Jump to content
25 posts in this topic

Recommended Posts

Hi friends..

 

Somebody fina a solution of chipset NForce3 and ATI 9600 Cards?? I tried any forms of configuration but i don't a sucessefull.

My machine for windows works fine, for games is excelent. I don't want swap my machine only this problem.

 

Somebody any sucess in this problem.

 

My configuration is:

A64 3.2 (no sse3)

MSI Neo Platinum (NForce3 250) (usb, sound, fireware, and sata working)

VGA Card ATI 9600XT 256 AGP

Monitor Acer 1916w (only 1024x768) native resolution is 1440x900

2 GB DDR

 

I think this machine is good. I have only problem with QE and CI on OSX. :thumbsdown_anim:

 

T+_+ Kill ABC

Link to comment
https://www.insanelymac.com/forum/topic/15250-nforce3ati-not-solution/
Share on other sites

Good topic!

 

I also have an nForce3 board with an ATI card, cannot work under OSX when QE3D is enabled. Everything is slow as hell.

 

I know it's not related to AMD since I had a Asrock 939dual before this one and QE3D was working fine.

 

It would be nice to collect ideas as where to look for to correct this problem. Is this because a kext somewhere has nforce ID's hard coded in it that prevents QE to work correctly. I tried many things already but without success.

 

To nForce users(or else), please post your experience! :blink:

Well I think now that the guilty is ATIRadeon9700GL.bundle, and it makes sense as it seems to be the one responsible for acceleration.

 

When I remove it everything works fine but no option to set QE3D. At least its responsible to load the OpenGL and OpenAl frameworks, so that makes me think the problem could also be from those frameworks.

 

I manage to get the ATIRadeonX1000GL and the ATIRadeon8500GL bundles to load by renaming the executable and replacing the Info.plist but it doesn't help either. It makes the screen garbled as soon as QE3D is activated.

  • 1 month later...
  • 2 weeks later...

Well I've been looking a solution to this problem for a long time now. The Boris method didn't work for me, I tried every possible combination of tweaks I can think of.

 

I believe that QE-3D is very slow because we run in our card in PCI mode...we miss the agp rate (4-8X) and the GART features. Our card runs like if we have a old radeon mach64 pci :(

 

Now I looked at the AppleI386 source and realized that the code was for old P2-P3 AGP2.0 bridge and vid cards. You can verify that by looking at the registers and compare them with Intel datasheet's:

 

http://www.intel.com/design/chipsets/datashts/29069302.pdf

 

With AMD64, many GART registers are set on the CPU so we need a major revamp of this driver. The reason that the AppleI386Driver ran without the ATI driver, is that it doesn't initialize anything, it just create a nub and exits. The ATI driver is responsible for loading the createagpspace etc.. functions so the AppleI386 is useless, it only make OSX see your card as an IOAGPDevice.

 

Cheers! ;)

I have the same problem and I have studied the cause too.

The MTRR isn't calculated correctly (at all) in the AppleMTRRSupport.kext

All OSes has the same problem, but at least Windows, Linux and FreeBSD have work-arounds.

 

I think we should fix the agp gart problem for other chipsets too. Buggy chipsets are probably the cause for tearing and other problems people are experiencing. The difference in our case is that we don't get a AGP port at all. These could and should be worked around relatively easily. It seems that just about any chipset has some problems.

 

Here's the FreeBSD source that fixes our chipset:

 

http://www.freebsd.org/cgi/cvsweb.cgi/src/...pci/agp_amd64.c

http://www.freebsd.org/cgi/cvsweb.cgi/~che...type=text/plain

 

 

Does anyone have time to port it as a patch to AppleMTRRSupport? (I'm too busy currently)

 

 

PS: I traced the problem all night, tracking the same problem on other os's.

I have the same problem and I have studied the cause too.

The MTRR isn't calculated correctly (at all) in the AppleMTRRSupport.kext

All OSes has the same problem, but at least Windows, Linux and FreeBSD have work-arounds.

 

I think we should fix the agp gart problem for other chipsets too. Buggy chipsets are probably the cause for tearing and other problems people are experiencing. The difference in our case is that we don't get a AGP port at all. These could and should be worked around relatively easily. It seems that just about any chipset has some problems.

 

Here's the FreeBSD source that fixes our chipset:

 

http://www.freebsd.org/cgi/cvsweb.cgi/src/...pci/agp_amd64.c

http://www.freebsd.org/cgi/cvsweb.cgi/~che...type=text/plain

Does anyone have time to port it as a patch to AppleMTRRSupport? (I'm too busy currently)

PS: I traced the problem all night, tracking the same problem on other os's.

 

 

Nice to see someone else working on this too :blink:

 

I found it funny to see your links to freebsd sources as I began last night to port these. lol

 

My plan was to make a replacement for AppleI386PCI, ie implementing IOPCI2PCIBridge methods... We know which methods to implement by having AppleI386 (and IOAGPDevice.h), so it should'nt be so hard. In fact I've port half of the linux agpgart project, but got stuck at implementing the gart memory, Linux becomes very different at this point... Freebsd sources seems better, but I only discovered these last night.

 

I agree we should port for other chipsets too, we have all the source we need...

I'll keep you updated

I dont know if this is relevant but i do remember on a system i used to have with a nforce3 and geforce6 that there was issues with stuttering? the fix for this was disableing fastwrites and wven when fastwrites was disabled in the bios it could be reenabled by the driver. is there a way to tell if fast writes are on in osx?

Fastwrites is an AGP feature , as described above our nforce3 boards are stuck in pci mode so I doubt it would matter. Nvidia cards aren't having this problem because they don't support acceleration yet meaning no QE+CI. Apple by default only allows QE on AGP cards because pci has insufficient bandwidth , our cards must be detected as AGP even though they're only operating at pci speeds or something like that.

 

 

This thread should be stickied , it's the only one that has come anywhere near finding a solution and it would help consolodate future questions and any progress. We just need some people to port the FreeBSD code that was discussed earlier. Really hope we can make something happen cause I've been considering building a new system in a few months because of this problem.

I dont know if this is relevant but i do remember on a system i used to have with a nforce3 and geforce6 that there was issues with stuttering? the fix for this was disableing fastwrites and wven when fastwrites was disabled in the bios it could be reenabled by the driver. is there a way to tell if fast writes are on in osx?

 

When writing the driver we can disable it, a register is responsible for it ...

 

This thread should be stickied , it's the only one that has come anywhere near finding a solution and it would help consolodate future questions and any progress. We just need some people to port the FreeBSD code that was discussed earlier.

 

We are already two working on this, we are to debug memory allocations routines. Every registers has been implemented already. Linux and freebsd are using structs and lists for their memory routines while apple has it's own implementation which is really not the same. It could work in no time once those routines are functionnal. If any dev see this post and would like to help, pm me. :)

 

Really hope we can make something happen cause I've been considering building a new system in a few months because of this problem.

 

I agree, I won't give away my X850 that easily neither.. :)

:construction: i have an ati 9600pro 128 agp and i cant get qe and ci but i already

make it to bee seen as agp in about this mac.

some one give me hope for this graphic card.

:blink: saludos, sorry about my english

about nforce 3 i already enable it !, but i got no sata recognition

×
×
  • Create New...