OpenGL3.2 on Catalyst Wine-GL-Exstansion [Archive]

Qaridarium

08-03-2009, 11:50 PM

OpenGL3.2 on Catalyst Wine-GL-Exstansion

All abaut OpenGL3.2 in catalyst and the Wine Speedup.

Now OpenGL3.2 is Out and the wine-nvidia-only-openGL-extansion come to AMD carts!!!!!!!!

"The new release adds features to improve performance, quality, accelerated geometry processing and easier portability of applications originally made using Direct3D,"

Wine now can handle DirektX up to 10.1!!!!!!!

first time on amd as the same as on nvidia!

in my hope we will see Opengl3.2 in the 9-9 catalyst final!

Nille

08-04-2009, 12:27 AM

*plonk*

.....

doubledr

08-04-2009, 12:35 AM

in my hope we will see Opengl3.2 in the 9-9 catalyst final!

wake up, my friend...

Nille

08-04-2009, 12:48 AM

in my hope we will see Opengl3.2 in the 9-9 catalyst final!

since some month we wait for OGL 3.1 final and you think ATI/AMD can release in 2 month an ogl 3.2 driver ? rofl

BTW ATI/AMD has a 3 month delay with new features or functions.

I`m an AMD fan to but this is nothing else as dreaming.

Qaridarium

08-04-2009, 02:30 AM

since some month we wait for OGL 3.1 final and you think ATI/AMD can release in 2 month an ogl 3.2 driver ? rofl

BTW ATI/AMD has a 3 month delay with new features or functions.

I`m an AMD fan to but this is nothing else as dreaming.

there are OpenGL3.2 beta drivers in the cloused beta!

in near future there will be a big bang!

becourse some importand chances will come!

they drop the fireGL-only fokus! and make bist investment in overall lead!

RC driver openbeta will come to!

Be Happy Somthing very big is going on!

tuxdriver

08-15-2009, 04:17 AM

since some month we wait for OGL 3.1 final...

Full OpenGL 3.1 support is already available with the Catalyst 9.8 drivers.

Catalyst 9.8 also supports geometry shaders and 33% of the OpenGL 3.2 features.

Unigine Tropics gets a 10% boost with the new extensions.

Somthing very big is going on!

A fix for the broken / crappy OGL vertex shader performance :)

mirv

08-15-2009, 04:44 AM

I'm fairly certain amd/ati already provide support for all the extensions that were moved to core in opengl3.2 - and the same for nvidia. A simple glxinfo | grep [whatever] will show that. I haven't gone through every single new addition, but the extensions that I know of that will be of benefit to wine are definitely there.

I also haven't noticed any broken vertex shaders.

BlackStar

08-15-2009, 05:12 AM

there are OpenGL3.2 beta drivers in the cloused beta!

in near future there will be a big bang!

becourse some importand chances will come!

they drop the fireGL-only fokus! and make bist investment in overall lead!

RC driver openbeta will come to!

Be Happy Somthing very big is going on!

I am actually intrigued. Open beta? Not bad at all.

Looking forward to September release.

energyman

08-15-2009, 05:31 AM

I would like to know where he gets that kind of information.

Heiko

08-15-2009, 05:41 AM

I would like to know where he gets that kind of information.

My guess is that he is in the closed beta program. But in that case he must have signed an NDA an is violating it badly overhere ;).

energyman

08-15-2009, 05:47 AM

well, since he told me he is not in the closed beta we can rule that out ;)

Heiko

08-15-2009, 05:58 AM

well, since he told me he is not in the closed beta we can rule that out ;)

Hmm, in that case I'm curious as well ;).
But... how does he know that all those beta drivers are so good when he isn't in the closed beta? At least he must have access to the drivers from the closed beta.

Qaridarium

08-15-2009, 08:39 AM

Hmm, in that case I'm curious as well ;).
But... how does he know that all those beta drivers are so good when he isn't in the closed beta? At least he must have access to the drivers from the closed beta.

if you search the internal driver numbers on google you will find something!

exampel there is an result for an unleaked driver named 8.65.4 google:"fglrx 8.65.4"

last leaket driver was 8.65.3

you see? thats easy!

Nille

08-15-2009, 09:09 AM

if you search the internal driver numbers on google you will find something!

exampel there is an result for an unleaked driver named 8.65.4 google:"fglrx 8.65.4"

last leaket driver was 8.65.3

you see? thats easy!

You find 1 for 8.65.4 and 2 for 8.65.3 and all links point to phoronix.

nanonyme

08-15-2009, 10:24 AM

well, since he told me he is not in the closed beta we can rule that out ;)A shame, no one can send a patrol of lawyers after him to keep him quiet for a while. :(

Heiko

08-15-2009, 10:45 AM

Searching on "fglrx 8.65.4" is very interesting!

