Jump to content

New AGPGart

Slice

By Slice,
in AGPGart

 Share
941 posts in this topic

Recommended Posts

Dan Druff
Dan Druff
Posted
Posted
hello

 

Ok you say with agp bus you get less fps than in pci mode ;) .Weird.. :)

 

PCI and AGP bus are always the same speed or so (from my own experience) with AGPGart, the video card kext doesn't seem to be able to tap entirely the AGP bus to operate but downgrade to PCI performances. My OpenMark results delta between PCI/AGP tests are always very similar to those of Ktrain.

 

This said, it is reported (once) that AGPgart worked for VIA chipset under Tiger and the raise in performance was more than clear, see post 255.

 

[EDIT: help needed]

 

I'm looking to use EFI string but I can't get them working (for now).

 

So I'm looking for injector (natit, ATIinject or others) that could inject values and work with vanilla kext, if any of you is aware of a method to get ATI X1600 pro (Dev. ID: 71C2 primary monitor and 71E2 for secondary) running with two monitor, I would be very happy.

Link to comment
Share on other sites
Dan Druff
Dan Druff
Posted
Posted

Hi,

 

I've reduced my setup to a near vanilla hackintosh here are all my kext modifications:

 

Here are my kext info:

 

Video (mix of ideas from dm_webd and cyclonefr, look here for later method)

AGPGart -> 2.7.1 with AGP_HOST: 0xf0000000

Natit -> with dual x1600 info.plist, all from DM_webd method look here

ATIRadeonX1000 -> vanilla with 71C2 key added

IONDRVSupport -> Vanilla 1.4.3 (from 10.4.7 combo update kexts)

 

ATA support

IOATAFamily.kext -> No version number (For ATA Support)

 

MB support (audio/networking)

AppleAC97Audio.kext -> 1.0.0 (for AC'97 sound support)

AppleAzaliaAudio.kext -> 1.0.0 (for AC'97 sound support)

IONetworkingFamily.kext -> 1.6.0 (for networking)

 

EFI support

AppleSMBIOS.kext -> 1.0.14 (For compatibility with EFI boot)

dsmos.kext -> No version number (Page description EFI module)

 

Rest is 10.5.4 vanilla kext.

 

 

Also results from OpenMark are different under 10.5.2 (450) and 10.5.4(517) with near 70 pts delta in favor of 10.5.4. Boot is also faster, maybe an effect of the vanilla kernel, or the two are distinct, OpenGL framework improvement and kernel effect.

 

BTW, I got QE/CI, rotation and everything is hardware accelerated.

Link to comment
Share on other sites
ktrain
ktrain
Posted
Posted

Just a little update here, I've all but given up on resolution change for my setup, I have tried every combo I can think of, the best i have so far is stuck at 1024x768@60 Hz, with HW QE/CI.

 

Every injector/plist edit I've tried gives me a blue screen with cursor, even with the IONDRVSupport.kext Slice mentioned earlier, makes no difference.

 

In short, X1300 Pro AGP ID 7142 is not currently fully compatible with Leo, at least in my setup with no-EFI.

Link to comment
Share on other sites
Dan Druff
Dan Druff
Posted
Posted
Just a little update here, I've all but given up on resolution change for my setup, I have tried every combo I can think of, the best i have so far is stuck at 1024x768@60 Hz, with HW QE/CI.

 

Every injector/plist edit I've tried gives me a blue screen with cursor, even with the IONDRVSupport.kext Slice mentioned earlier, makes no difference.

 

In short, X1300 Pro AGP ID 7142 is not currently fully compatible with Leo, at least in my setup with no-EFI.

 

You should try my setup with Dev ID changed to 7142, it may work, but you must have EFI installed.

Link to comment
Share on other sites
Dan Druff
Dan Druff
Posted
Posted
2 Dan Druff

Could you test different flags, please?

AGPGart.info.plist

            <key>IOAGPFlags</key><BR>              <integer>5</integer>

Possible values 1,5,0x27, 0x10001, 0x10005, 0x10027

 

Here are my tests, please note that dmesg keeps loaded with junk (at the end), I don't know why.

