Jump to content

Best method for graphics performance?


18 posts in this topic

Recommended Posts

Performance has nothing to do with injection, unless you missed something (like for example adding your device ID to NVDAResman.kext).

 

It's the drivers that drive the video card, not the injection method, once OS X is running whatever injection method you're using has done its job.

 

There are kernel extensions that control GPU throttling, If they are loaded at all on your system you may have to modify them to get optimal performance. But again the injection method does not interfere with this.

 

Of course the situation could be totally different on ATI cards, I don't know anything about them.

Link to comment
Share on other sites

It's the drivers that drive the video card, not the injection method, once OS X is running whatever injection method you're using has done its job.

What drivers are we talking about here? (nVidia 285)

What happens with the official nVidia 285 Mac OS X drivers if they are installed? Can they be installed?

Link to comment
Share on other sites

What drivers are we talking about here? (nVidia 285)

We're in the Snow Leopard forum so I thought we were talking about the nvidia drivers that ship with Snow Leopard.

What happens with the official nVidia 285 Mac OS X drivers if they are installed? Can they be installed?

Yes they can, but I would only install them on 10.5.x (10.5.6 and up if I recall correctly?). I saw a good performance boost at the time. The drivers support many more cards than just the 285 GTX.

 

The Snow Leopard drivers are newer and have full 64-bit support. I'm sure Snow Leopard would break in terrible ways if you installed those old nvidia drivers - they're from July 2009!

 

FWIW I get the same level of performance with the drivers in Snow Leopard 10.6.3 - around 4000 points in OpenGL Extensions Viewer for everything except OpenGL 2.1 and 3.0, which appears to be normal, nobody gets good scores on those yet.

post-188427-1273011253_thumb.png

If you want to play with the 2009 Nvidia drivers make a test install of 10.5.x somewhere. Back when the drivers came out I posted a small install guide here:

http://www.projectosx.com/forum/index.php?...post&p=1677

Link to comment
Share on other sites

I tried with and without nVidia drivers in 10.5.8. Made no difference at all as can be seen below.

The Snow results were about 25% higher. Interesting. Thanks.

Order is Leo, Leo with driver and Snow.

post-448042-1273043560_thumb.jpg

post-448042-1273043567_thumb.jpg

post-448042-1273043583_thumb.jpg

Link to comment
Share on other sites

I'd root for Snow Leo. Though, I've yet to post my scores in the Forum that has a special Geekbench thread.

 

Coz Snow Leopard has 64 bit capabilities, and the overall data throughput between the CPU, RAM and GFX RAM, is a tad faster. I've seen that.

Specially, the OpenCL library. That too is improvised. Try running the Galaxies and qJulia tests. etc.

You'll get a better realistic performance score.

 

Regards,

Freaky Chokra :robot:

Link to comment
Share on other sites

You could test if chokra's theory holds water - run OGL Extensions Viewer on Snow Leopard in 32-bit mode.

 

If he's right, scores should match with what you got in 10.5.8. I'm curious to see the results myself.

Link to comment
Share on other sites

I'd root for Snow Leo. Though, I've to post my scores in the Forum that has a special Geekbench thread.

 

Coz Snow Leopard has 64 bit capabilities, and the overall data throughput between the CPU, RAM and GFX RAM, is a tad faster. I've seen that.

Specially, the OpenCL library. That too is improvised. Try running the Galaxies and qJulia tests. etc.

You'll get a better realistic performance score.

 

Regards,

Freaky Chokra :robot:

OOOOOPPPSSSSSS!

Thanks but I think you've missed the point of the thread! We're not discussing OSs.

Ya you are right! I was replying something and forgot to add the MAIN point!

HERE:

THe best method I think is DSDT! Not EFI. And DSDT editing + KEXT Patching.

That way, you can use full capabilities of the KEXTs, and Tap all the powerful abilities of your Display Card.

 

@Beerkex'd. Thanks for that! ;)

That way every Hackintosh can check for compatibilities and the potential of his card, if it is being used by OS X or not. ;)

 

Lastly, That OpenGL Ext. Viewer. Isn't that capable. Yes, it is much better for a diagnostic tool. Just like DxDiag that can be seen since Win98. ;)

But, both are just reporting tools. Though, O.E.V. is way better than DxDiag, itstill can't give you real world or realistic run time scores like Galaxies, qJulia, or even XBench or GeekBench. I'd say XBench is a notch higher. You should not miss the way it processes GFX output. It has all -> 2D images, 3D images rendering as well as text output performance scripts too!

