is committed to doing that review exercise a second time for the earlier UVDs?

Thanks.

Yes, discussions have been underway for a while now.

Just like last time though -- no guarantees that we'll actually be able to release it no matter how much time we spend on it -- and the older UVDs have even more IP-related challenges than the more recent designs.

[bridgman]

The ATI naming scheme sure is wacky. How come rv710 and rv730 come after rv770 and rv790...

We used to develop the "big chip" first (big chips have a higher number in the second digit) then create smaller "derivative" chips once the big chip had finished most of the design verification. Sometimes the smaller chips would be able to pick up new versions of certain IP blocks.

Eventually the big chips got too big to fit on the megabuck HW emulators and so a "smaller chip first" model started to look more attractive

[Kostas]

Is there any webpage tracking Radeon's progress on the post-HD6000 GPUs?