I've got the following system:
Asus P5K-E
Intel Core 2 Duo E6400 (OC to 2.94 GHz)
5 GB OCZ Gold DDR2 800 MHz
8800 GT 512 MB (flashed with mac 8800 GT Rom)
I've got Snow Leopard up and running with the standard DSDT generated by the patcher and the following kexts:
ADI1988b (enables sound)
fakesmc (newest version)
JMicronATA
NullCPUPowerManagement
OpenHaltRestart
PlatformUUID
SleepEnabler
VoodooPower
Everything is running smoothly in either 64 or 32 bit (I'm in 32 bit presently because there is no Voodoo Power kext for 64 bit as of yet.) The problem is that my open GL performance is less than half of what I see in Leopard (According to XBench, anyhow. OpenMark shows a 15 percent drop.) Anyone else seeing poor nVidia performance in Snow Leo?
62 replies to this topic
#1
Posted 24 August 2009 - 02:01 PM
#2
Posted 24 August 2009 - 02:14 PM
SA22C, on Aug 24 2009, 03:01 PM, said:
I've got the following system:
Asus P5K-E
Intel Core 2 Duo E6400 (OC to 2.94 GHz)
5 GB OCZ Gold DDR2 800 MHz
8800 GT 512 MB (flashed with mac 8800 GT Rom)
I've got Snow Leopard up and running with the standard DSDT generated by the patcher and the following kexts:
ADI1988b (enables sound)
fakesmc (newest version)
JMicronATA
NullCPUPowerManagement
OpenHaltRestart
PlatformUUID
SleepEnabler
VoodooPower
Everything is running smoothly in either 64 or 32 bit (I'm in 32 bit presently because there is no Voodoo Power kext for 64 bit as of yet.) The problem is that my open GL performance is less than half of what I see in Leopard (According to XBench, anyhow. OpenMark shows a 15 percent drop.) Anyone else seeing poor nVidia performance in Snow Leo?
Asus P5K-E
Intel Core 2 Duo E6400 (OC to 2.94 GHz)
5 GB OCZ Gold DDR2 800 MHz
8800 GT 512 MB (flashed with mac 8800 GT Rom)
I've got Snow Leopard up and running with the standard DSDT generated by the patcher and the following kexts:
ADI1988b (enables sound)
fakesmc (newest version)
JMicronATA
NullCPUPowerManagement
OpenHaltRestart
PlatformUUID
SleepEnabler
VoodooPower
Everything is running smoothly in either 64 or 32 bit (I'm in 32 bit presently because there is no Voodoo Power kext for 64 bit as of yet.) The problem is that my open GL performance is less than half of what I see in Leopard (According to XBench, anyhow. OpenMark shows a 15 percent drop.) Anyone else seeing poor nVidia performance in Snow Leo?
Yes opengl spinnig squares score with xbench is kinda slow 88 i get and in leopard 180+ , but the system runs sweet
#3
Posted 24 August 2009 - 03:33 PM
Do you see any difference in fonts too?
#4
Posted 24 August 2009 - 06:35 PM
hchinasky, on Aug 24 2009, 09:33 AM, said:
Same here with 9300m gs. Font rendering is much better in leopard. I don´t see fonts clear under snow leopard.
Do you see any difference in fonts too?
Do you see any difference in fonts too?
No, rendering in Snow looks great for me. It's also worth noting that the system feels responsive and snappy, it's just not performing well in the benchmarks. Since I don't game with OS X, it really doesn't matter in the end.
I'd bet that MacPro owners are seeing the same problem with their 8800GT's.
#5
Posted 24 August 2009 - 08:22 PM
yes, the same issue.
GeForce 8500GT 512MB (Xbench OpenGL Leopard: 180 / Xbench OpenGL Snow Leopard: 80).
GeForce 8500GT 512MB (Xbench OpenGL Leopard: 180 / Xbench OpenGL Snow Leopard: 80).
#6
Posted 24 August 2009 - 08:51 PM
Rendering is just fine tho and the fonts looks ok , just the opengl test in xbench is 100 or so fps slower lol to test it out i tried some off the osx scene demos and all ran excellent so the slow opengl score doesnt seem to bad overall
#7
Posted 24 August 2009 - 09:58 PM
I think, xbench is too old. Try: http://www.realtech-vr.com/glview/
#8
Posted 26 August 2009 - 08:33 PM
Me too.
Test with GioFX OpenMark
Launch the benchmark from the OpenMark Menu
Waiting for the result after 9 FPS, open "OpenMark result.tga" file in the folder
10.5.8 ± 26000
10.6 ± 14000 ?
Test with GioFX OpenMark
Launch the benchmark from the OpenMark Menu
Waiting for the result after 9 FPS, open "OpenMark result.tga" file in the folder
10.5.8 ± 26000
10.6 ± 14000 ?
Attached Files
#9
Posted 27 August 2009 - 02:55 PM
It's a problem with Snow Leopard, not your hacks. It seems as if OGL performance is mucked up, and CL is apparently spotty for ATI cards.
#10
Posted 29 August 2009 - 01:36 PM
Same issue here (9600GT) and spaces is absolutely not smooth in SL. Hope Apple quickly releases updates...
#11
Posted 29 August 2009 - 05:31 PM
I've noticed this as well (9800 GT). World of Warcraft performance is much worse than Leopard, and lines look much more jagged. No matter what I set the anti-aliasing on, it doesn't look like it's working.
