Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
i used the leaked ones.
Having no Problems detected until now :)
I just setup my mac 4 weeks ago and have an up to date backup so i just thought "um... do it" ^^
here's my test with my 13" MacBook Unibody first generation with 9400m
almost no changes with the new drivers :(
But i'm VERY happy with the results in my iMac.
For example WoW run ok (everything max no FSAA), still MUCH better than with windows 7 and now i can play with 2 - 4x FSAA mostly smooth.
Can't wait to see the first final release of the coming updates. As far as i know it's not the only one we can expect in near future :)

Later this evening or tomorrow i will test Starcraft 2 and TF2.
But Starcraft has an "intern" bug with the Shaders on Mac OS. At least i read it. They are working on it.

Blizzard Blue Post:
right now, the Mac client is seeing some pretty major performance issues due to bugs with both the nVidia graphics drivers in 10.6.4. Snow Leopard, and a bug within Starcraft 2: Wings of Liberty.

The Shaders, Lighting, Shadow options eat up all of the available video memory, or VRAM, very quickly. This results in the computer paging out to the standard RAM and to the Hard Drive to store the shader details, which is much slower than VRAM. This makes the game slow to a crawl.

I personally recommend capping the frame rate of Starcraft 2 so your computer doesn't overwork itself, wasting resources on the shaders, leaking all of the VRAM and overheating itself in the process. See below in "Things to Know Before Changing Any Settings" for more information on how to do this

...

Limiting the in-game frame rate to 30 or 40fps will help keep your Mac cool and prevent overheating. The Macbook Pros especially tend to run a little bit hot being in that nice tiny form factor. To limit the frame rate, type in and save the following into the Variables text file:

frameratecap=30
frameratecapGlue=30

"30" being the value of whatever number you want to cap it at. You will find the Variables text file under:

“/Macintosh HD/Users/TheUserYouInstalledSC2Under/Documents/Blizzard/StarCraft II/Variables.txt"

To see the current frame rate while playing the game, hold down Control+Option, and tap the F key.
 

Attachments

  • pre.png
    pre.png
    98.4 KB · Views: 175
  • after.png
    after.png
    108.8 KB · Views: 144
i used the leaked ones.
Having no Problems detected until now :)

Did you use snow_leopard_graphics_update_10f2074.dmg, or did you just place the different .kext-files in their right place? Would be very kind of you to describe in detail how you did, since I wanna try it myself, but don't want a corrupt system again (as happened the first time I tried it).
 
I got the snow_leopard_graphics_update_10f2074.dmg and simply run the installer. That was all I did.
 
I got the snow_leopard_graphics_update_10f2074.dmg and simply run the installer. That was all I did.

Gonna try this myself now then. And hope for the best. :)

Think I'm gonna try some in-game FPS as well, so we can see how it compares in "real world" as well. Hope to see you guys soon. :)
 
Ok, here are my results. This is with an iMac 2,4 Ghz with ATI Radeon HD 2600 Pro. I installed using snow_leopard_graphics_update_10f2074.dmg. First, the OpenGL Extensions Viewer test (http://www.realtech-vr.com/glview/download.html), is really promising. Look at the pre and post-graphicsdriver update (attached). The "real world" test I did, using Team Fortress 2, HL2 Episode 1 and Episode 2 is not that impressive. However, I guess there's a huge chance for these results not being very scientific correct, since all I did was record a timedemo in each game (about 1 minute of gameplay), and then ran the timedemo before and after I installed the update. Now that I think of it, I had an external 1080p TV connected as well, don't know if that makes any difference. I don't know the exact settings I used (this is the settings I use when I played the games, so can't remember, there's no AA for sure).

Can someone with more knowledge to this explain why the OpenGL-test is so remarkable, and the gaming FPS doesn't change that much?

The results:

Team Fortress 2:
Pre: 26,54 FPS. 5,32 fps variability
Post: 33,21 FPS. 4,98 fps variability

HalfLife 2 Episode One:
Pre: 31,07 FPS. 7,54 fps variability
Post: 33,38 FPS. 7,91 fps variability

HalfLife 2 Episode Two:
Pre: 19,56 FPS. 5,37 fps variability
Post: 21,61 FPS. 6,36 fps variability
 

Attachments

  • pre.png
    pre.png
    108.2 KB · Views: 138
  • post.png
    post.png
    107.8 KB · Views: 69
The Source engine is not yet optimized for the new drivers, and does not fully leverage the optimizations. That's basically what RBarris (OS X dev at Valve) has said on the Steam forum.
 
The Source engine is not yet optimized for the new drivers, and does not fully leverage the optimizations. That's basically what RBarris (OS X dev at Valve) has said on the Steam forum.