Graphics: EG Cypress (new DirectX 11 card of AMD is called Evergreen Cypress).

AdrenalineJunky

08-15-2009, 10:57 AM

Searching on "fglrx 8.65.4" is very interesting!

Graphics: EG Cypress (new DirectX 11 card of AMD is called Evergreen Cypress).

is it just me or do those results seem really random?

126fps in uniguine sanctuary, but only 76 in ET: quake wars? etc.

Heiko

08-15-2009, 11:22 AM

is it just me or do those results seem really random?

126fps in uniguine sanctuary, but only 76 in ET: quake wars? etc.

Well, the results might be off... Could be because these Linux drivers are not yet optimal for the new GPU. But it definitely looks like leaked AMD R8xx tests as performed by Phoronix (the url from google shows u=root, so I assume it is a test done by Michael).

Michael

08-15-2009, 12:00 PM

No Cypress testing was done by me or anyone affiliated with Phoronix.

Heiko

08-15-2009, 12:15 PM

No Cypress testing was done by me or anyone affiliated with Phoronix.

Hmm, I did some testing and I saw that if you run the test as root, the link to the results will include `u=root'. So it could indeed very well be that somebody else ran some benchmarks... perhaps use some kind of hack to change the name of the graphics card. Or perhaps not...

Michael

08-15-2009, 12:33 PM

Hmm, I did some testing and I saw that if you run the test as root, the link to the results will include `u=root'. So it could indeed very well be that somebody else ran some benchmarks... perhaps use some kind of hack to change the name of the graphics card. Or perhaps not...

The u=root portion just meant the user was running the tests as root on their own system.

Heiko

08-15-2009, 01:03 PM

The u=root portion just meant the user was running the tests as root on their own system.

Yes, I figured that (and changed my previous post accordingly). Now the question is: did somebody really have a Cypress card and tested it using the Phoronix Test Suite? (perhaps some of the AMD guys on this forum? Or perhaps it was Qaridarium :p, or maybe Phoronix is denying everything because there must be a very strict NDA on publishing about the new hardware).

Or did somebody mess with the submitted results and found a way to submit a dummy graphics card name?

AdrenalineJunky

08-15-2009, 04:40 PM

Qaridarium

08-15-2009, 05:37 PM

Qaridarium

08-15-2009, 05:38 PM

Yes, I figured that (and changed my previous post accordingly). Now the question is: did somebody really have a Cypress card and tested it using the Phoronix Test Suite? (perhaps some of the AMD guys on this forum? Or perhaps it was Qaridarium :p, or maybe Phoronix is denying everything because there must be a very strict NDA on publishing about the new hardware).

Or did somebody mess with the submitted results and found a way to submit a dummy graphics card name?

no thats not my work! i only become a information of 5870 hartware benchmark in the PTS database!

ObiWan

08-15-2009, 05:40 PM

Heiko

08-15-2009, 05:46 PM

my bet would be someone messed with submitted results...

it is concievable that it could have been one of the AMD guys, i kinda doubt it though. i don't see them putting benchmarks of a pre-production card on the internet like that.

Well, I tried running a benchmark myself. The default is to submit to the database after running the benchmark, so if you don't pay attention it might happen. I don't think there is a way of removing the results from the database after you submitted it...

nanonyme

08-15-2009, 05:50 PM

Well, I tried running a benchmark myself. The default is to submit to the database after running the benchmark, so if you don't pay attention it might happen. I don't think there is a way of removing the results from the database after you submitted it...I'd assume though that inhouse testing with the cards happens in an environment without an Internet connection. ;) That's what I'd do in an experimention environment anyway. ^^ And I don't see any reason why the pros would be any more stupid than me.

AdrenalineJunky

08-15-2009, 05:56 PM

there is no hack! thats the 5870! direktX11 VGA!

you have absolutely no way of knowing that.

Heiko

08-15-2009, 06:32 PM

I'd assume though that inhouse testing with the cards happens in an environment without an Internet connection. ;) That's what I'd do in an experimention environment anyway. ^^ And I don't see any reason why the pros would be any more stupid than me.

I know that some people, those who have the right connections, sometimes get a nice surprise from hardware manufacturers. Like a new piece of hardware that isn't yet available to the general public. It could be the same in this case... or perhaps AMD gave a couple of gpu's to some of their employees, just for home use.

You never know... but could still be that somebody found a way to change the name of the graphics card. But still... if somebody wants to start a rumour, he would make it more obvious probably. Would you call it `EG Cypress'? Nah... `HD5870' or `EverGreen Cypress' makes much more sense.