 

Besides, looks like this have no effect at all.

 

 

Value

 

0x1

Openmark: 510

Cinebench: 3456

Other: Boots fine, no problem, idem to default setup on AGPGart 2.7.1

 

0x5

Openmark: 510

Cinebench: 3447

Other: Default setup, some image tearing under OpenGL rendering

 

0x27

Openmark: 510

Cinebench: 3461

Other: Boots fine, no problem, idem to default setup on AGPGart 2.7.1

 

0x10001

Openmark: 510

Cinebench: 3440

Other: Boots fine, no problem, idem to default setup on AGPGart 2.7.1

 

0x10005

Openmark: 510

Cinebench: 3450

Other: Boots fine, no problem, idem to default setup on AGPGart 2.7.1

 

0x10027

Openmark: 510

Cinebench: 3460

Other: Boots fine, no problem, idem to default setup on AGPGart 2.7.1

 

 

I hope this help.

results.zip

Link to comment
Share on other sites
Dan Druff
Dan Druff
Posted
Posted

With an nearly identical setup I upgraded to 10.5.5, all is working (except sleep).

 

OpenGL performances identical to 10.5.4

 

Here are my kext info:

 

Video (mix of ideas from dm_webd and cyclonefr, look here for later method)

AGPGart -> 2.7.1 with AGP_HOST: 0xf0000000

Natit -> with dual x1600 info.plist, all from DM_webd method look here

ATIRadeonX1000 -> vanilla with 71C2 key added

IONDRVSupport -> Vanilla 1.4.3 (from 10.4.7 combo update kexts)

 

ATA support

IOATAFamily.kext -> No version number (For ATA Support)

 

MB support (audio/networking)

AppleAC97Audio.kext -> 1.0.0 (for AC'97 sound support)

AppleAzaliaAudio.kext -> 1.0.0 (for AC'97 sound support)

IONetworkingFamily.kext -> 1.6.0 (for networking)

 

EFI support

AppleSMBIOS.kext -> 1.0.14 (For compatibility with EFI boot)

dsmos.kext -> No version number (Page description EFI module)

 

Rest is 10.5.5 vanilla kext.

Link to comment
Share on other sites
  • 3 weeks later...
Slice
Slice
Posted
  • Author
Posted
Do I have to remove Titan before this will work? It KPs my machine at boot. Also, is the latest version okay for 10.4.11?

Yes, my AGPGart works in systems 10.4.6 - 10.5.5, no matter. But I never see good reports with nVidia card on Intel motherboard.

Sorry, I have no AGP Nvidia to check what is wrong and no other developer to look the problem.

You need to have Titan/natit/nvinject or else to have QE/CI with and without AGPGart.

Link to comment
Share on other sites
scj312
scj312
Posted
Posted
Yes, my AGPGart works in systems 10.4.6 - 10.5.5, no matter. But I never see good reports with nVidia card on Intel motherboard.

Sorry, I have no AGP Nvidia to check what is wrong and no other developer to look the problem.

You need to have Titan/natit/nvinject or else to have QE/CI with and without AGPGart.

So basically I'm SOL getting this to use AGP fully?

Link to comment
Share on other sites
Valdrone
Valdrone
Posted
Posted

For some reason this AGPGart causes a kernel panic with this Geforce 9300GT 512mb and not with my Radeon X1600 Pro 512mb. I've tried installing both attached files but every time I get a kernel panic. Here's the last output: "AGP: Found VIA VT83xx/VT87xx/KTxxx/Px8xx host to PCI bridge

AGP create nub for AGP Card=02e210de bus=1 cap=68" before the panic. 0x02e210de is my card, so that is correct.

 

When I install the modified IOPCIFamily, it doesn't even output extra information about the panic. However, without it, I get the following output towards the end of the panic dump: "com.slice.driver.AGPGart(2.7.1)00x34591000->0x34590fff

dependency: com.apple.iokit.IOPCIFamily(2.4.4)00x2dfb6000"

 

I've set up my injector properly to get display with QE/CI, but there's lots of artifacting and distortion that increases until it becomes unusable. In the system profiler, it gets recognized as PCI instead of AGP (but the VRAM is correct). I'm assuming that the distortion is caused by the reduced bus speed. Can you help me here?

