New AGPGart

[PHICO]

SUPPORT LeadTek 7300GT-256Mb-AGP ？？

 

my X1600pro-256Mb-AGP（fine on leopard） died for unknown reason，so i need to buy a new VGAcard

 

between 7300GT-256Mb-AGP & 7600GS-256Mb-AGP,which is better for leopard now??!!!

[Dan Druff]

I've made some tests, hope this will help you Slice.

 

Leo 10.5.2

AGPGart 2.6.5.b2

IONDRVsupport 10.4.5

atindrv 10.4.5

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

 

First, I've made the changes to AGPGart info.plist and set AGP_Base to 0xe0000000.

 

First boot with 256MB Aperture -> No OpenGL -> Starting GioFX crashed the computer (I've no data on this)

Second boot with 64MB Aperture -> No boot but get (1) blue screen, (2) boot to GUI but image all scrambled (icons image mixed up with others on black screen, no data either)

Third boot with 128MB Aperture -> Boot OK and test given below

 

 

After, I've set AGP_Base to 0xf0000000.

 

First boot with 64MB Aperture -> Boot to GUI and test below

Second boot with 128MB Aperture -> Boot to GUI and test below

 

Image of test AGP_Base to 0xe0000000 128MB Aperture   

 

 

Image of test AGP_Base to 0xf0000000 64MB Aperture     

 

 

Image of test AGP_Base to 0xf0000000 128MB Aperture   

 

 

@ Slice

I've been installing Ubuntu on my computer and ran into problems with the ATI X1600 drivers and had to download the proprietary drivers from ATI. It made me think about this, does these drivers, even if coded for Linux, can be useful to you?

dmesg_AGP_Base_0xe0000000_128MB.txt

dmesg_AGP_Base_0xf0000000_64MB.txt

dmesg_AGP_Base_0xf0000000_128MB.txt

ioreg_AGP_Base_0xe0000000_128MB.txt

ioreg_AGP_Base_0xf0000000_64MB.txt

ioreg_AGP_Base_0xf0000000_128MB.txt

[cyclonefr]

Dan tu sais que tu peux utiliser ATINDRV 10.5.2 et ses drivers....Le seul fichier nécessaire est IONDRV .7 voila

 

After this little french introduction just to let u know guys that at this moment tests in Aperture Size = 128Mb are the only settings usefull. For god knows AGPGART doesnt work at this moment on iNTeL but im sure Slice will bend his neck to this Apple's lack of AGP support !

So who claims it does increase anything dare to proove it and run Xbench with and without AGPGART and with a sys profiler snapshot (to check Ci:QE)

[cybland]

Hi Dan Druff,

 

Could you please run the Cinebench tests with and without AGPGart and post screenshots of scores? Your images only show scores with AGPGart. Also I would be very interested to see your Xbench scores, prefereably with and without AGPGart.

[Slice]

Without AGPGart we have no influence on aperture size.

As Dan Druff have different results with different aperture, so AGPGart works! And Cinebench score is not so bad 1600 as I have with my Hackintosh #1.

 

As I see different IONDRVSupport version has influence on result. I can propose you my compilation. Dunno what is the change but smaller code size (more optimized?)

 

There is one more actor to play IOGraphicsFamily that perform connect to GA.plugin and so to AGPDriver. What about test with other version?

[Dan Druff]

Here are the results I have to expect from my GFX Card

 

CINEBENCH R10

****************************************************

Tester : Processor : Intel® Celeron® CPU 2.53GHz

MHz : 2525 Mhz

Number of CPUs : 1

Operating System : WINDOWS 32 BIT 5.1.2600

 

Graphics Card : Sapphire RADEON X1600 PRO

Resolution : 1400-1050

Color Depth : 32 bits

 

****************************************************

 

Rendering (Single CPU): --- CB-CPU

Rendering (Multiple CPU): --- CB-CPU

 

Shading (OpenGL Standard) : 2082 CB-GFX

 

****************************************************

 

Also, I'd like to point that GioFX scores are always lower with AGPGart, any idea why this pattern of results is the opposite of Cinebench?

[riws]

Also, I'd like to point that GioFX scores are always lower with AGPGart, any idea why this pattern of results is the opposite of Cinebench?

How you know it? None benchmark runs or from the FPS measure?

[Dan Druff]

I'm mistaken, ATINDRV.kext has a big effect, look at the tests I've just run.

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport 10.4.5

atindrv 10.5.2

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

Look at the OpenGL tests:  

 

From my OpenGL results from Cinebench are crappy (with tearing in bonus!) but GioFX runs better than ever (in fact better than with IONDRVsupport 10.5.2).But, the real new point is that my card is seen as AGP with those results onGioFX, never able to achieve that before!

 

@ RiwsFPS Measure...

 

Look at the previous posts plz.

dmesg_AGP_Base_0xf0000000_128MB_ATINDRV_10.5.2.txt

ioreg_AGP_Base_0xf0000000_128MB_ATINDRV_10.5.2.txt

[Dan Druff]

I've tested this setup

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport provided by Slice

atindrv 10.5.2

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

No boot to GUI, no that's not true. I've got to GUI, resolution was 1400x1050, I had mouse and blue screen... but nothing else. Two rebbots and nothing. So the IONDRVsupport from Slice is incompatible with ATINDRV from 10.5.2.