I think it might be real... however, who submitted this will probably remain a mystery.

energyman

08-15-2009, 07:03 PM

A shame, no one can send a patrol of lawyers after him to keep him quiet for a while. :(

well, I suppose since he did not sign anything he does not have a valid licence to use the drivers in the first place - but I am not a lawyer ;)

The only problem I see is: in the past he claimed this and that great improvement 'because it was in a beta' and then it wasn't there in the final. Stuff like that hurts amd.

nanonyme

08-15-2009, 08:43 PM

Yeah, you're right. And they should also be perfectly allowed to drop features from final versions even if they are in betas if they consider the features not to be mature enough yet. I'd be kinda interested to read the beta NDA some day mind. but I doubt they'd want to send me one unless I actually want to sign it. (which I don't, I've antipathies towards NDA's)

tuxdriver

08-15-2009, 10:05 PM

Catalyst 9.8 already has 33% of the OpenGL 3.2 features.

Corrected.

tuxdriver

08-16-2009, 12:06 AM

I also haven't noticed any broken vertex shaders.

The problem is really bad performance (especially at lower resolutions) with R600 cards, even under Windows XP [2.5x slower than Direct3D]

Crappy vertex shader performance with complex shaders and a weird issue with the memory management.

If you have an R600 card,try the Unigine Tropics benchmark under Windows XP with both renderers (Direct3D9 and OpenGL) at a lower resolution and compare the results. With OpenGL,aside from the ridiculously low framerate,you'll get occasional freezes/stutters as a bonus.
With D3D9 it runs really fast and smooth as silk.

Or try to play Serious Sam 2 in OpenGL mode.

cutterjohn

08-16-2009, 01:35 AM

Forget it, catalyst drivers suck hard, and I fail to see hoiw AMD can manage to stay in the GPU business with such crap drivers.

Anyways, windows 9.8 catalyst will be out Monday, and can be had already if you know where to look. Let's just hope that the apocalypse has come and the linux catalysts actually have some sort of improvements in store, but I don't really have any hope as I gave up a couple releases ago... and especially since I've notice no mindbending chaniges inthe space-time continuum to indicate a wrinkle in the space-time continuum indicating quality AMD driver release.. OTOH I've noticed heightened aimless marketing BS from marketroids... along with teh supa awesome phenom ii 965 "wish we could catch up" release (2006 called and said he wants his CPUs back AMD)

Anyone have a GT280M drop in for a GT725 notebook? I need a GPU that actually has working drivers... i.e. not an AMD part

A shame, no one can send a patrol of lawyers after him to keep him quiet for a while. A Shame that no one can send a patrol of lawyers after AMD for not providing useable support for their products...

Heiko

08-16-2009, 03:20 AM

Forget it, catalyst drivers suck hard, and I fail to see hoiw AMD can manage to stay in the GPU business with such crap drivers.

Anyways, windows 9.8 catalyst will be out Monday, and can be had already if you know where to look. Let's just hope that the apocalypse has come and the linux catalysts actually have some sort of improvements in store, but I don't really have any hope as I gave up a couple releases ago... and especially since I've notice no mindbending chaniges inthe space-time continuum to indicate a wrinkle in the space-time continuum indicating quality AMD driver release.. OTOH I've noticed heightened aimless marketing BS from marketroids... along with teh supa awesome phenom ii 965 "wish we could catch up" release (2006 called and said he wants his CPUs back AMD)

Anyone have a GT280M drop in for a GT725 notebook? I need a GPU that actually has working drivers... i.e. not an AMD part

A Shame that no one can send a patrol of lawyers after AMD for not providing useable support for their products...

Ah, the trolls are back. I was already missing them...

Heiko

08-16-2009, 03:35 AM

Hmm there are at least 2 tests from that guy

One with 8.65.3(ATI Open GL Version 2.1.9000)
http://global.phoronix-test-suite.com/?k=profile&u=root-6923-9841-16162

and one with 8.65.4(OGL 2.1.9000)
http://global.phoronix-test-suite.com/?k=profile&u=root-13430-28667-14083

The leaked one was 8.54.2(OGL 2.1.8975) not .3 or was there anonther leak i'm not aware of?