Link to comment
Share on other sites
Slice
Slice
Posted
  • Author
Posted
For some reason this AGPGart causes a kernel panic with this Geforce 9300GT 512mb and not with my Radeon X1600 Pro 512mb. I've tried installing both attached files but every time I get a kernel panic. Here's the last output: "AGP: Found VIA VT83xx/VT87xx/KTxxx/Px8xx host to PCI bridge

AGP create nub for AGP Card=02e210de bus=1 cap=68" before the panic. 0x02e210de is my card, so that is correct.

 

When I install the modified IOPCIFamily, it doesn't even output extra information about the panic. However, without it, I get the following output towards the end of the panic dump: "com.slice.driver.AGPGart(2.7.1)00x34591000->0x34590fff

dependency: com.apple.iokit.IOPCIFamily(2.4.4)00x2dfb6000"

 

I've set up my injector properly to get display with QE/CI, but there's lots of artifacting and distortion that increases until it becomes unusable. In the system profiler, it gets recognized as PCI instead of AGP (but the VRAM is correct). I'm assuming that the distortion is caused by the reduced bus speed. Can you help me here?

As I see you have VIA chipset. It is most complex for me (VIA_LINK) because I have no good documentation how to tune up it and very few test with it.

Did you try AGPGart 2.6.7 or old versions 2.1 and 2.2?

 

A lot of artifact may be cause by incorrect flushing of the chipset. The AGPGart must do it more correctly but I don't know if it right.

Link to comment
Share on other sites
Valdrone
Valdrone
Posted
Posted

No, I haven't tried them, but I'm afraid to, since if it fails I can't even boot using single user mode. Perhaps I might stand a chance of booting into linux and removing it, but last time I checked linux HFS+ write compatibility was failing. I'd be much more willing to test if there was a more specific version to try.

 

http://www.newegg.com/Product/Product.aspx...N82E16813138262

This is my motherboard. (I had a one that was 3x better, but a faulty PSU killed it).

Link to comment
Share on other sites
hotcorez
hotcorez
Posted
Posted

Hello!

 

Try using Transmac or MacDrive on a pc boot. If you are trying to write data from linux to HFS+, make sure you have

 

HFS+ non Journalled.

 

This can be disable easily enough once you have booted back into mac!! Otherwise it may be possible to do this form

 

a boot DVD using terminal?

 

I have alsa heard that if you disable journalling, it will boost your performance.

 

http://community.sonikmatter.com/forums/in...p;mode=threaded

 

Use Google cache to read this article above.

 

Here are the benefits FOR enabling it:-

 

"Journaling" is a feature that helps protect the file system against power outages or hardware component failures,

 

reducing the need for repairs. Journaling was first introduced in Mac OS X Server 10.2.2, then to the non-server OS

 

in Mac OS X 10.3 Panther. This document explains some of the benefits of using this feature and how it works.

 

btw, EXT3 for linux is like HFS+ J'd and EXT2 is like HFS+ non j'd (equivalent, not the same!)

 

Hope this helps.

Link to comment
Share on other sites
Beerkex'd
Beerkex'd
Posted
Posted
Unfortunately, unless I'm mistaken, I can't run EFI on this old P4

 

Oh yes you can, even if it's SSE2 only. Install the Chameleon bootloader.

 

Also; someone in Russia, for the love of god, please send a Nvidia AGP card to Slice!

Link to comment
Share on other sites
Slice
Slice
Posted
  • Author
Posted
Also; someone in Russia, for the love of god, please send a Nvidia AGP card to Slice!

Thank you in advance! I always dream about it... :D

But I prefer by cache or by PayPal.

 

Or help me with theory: how QE/CI is working in MacOSX.

What is missing in AGPGart? Or simply nVidia drivers is not intended to be AGP?

Link to comment
Share on other sites
 Share
Go to topic listing
×
×
  • Create New...