Yeah, that sounds about right. Too bad I didn't try out any non-steam games, although one would think that the same reasoning applies to all games? Gonna try some others now and see if I can notice any difference..
 
Nearly no improvements in Starcraft 2 :(

I hope it will become better, like WoW where the Mac Version runs much smoother compared to Win7 :)
 
thanks **ChuckBlack**

//edit


Ms x4 / Ani x16
Zrzut_ekranu_2010-08-13_godz_17_18_44.png


Look above these are the correct settings :











Please everybody doing these tests:

use OEV : 3.3.1 (they updated the new version)

use benchmarkmode

use Multisample: full

use Anisotropy : full

we have no comparable tests without this.


HD 5770
:

First test : benchmarkmode : multisample /anisotropy : 1


http://att.macrumors.com/attachment.php?attachmentid=246500&stc=1&d=1281953053


second test : benchmarkmode : multisample/anisotropy :full = 8/16


http://att.macrumors.com/attachment.php?attachmentid=246501&stc=1&d=1281953072


And look at this ! :

very strange : no difference between the 2 modes :



same framerate about 300 fr. in non AA mode and 16 mode
 

Attachments

  • Bildschirmfoto 2010-08-16 um 11.47.31.png
    Bildschirmfoto 2010-08-16 um 11.47.31.png
    71.9 KB · Views: 168
  • Bildschirmfoto 2010-08-16 um 11.53.14.png
    Bildschirmfoto 2010-08-16 um 11.53.14.png
    72.8 KB · Views: 166
Ok, here are my results.

Can someone with more knowledge to this explain why the OpenGL-test is so remarkable, and the gaming FPS doesn't change that much?

The results:

Team Fortress 2:
Pre: 26,54 FPS. 5,32 fps variability
Post: 33,21 FPS. 4,98 fps variability

HalfLife 2 Episode One:
Pre: 31,07 FPS. 7,54 fps variability
Post: 33,38 FPS. 7,91 fps variability

HalfLife 2 Episode Two:
Pre: 19,56 FPS. 5,37 fps variability
Post: 21,61 FPS. 6,36 fps variability

Interesting :

But please use the settings mentioned above.
i have a second diplay attached as well
no change in the benchmarkmode...
 
The update doesn't fix a choppy scrolling issue on Firefox. I still should stick with 10.6.3 drivers.
 
Can someone with more knowledge to this explain why the OpenGL-test is so remarkable, and the gaming FPS doesn't change that much?

Well, the numbers you gave shows a 20-25% improvement, which is pretty huge imo. But I agree, on my iMac I got about a 50% improvement in the OpenGL tests.

My guess is that this is because games do so much more than just render graphics - it handles user input, plays music/video, calculates AI for enemies in the game, sending network traffic and so on. If the game spends 50% of the time to draw the game scene and the rest on above mentioned stuff, you'd see about a 25% improvement in overall game performance. I guess this is the case in real gaming tests vs OpenGL tests.
 
Still no difference

FYI

I tried the iMac update, the pre-release SLGU and the newly released SLGU 1.0 and still have no performance difference at all.

EDIT: I noticed one problem with the SLGU. Portal and Halflife 2 ep1 won't boot anymore...
EDIT EDIT: I rolled back the drivers to the pre-release of the Snow Leopard Graphics update and everything is booting again.

Screen%20shot%202010-08-18%20at%201.22.23%20AM.jpg
 
ITS WORKING!!!!!


SOLUTION


10.6.4 will severely decrease NVIDIA GPU performance by causing it's three clocks to always run at level 0 low power mode.

Edit the AppleGraphicsPowerManagement.kext/Contents/Info.plist's MacPro4,1 section and change Vendor10deDevice05e2 to Vendor10deDevice05e3.
The 05e3 part is for a GTX 285, see nvclock0.8b5_7-darwin/src/backend/info.c (link below) for others.
This is to allow the GPU to have power management; level 2 high power when stressed, and down clocking to level 0 when not stressed (~10C lower).
The GPU clock will be at level 0 after sleep and wake, but the shader and memory clocks will temporarily be stuck at level 2.
However, if load is applied and then reduced, shader and memory will return to level 0.

Delete AppleGraphicsPowerManagement.kext.
or
Use the AppleGraphicsPowerManagement.kext from 10.6.0/10.6.2/10.6.3.



Temperatures and clock speeds can be seen with nvclock, and for lower temperatures it's recommended to set fanspeed mode to auto with `nvclock -F auto -f', which needs to be set again after sleep and wake

http://www.projectosx.com/forum/index.php?showtopic=1246



FULL SPEED - SCII 1920x1200 Ultra non lag !!!!!!
 
