10.6.4 Really bad OpenGL performance with HD 5xx Solved

[Eisofen]

Moin,

 

here are my results, not THAT shabby.. Rig is an old GA-P35-DS3P, Intel C2D 6750, 4Gig, Graphics is a Powercolor 5770, driver is a patched netkas-driver (replaced Uakari with Vervet)

 

 

Ofen

[sch8mid]

so we see all the same behaviour

 

frames in the 200 range -and we should exspect > 1000

hope new frameworks and drivers will fix it

[dignouf]

so we see all the same behaviour

 

frames in the 200 range -and we should exspect > 1000

hope new frameworks and drivers will fix it

 

The 1000 fps picture uses an older version of opengl extension viewer. The benchmark must have been updated on the latest.

[mitch_de]

Yepp

same here

did a Timemachine backup

...installed opengl framework with Pacifist -as well as the new framebuffers and the quartz frameworks)

though do not know if all went well ..but had no KP

restarted without any problems

but

just a little lower as your score

 

under 200 fr.

 

my 3850 scored 1400 in Benchmarkmode 4 years ago

 

Thinks we have to wait for the champs Netkas and Morpheus

 

best

armin

 

Had you also installed the updated OpenCL Framework ?

Maybe help smalluxGPU get instances scene benchmark (in V171_V3) or instances interactive scene (V171_V2+) working.

[sch8mid]

Had you also installed the updated OpenCL Framework ?

Maybe help smalluxGPU get instances scene benchmark (in V171_V3) or instances interactive scene (V171_V2+) working.

 

hi Andreas

yes even the Quartz frameworks

nothing changed here

[mitch_de]

Thanks reporting this (slg instances scene OpenCL not working).

iMac 5770 get 43.8 sec on Luxball Ultra GPU and 46.0 sec on Sponza Ultra GPU. (barefeats soon)

[sch8mid]

Thanks reporting this (slg instances scene OpenCL not working).

iMac 5770 get 43.8 sec on Luxball Ultra GPU and 46.0 sec on Sponza Ultra GPU. (barefeats soon)

 

so my initial results with a fresh 10.6.4 install ,

Asere 1.9(vervet) in 64 bit and without ATI-Init,

and my Sappire HD 5770 seem rather o.k.

 

bench luxball midrange : 8,6 sec.

bench ultraGPU :34,9 sec

bench sponza :44,8 sec.

[mitch_de]

Thanks updating your results. Final version will have collected reference times (no speed changes) and very nice flat scene.

rob from barefeats also tested instances scene on new macs - no way with ATI. Driver updates in future will fix that - happend before also and fixed by Apple/ATI updates.

[sch8mid]

Just got the information:

 

 

"Attaching benchies, keep in mind that OpenGl View just got updated, sadly they didn't update the results page with a version number or anything, and since scores are now a fraction of what they were, we all need to be sure we are on same version."

 

Updated version is 3.31 .

 

So please be so kind and use all this new version so results are compatible.

[teeck2000]

Hey were you guys able to find a solution, sorry I read through the thread and couldn't find anything conclusive.

 

I don't have benchmark stills but I am getting really bad performance as well with the 5870. I was using a GTX285 in 10.6.4 and getting really bad performance so I thought I might try the 5870 since ATI has been better under 10.6.4, but I am getting about the same as I had with the GTX 285 in the 180-200 range.

 

I would really like to find a solution as everything else is running great. Let me know how I can help test.

 

I tried the new opengl frameworks with pacifist and it made my system unbootable, so I don't want to try that again, although I made a clone so I can get back running fast, if it does get bricked again.

[Dreamboy21]

my asus ati radeon 5770 cucore 1GB ddr5 performs great

 

 

my graphic card:

http://www.asus.com/product.aspx?P_ID=A3e6...&templete=2

[jmacie]

my asus ati radeon 5770 cucore 1GB ddr5 performs great

 

 

my graphic card:

http://www.asus.com/product.aspx?P_ID=A3e6...&templete=2

Which set of gl\cl frameworks are you using? This is the first report after the OP, to have higher numbers. If you used 10f2074, which parts did you use? Thx, johnm

[sch8mid]

"Attaching benchies, keep in mind that OpenGl View just got updated, sadly they didn't update the results page with a version number or anything, and since scores are now a fraction of what they were, we all need to be sure we are on same version."

 

Updated version is 3.31 .

 

So please be so kind and use all this new version so results are compatible.

 

 

 

 

sorry

 

old version of opengl extension bench I am pretty sure

 

and it is useless if he shows the 3.3.1 downloadpage

 

in the benchwindow only 2.1 mode is displayed , no 3.0. mode .....

with that version I got a score of 5260 with a WC HD 3870

and 1600fr. in benchmarkemode...3 years ago

 

 

5770 with old version (Benchmarkmode..Multisample 4- Anisotopry 16)

 

 

[teeck2000]

So what was happening was that starcraft II was throwing me off and so were the scores of open GL extension viewer version 3.31. I didn't know the benchmarks score system was changed. I ran the open GL extension viewer on my actual mac pro and the results were around what I got on this hack.

 

The problem was I kept running starcraft and getting like 15-20 FPS to test my performance, and it was making me think the open gl performance was awful, but I ran World of warcraft on ultra, 8x AA and the performance was good. Of course the performance could be better, since Win 7 runs it a lot better, but at least it's not as bad as I thought. So currently Starcraft II in OS X 10.6.4 has some huge issues and runs god awful on this hack. Although it was running ok on my actual Mac Pro 3,1 with GTX 285 in 10.6.4, so I a not sure what the issues are.

 

Here are my benchmarks for the my hack, 5870, 10.6.4.

 

[_MAC_BABY_]

MY HD5670 on 10.6.4

 

 

Why. FPS rate higher ?

[sch8mid]

now tick

Multisample 8

anisotropy 16

tick benchmark checkbox

 

and post results like advised here in this thread

 

if your unsure just look in teeck2000 post

there you see the correct config.

[daryoon]

Morpheus original screenshot had the following settings:

 

Display Mode: 1920x1200x32 bpp

Multisample: 4

Anisotropy: 15

Benchmark checkbox: ticked

 

I set the same settings and used OpenGL Extensions Viewer: glview 3.16. It's probably important to figure out which version of glview Morpheus used in his original test. As the benchmark result varied wildly.

 

Here's the results of my 5770 with glview 3.16:

 

This is the result with glview 3.31:

 

I wasn't able to locate a copy of glview 3.15. I believe that is the version that introduced "test 3.0."

 

Anyways, my point being that the glview version used can be the reason for Morpheus high benchmark results.

811079_glview316.dmg