There is definitely something wrong with OpenGL.
There is definitely something wrong with OpenGL.
#12
Posted 29 August 2009 - 09:58 PM
I have just tried the OpenGL Extensions Viewer link that zoliky posted.
I loaded the app and used defaults settings and my SnowLeopard system posted higher scores than 10.5.8, apart from test 2.1. I am using an MSI 8800GT 512MB OC.
EDIT: I have just run the same test again but with time with 10.6.1. There is a slight improvement in some tests and a slight loss in others, so an average no real change.
I loaded the app and used defaults settings and my SnowLeopard system posted higher scores than 10.5.8, apart from test 2.1. I am using an MSI 8800GT 512MB OC.
EDIT: I have just run the same test again but with time with 10.6.1. There is a slight improvement in some tests and a slight loss in others, so an average no real change.
Attached Files
#13
Posted 30 August 2009 - 06:15 AM
A second test please ?
Test with GioFX OpenMark
Test with GioFX OpenMark
#15
Posted 30 August 2009 - 09:26 AM
#16
Posted 30 August 2009 - 09:56 AM
#17
Posted 30 August 2009 - 12:57 PM
FORGET XBench & OpenMark benches for 10.6 + 10.5 on newer Intel "Macs" !!
Both are much to old and need to be updated. They are both very old compiled with very old OS X frameworks / very old GCC 3.x (UB bot not good optimized Intel code). Now we have GCC 4.x !
Only use the new + very valide OpenGLExtensionsviewer for Tests .
To compare GPU speed differences between GPUs:
OPTIONS: full screen, benchmark, higher resulution (at least 1280x800, the more the better), 32 Bit and 4* Multisampling (FSAA)
gives: high gpu load = gpu limits bench results = less cpu+driver dependend bench results
To compare driver speed differences on same GPU:
OPTIONS: full screen, benchmark, lower resulution (less than 800x600), 32 Bit no Multisamling
gives: less/moderate gpu load = CPU(drivers) + gpu limits bench results = much cpu+driver dependend benhc results
Both are much to old and need to be updated. They are both very old compiled with very old OS X frameworks / very old GCC 3.x (UB bot not good optimized Intel code). Now we have GCC 4.x !
Only use the new + very valide OpenGLExtensionsviewer for Tests .
To compare GPU speed differences between GPUs:
OPTIONS: full screen, benchmark, higher resulution (at least 1280x800, the more the better), 32 Bit and 4* Multisampling (FSAA)
gives: high gpu load = gpu limits bench results = less cpu+driver dependend bench results
To compare driver speed differences on same GPU:
OPTIONS: full screen, benchmark, lower resulution (less than 800x600), 32 Bit no Multisamling
gives: less/moderate gpu load = CPU(drivers) + gpu limits bench results = much cpu+driver dependend benhc results
#18
Posted 30 August 2009 - 02:42 PM
yes thats right Mitch_de, I did a test with OpenglExtensionsviewer...
and the results are not loss any performance in SL
and the results are not loss any performance in SL
#19
Posted 30 August 2009 - 02:49 PM
mitch_de, on Aug 30 2009, 02:57 PM, said:
FORGET XBench & OpenMark benches for 10.6 + 10.5 on newer Intel "Macs" !!
Both are much to old and need to be updated. They are both very old compiled with very old OS X frameworks / very old GCC 3.x (UB bot not good optimized Intel code). Now we have GCC 4.x !
Only use the new + very valide OpenGLExtensionsviewer for Tests .
To compare GPU speed differences between GPUs:
OPTIONS: full screen, benchmark, higher resulution (at least 1280x800, the more the better), 32 Bit and 4* Multisampling (FSAA)
gives: high gpu load = gpu limits bench results = less cpu+driver dependend bench results
To compare driver speed differences on same GPU:
OPTIONS: full screen, benchmark, lower resulution (less than 800x600), 32 Bit no Multisamling
gives: less/moderate gpu load = CPU(drivers) + gpu limits bench results = much cpu+driver dependend benhc results
Both are much to old and need to be updated. They are both very old compiled with very old OS X frameworks / very old GCC 3.x (UB bot not good optimized Intel code). Now we have GCC 4.x !
Only use the new + very valide OpenGLExtensionsviewer for Tests .
To compare GPU speed differences between GPUs:
OPTIONS: full screen, benchmark, higher resulution (at least 1280x800, the more the better), 32 Bit and 4* Multisampling (FSAA)
gives: high gpu load = gpu limits bench results = less cpu+driver dependend bench results
To compare driver speed differences on same GPU:
OPTIONS: full screen, benchmark, lower resulution (less than 800x600), 32 Bit no Multisamling
gives: less/moderate gpu load = CPU(drivers) + gpu limits bench results = much cpu+driver dependend benhc results
Thank you !!!
#20
Posted 30 August 2009 - 03:08 PM
Did the OpenGLExtensionsviewer test with my 9600GT in 1920x1200 4xAA / 16xAF ... Scores are better in Leopard (GL 1.1 to 2.0 ~700fps in 10.5.8, ~600fps in SL, GL 2.1 is the same) but not much so the performance is ok. The slow spaces seems to be a problem with beamsync as I have only 10fps with beamsync, in leopard it's 50-60fps... without beamsync >100.
0 user(s) are reading this topic
0 members, 0 guests, 0 anonymous users



Sign In
Create Account