OpenGL Extensions Viewer 3.32 NEW

Zrzut_ekranu_2010-09-08_godz_22_18_29.png
Zrzut_ekranu_2010-09-08_godz_22_20_25.png


Zrzut_ekranu_2010-09-08_godz_22_21_46.png
Zrzut_ekranu_2010-09-08_godz_22_23_08.png



Correct settings :

use OpenGL Extensions: Viewer 3.32
use Multisample: full (x4)
use Anisotropy : full (16x)


Zrzut_ekranu_2010-09-08_godz_22_36_37.png
Zrzut_ekranu_2010-09-08_godz_22_38_50.png

Zrzut_ekranu_2010-09-08_godz_22_40_08.png
Zrzut_ekranu_2010-09-08_godz_22_41_25.png


back to the game
 
I don't get it, are you saying just to delete the kext file and restart?

I'm on MacBook Pro with Geforce 9600M GT. How can i get the graphics performance up as in 10.6.3?
 
Top scores in OpenGL framerates

Built a new rig last week , some screens for you:

the new built is actually number 1 in Cinebench 11.5 unofficial database OpenGl Rendering
in Win 7 and OSX 10.6.4 (ATI Radeon HD 5770)

Some stock data:

Mac Pro 2010 (late August) 6 cores -12 threads
Framerate @ 3,33 GHz = 34,57 fr.

here:

Custom built Mac 4 cores -8 threads
Framerate @3,33 Ghz = 36, 82 fr.


Comparison : Apple MacPro 2010 vs. Custom Mac

Case : Lian Li ..(better cooling , better air flow , better fans ,dead silent)
CPU : i7 950 @> 4,0 (better cooling with 2 x 800 rpm silent fans in push pull)
GPU : ATI Radeon HD 5770 by Sapphire (advanced heatpipecooling -3rd generation)
RAM : 6 GB Mushkin Redlines 1600 6-6-8 (better RAMs -higher speed)
MB: Gigabyte ( USB 3 - Esata - 6 Gb Sata - 3 Hi-Speed PCIE-slots)

899 € with Win 7 on HD 1 ......MacOSX 10.6.4 on HD 2 - boot selection : F12
_______________________________________

Best results :

My Custom built Mac OC :

Cinebench : OSX 10.6.4 : 45,13 fr.

Cinebench : Win 7 : 88,16 fr. (32 bit !)

OpenGL CPU : 7,1 pts.

OpenCL : 6,8 sec.

Geekbench : 14664




I really was not willing at all to pay that ridiculous amount of money for a computer with :



a. outdated technology : a case designed in 2004 with bad ventilation , a graphiccard that uses a 2008 reference design @ double cost with
slow drivers , a motherboard without 3 Hi-speed PCI-E slots , without fast 6 GB Sata , without USB 3, without e- sata
without Blu-ray and with cheap and slow RAM .


So I am quite happy for now in my fast 900 $ non reality distorsion field , Mister Jobs and shareholders.
And I don´t feel like an idiot paying more than 2000 bucks for this kind of slow " Recycling Computer" ,
a nice but ugly try to squeeze the most money out of faithful customers, Mr.Jobs.
And I am on my way to leave the Apple Pro App boat as well.
One of the best NLE Editing systems (FinalCutPro) the world has seen , left in the dust for years now.

Regards

S ( first Mac Pro 1988 - last MacPro for now 2008) ..until Apple returns from Disneyland and the orientally flavoured marketing methods of its CEO.



First. screen: Apple Mac Pro (Pro ?...rofl) 2010 late August . other screens my custom built....
_______________________________________
 

Attachments

  • cinebenchmacpro.png
    cinebenchmacpro.png
    134.7 KB · Views: 156
  • cinebench database.jpg
    cinebench database.jpg
    96.3 KB · Views: 174
  • best mac.png
    best mac.png
    151.5 KB · Views: 144
  • 88.jpg
    88.jpg
    79.6 KB · Views: 161
  • geek.jpg
    geek.jpg
    31.8 KB · Views: 183
Simple solution

Simple solution: Just play a quicktime H264 video in the background.

It will throttle the dedicated GPU up, since it uses hardware acceleration. Keep it playing in a loop, because why not.

It will eat maybe a few % of the CPU but that will STILL be a way better solution than disabling the GPU power management kext, which will decrease battery life ca. 2-3x.

I'm not sure if this kicks the GPU all the way up, but playing any qt h264 video makes Expose/Desktop/Spaces etc. an awful lot smoother. Enough so to forget about the issue. (Which is there all the time, and I'm really pissed at Apple).
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.