These are all the links I found:

fglrx 8.65.4, OpenGL: 2.1.9007, 2009-07-28 07:06, Graphics: EG CYPRESS 1024MB (http://global.phoronix-test-suite.com/?k=profile&u=root-13488-18327-24183) cypress_run
fglrx 8.65.4, OpenGL: 2.1.9008, 2009-07-29 09:10, Graphics: EG CYPRESS 1024MB (http://global.phoronix-test-suite.com/?k=profile&u=root-27628-3404-18707) startup
fglrx 8.65.4, OpenGL: 2.1.9003, 2009-07-30 07:27, Graphics: EG CYPRESS (http://global.phoronix-test-suite.com/?k=profile&u=root-13430-28667-14083) 8.65_full
fglrx 8.65.3, OpenGL: 2.1.9000, 2009-07-30 10:49, Graphics: EG CYPRESS 1024MB (http://global.phoronix-test-suite.com/?k=profile&u=root-6923-9841-16162) before_full_run
fglrx 8.65.3, OpenGL: 2.1.9000, 2009-07-30 14:20, Graphics: EG CYPRESS 1024MB (http://global.phoronix-test-suite.com/?k=profile&u=root-30516-15871-25631) try_again

mirv

08-16-2009, 03:51 AM

The problem is really bad performance (especially at lower resolutions) with R600 cards, even under Windows XP [2.5x slower than Direct3D]

Crappy vertex shader performance with complex shaders and a weird issue with the memory management.

If you have an R600 card,try the Unigine Tropics benchmark under Windows XP with both renderers (Direct3D9 and OpenGL) at a lower resolution and compare the results. With OpenGL,aside from the ridiculously low framerate,you'll get occasional freezes/stutters as a bonus.
With D3D9 it runs really fast and smooth as silk.

Or try to play Serious Sam 2 in OpenGL mode.

Not that sadly, I really don't use windows much anymore. No reason to boot into it basically, as I don't have that much time spare to play games these days (too busy programming and watching movies when not working).
Is there anything else you can suggest I try? I'm just interested in if it's really the drivers that are to blame, or game engines' handling of shaders.

AdrenalineJunky

08-16-2009, 06:02 AM

just noticed that the tests with the eg cypress were done with an intel processor... kinda seems unlikely it would be from AMD testing....

though it is still possible someone got thier hands on a pre-production sample.

nanonyme

08-16-2009, 07:09 AM

The problem is really bad performance (especially at lower resolutions) with R600 cards, even under Windows XP [2.5x slower than Direct3D]Hardly that surprising, vendors have to ship their own fast path implementations for OpenGL as far as I've heard. The good side is that if a vendor has a fast implementation on Wndows, they probably also have in Linux. (like nVidi) The bad side is that if you don't have a fast OpenGL implementation to begin with, you don't just magically get it on either platform. (like with ATi)

Qaridarium

08-16-2009, 09:22 AM

you have absolutely no way of knowing that.

i know it! there was also a PTS Profil DELL system witz 2x8core-intel and a 5870 and 64gb ram...

but michael delete this profil!

Qaridarium

08-16-2009, 09:24 AM

nanonyme

08-16-2009, 10:27 AM

XAA? You've got to be kidding! :D

ObiWan

08-16-2009, 12:32 PM

the leaket driver was name:"8.65.0"/8.65.3 <--

The leaked Ubuntu Driver was 8.65.2
(the one from the archive fglrx-8.650-ubuntu9.04-amd64only.7z)

at least thats what the driver calls itself on my system.

dmesg | grep fglrx
fglrx: module license 'Proprietary. (C) 2002 - ATI Technologies, Starnberg, GERMANY' taints kernel.
[fglrx] Maximum main memory to use for locked dma buffers: 7760 MBytes.
[fglrx] vendor: 1002 device: 9440 count: 1
[fglrx] ioport: bar 4, base 0xb000, size: 0x100
[fglrx] Kernel PAT support is enabled
[fglrx] module loaded - fglrx 8.65.2 [Jul 14 2009] with 1 minors
http://img80.imageshack.us/img80/5959/fglrxversion2.png

nanonyme

08-16-2009, 02:36 PM

Heh, seems the results Qaridarium linked to have been promptly removed. ^^

Heiko

08-16-2009, 04:14 PM

All the tests I have linked to are removed as well. Very strange.
However... google cache is your friend ;).

http://www.google.nl/#q=%22EG+CYPRESS%22+site:global.phoronix-test-suite.com&hl=nl&safe=off&filter=0&fp=ed4b95e7430cfd12

nanonyme

08-16-2009, 04:35 PM

Heiko

08-16-2009, 04:52 PM

Yeah, strange as in makes me be even more convinced in that they were fake. :)

Or not and it is all a big cover up (that is much more fun to believe ;)). How could Phoronix know whether the results were fake or not? Why weren't they removed yesterday when Michael responded in this topic (if they were fake he could have removed them back than).