 

GTG! Keep me interested! ;) :pirate2:

Regards,

Freaky Chokra :ninja:

Link to comment
Share on other sites

OOOOOPPPSSSSSS!

 

Ya you are right! I was replying something and forgot to add the MAIN point!

HERE:

THe best method I think is DSDT! Not EFI. And DSDT editing + KEXT Patching.

That way, you can use full capabilities of the KEXTs, and Tap all the powerful abilities of your Display Card.

 

 

GTG! Keep me interested! ;):robot:

Regards,

Freaky Chokra :ninja:

 

I wanted to fix my resolution when I got my iAtkos running a week ago, and I found a topic for my video card with kexts and dsdt.

I installed the kexts, but not the dsdt cause I couldn't find out how to get to Root, and my resolution went to normal.

What does DSDT do? I might want to give this another try

Link to comment
Share on other sites

You could test if chokra's theory holds water - run OGL Extensions Viewer on Snow Leopard in 32-bit mode.

 

If he's right, scores should match with what you got in 10.5.8. I'm curious to see the results myself.

OK, here we are.

Scores on Snow 32 bit are no different to 64 bit. So, that's that theory blown away!

 

 

 

Just to add that X-Bench Leo scores were actually higher than 64-bit Snow!

post-448042-1273229908_thumb.jpg

post-448042-1273230960_thumb.jpg

post-448042-1273230972_thumb.jpg

Link to comment
Share on other sites

OK, here we are.

Scores on Snow 32 bit are no different to 64 bit. So, that's that theory blown away!

Just to add that X-Bench Leo scores were actually higher than 64-bit Snow!

Lucky You!

 

1st)Well, that's something new to me. I've never booted t oSnow Leopard 32-bit mode ever. Except when I'm trying something new and have to rebuild the mkext files.

 

2nd) It's amazing to see Leopard figures to be higher than Snow Leopard.

Though, if you notice, KBeziers score is higher in Snow Leopard than in Leopard.

 

I'll try to post my results asap I can.

 

Regards,

Freaky Chokra :D

Link to comment
Share on other sites

That was unexpected. I suppose it's possible that OpenGL Extensions Viewer doesn't do anything that would take advantage of 64-bitness..? Or maybe that has nothing to do with it.. if drivers are 64-bit you should get 64-bit performance no matter what OpenGL Extensions Viewer is doing. Dunno.

 

Anyway if that really is the case I guess the only way to find out would be to find a benchmarking utility that has separate 32- and 64 bit modes. I know Geekbench has but it's not free.

Link to comment
Share on other sites

bigfish: that is what I use, but have you anything to back your statement up?

 

Beerkex'd: I didn't think that Geekbench measured graphic performance at all. FWIW, on 64bit SL, I get around 14000 using 64bit GB and 12000 using 32bit GB.

Link to comment
Share on other sites

Yeah you're right, it doesn't.

 

bigfish: We're discussing performance, not ease of use.

 

If you see better performance with GraphicsEnabler=y than with whatever you used before, I would like to see something to back that up as well.

Link to comment
Share on other sites

Well, this is becoming more interesting.

I just tried the OEV benchmarks with an EFI string from OS X Tools, and my results went from the 5000s (with graphics enabler) to the 20s. Yes twenties!

Link to comment
Share on other sites

Holy {censored}. Are you sure QE/CI is still enabled? And that you didn't switch to the Software Renderer in OGLEV by mistake..?

 

If you want to compare what's being done from one injection method to the other, you can make ioreg dumps and convert them from hex to readable plists with gfxutil.

ioreg -lw0 -p IODeviceTree | grep device-properties (not totally sure about this one, google "nvidia ioreg" or something and see what turns up)

gfxutil:

http://forum.netkas.org/index.php?topic=64.0

 

More reading material/clues on how to get at the info:

http://www.projectosx.com/forum/index.php?showtopic=370

http://forum.voodooprojects.org/index.php/topic,725

Link to comment
Share on other sites

 Share

×
×
  • Create New...