I think somebody messed up (could be Phoronix, an AMD guy or some lucky guy who got his hands on a sample) and now they try to cover everything up.

nanonyme

08-16-2009, 05:05 PM

Or not and it is all a big cover up (that is much more fun to believe ;)). How could Phoronix know whether the results were fake or not? Why weren't they removed yesterday when Michael responded in this topic (if they were fake he could have removed them back than).

I think somebody messed up (could be Phoronix, an AMD guy or some lucky guy who got his hands on a sample) and now they try to cover everything up.On the other paw, why should Michael have to care if someone did that? I mean, it's not as if he's juridically responsible for the test results he hasn't himself sent...
It probably is semi-trivial to fake those results at worst by using a hex editor, at best an editor and a compiler, so I wouldn't really trust their validity much unless someone actually confirms them. (which doesn't seem likely)

Kano

08-16-2009, 05:08 PM

The results did not look like fakes, but the cpu was really slow. Of course faking is always possible, but i did not see any obivious fakes yet. But wrong results (due to errors which made em run much quicker than normal) i saw already several. Like when you bench against a profile and your display size is smaller then you get better results. Extreme results you see then for vdrift as not the original settings are used then but a fallback default. Other results are also faster but not so extremely false.

nanonyme

08-16-2009, 05:17 PM

I mostly meant that if I'd make the driver lie its version number, who would ever know the difference? It's impossible to say something is or is not a fake just by how it looks.

Kano

08-16-2009, 05:25 PM

The version number does not matter for me. But the results are not that phenomenal that it would require faking.

BlackStar

08-16-2009, 05:26 PM

I mostly meant that if I'd make the driver lie its version number, who would ever know the difference? It's impossible to say something is or is not a fake just by how it looks.

I wonder if it's possible for PTS to somehow verify the video card... maybe matching the video BIOS checksum against a database? Modified BIOSes will then show up as "unverified".

Kano

08-16-2009, 05:31 PM

Maybe think of some other normal solutions without a fake ;)

nanonyme

08-16-2009, 05:59 PM

Maybe think of some other normal solutions without a fake ;)Other normal solutions like a conspiracy theory instead of there just being some attention-seeking user? :p Yeah, that always makes more sense.

Heiko

08-16-2009, 06:08 PM

Other normal solutions like a conspiracy theory instead of there just being some attention-seeking user? :p Yeah, that always makes more sense.

If somebody faked it, it would make much more sense to fake the results as well so that it looked like the graphics card performed very fast. It would also make more sense to name the graphics card somewhat clearer (HD5870 would be clearer than EG Cypress).

Anyway... we'll probably never know for sure.

Qaridarium

08-16-2009, 07:09 PM

Yeah, strange as in makes me be even more convinced in that they were fake. :)

there is no fake!

amd' put this pts profiles online.

Qaridarium

08-16-2009, 07:10 PM

tuxdriver

08-16-2009, 07:35 PM

Its an amd guy! ...

...running tests on a Core2 Duo box? LOL :)

Qaridarium

08-16-2009, 09:35 PM

...running tests on a Core2 Duo box? LOL :)

ATI has many test systems with intel cpus!

and in the past ati only test on intel systems...

i know a bug catalyst 8.11 8.12 9.1 9.2 on 64bit systems with more than 2gb ram...

does not work becourse an bug only on amd cpu bases systems and in the internal testing they do not found the bug becourse of the intel testing maschines...

2 years ago wen AMD Buy ATI there was 90% of intel test-systems in ATI!

today amd systems come up in "ATI"

nanonyme

08-17-2009, 02:34 AM

If somebody faked it, it would make much more sense to fake the results as well so that it looked like the graphics card performed very fast. It would also make more sense to name the graphics card somewhat clearer (HD5870 would be clearer than EG Cypress).Actually no. If someone were to fake what the driver says the card is and the features but still run the test normally, results would be just as good as with a normal older card. Plus Qaridarium is raving about it not being a fake so it must be. QED. ;)

AdrenalineJunky

08-17-2009, 02:41 AM

Plus Qaridarium is raving about it not being a fake so it must be. QED. ;)

good point.

Heiko

08-17-2009, 03:30 AM

Actually no. If someone were to fake what the driver says the card is and the features but still run the test normally, results would be just as good as with a normal older card. Plus Qaridarium is raving about it not being a fake so it must be. QED. ;)

That is not what I meant: what I meant is that it would be much more interesting, if you want some attention, to fake the GPU name AND the results. Don't you think very high performance would draw a lot of attention? I believe somebody in this topic already pointed out that it is quite easy to influence the results of the test suite.

So my point is: if it was somebody who wanted attention, he might have gone for faking the results as well.