 

 

I've tested another setup

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport provided by Slice

atindrv 10.4.5

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

No boot to GUI either. Again, I've got to GUI, resolution was 1400x1050, I had mouse and blue screen... but nothing else. Two rebbots and nothing. So the IONDRVsupport from Slice seems incompatible with ATINDRV of any version.

 

I made additionnal tests

 

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport 10.4.5

atindrv 10.4.5

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

Here are the results:   

 

As you see, My setup is identical to the setup in my post 452 but my results are less shiny. There's a change that I've lost, and it seems critical... I'll try to find out what change is it.

 

 

And this last setup too

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport 10.4.5

atindrv 10.4.7

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

Here are the results:   

 

In this last setup, I've got image tearing, like with 10.5.2, in all not a good combination.

 

 

Edit

When I'm in the blue screen with my mouse, I can move my mouse, hear the HDD do their stuff. But also I can shut down the computer WITH the mouse (apple -> Shutdown) and if I miss Shutdown and press Sleep, the GUI appears briefly before the sleep (that doesn't wake up on my comp!). Then, even if the kexts aren't able to display the desktop on the monitor, they are able to get the GFX card to work.

dmesg_IONDRVsupport_10.4.5_ATINDRV_10.4.5.txt

ioreg_IONDRVsupport_10.4.5_ATINDRV_10.4.5.txt

dmesg_IONDRVsupport_10.4.5_ATINDRV_10.4.7.txt

ioreg_IONDRVsupport_10.4.5_ATINDRV_10.4.7.txt

[Dan Druff]

I was thinking about that, if AGP support still exist for PPC mac comp, it's because there's a need for that.

But, as there's no Intel mac with AGP bus, it may had never been written for Leopard because there's no need to write this piece of software.

 

What do you think?

[cyclonefr]

u ave to use ATINDRV 10.5.2 and IONDRV .7 DAN that's the only version that works on X1600 AGP

[Dan Druff]

@cyclonefr

 

Explain me where to find IONDRVsupport 10.4.7, it's not in the Apple 10.4.7 combo update package (I explored the package with Pacifist, it's not there)!

 

If you have a version tagged 10.4.7 that is comming from elswhere than Apple, then it's not 10.4.7, but a mod version.

 

Could you send me your version? Official or not, if it can make AGP works!

[Slice]

I was thinking about that, if AGP support still exist for PPC mac comp, it's because there's a need for that.

But, as there's no Intel mac with AGP bus, it may had never been written for Leopard because there's no need to write this piece of software.

 

What do you think?

You are right. AGP support for PowerPC exists by the driver AppleMacRiscPCI which contains hardware specific codes. But no analogous driver by Apple for Intel hardware. And more. IONDRVSupport calls specific driver "Driver,AAPL,MacOS,PowerPC" obtained from EFI. Because of lack of the driver we have problems with AGP Graphics. Is it deadend? I think no! Because AGPGart works fine with AMD64 on Tiger. So Tiger contains enough support for AGP graphics.

 

Thank you Dan for numerous attempts! It is very useful for the problem. One more attempt: what about IOGraphicsFamily from Tiger? I have 10.4.6 so I upload default kext from my system.

 

But 1 strange question. Why my version of IONRDVSupport don't work??? It is clear recompilation of 10.4.11 kext. It works for my Tiger. I'll search a difference with 10.4.7. It might be a key for the problem.

IOGraphicsFamily.kext.zip

IONDRVSupport.kext.zip

[yorkie]

I've got Leopard 10.5.2 on an AMD (K8T800 chipset) with Nivdia 6800GT 256MB using NVinject 0.2.0. I'd like to get AGP working. I can't tell from this thread whether AGPgart is known to be working on this system, and if so which files I need to download?

 

If it's not working yet I am happy to try anything for you and give you the results.

[cyclonefr]

1) to SLice : IONDRVSUpport 10.4.5 and 10.4.7 are the only version working on X1600 cause it does seek for 71C2 (first cpu on the card) to send the signal to the outup (VGA/DVI). When u use IONDRVSupport 10.4.11 for example it will result in a blue screen : no video signal is send to the right CPU core. So if u base ur compilation on .7 version there shouldnt be any problem.

 

2) Dan Druff : the file was in 10.4.7 COMBO i dont know why u couldnt find it maybe it's the same as .5 i may be wrong...Anyway ive already attached the file in the X1600 AGP Topic u can easily find it. But mainly what I was saying is X1600AGP can work without artifacts etc with ATINDRV 10.5.2 and original 10.5.2 X1600 drivers. U wont get AGP enabled and accelerated AGP at the moment, that's why Slice is working like a log But ATINDRV 10.5.2 is more optimized u'll get a little more perf. Enjoy

[Slice]

I've got Leopard 10.5.2 on an AMD (K8T800 chipset) with Nivdia 6800GT 256MB using NVinject 0.2.0. I'd like to get AGP working. I can't tell from this thread whether AGPgart is known to be working on this system, and if so which files I need to download?

 

If it's not working yet I am happy to try anything for you and give you the results.

I correct topic with new link. Look it!

 

1) to SLice : IONDRVSUpport 10.4.5 and 10.4.7 are the only version working on X1600 cause it does seek for 71C2 (first cpu on the card) to send the signal to the outup (VGA/DVI). When u use IONDRVSupport 10.4.11 for example it will result in a blue screen : no video signal is send to the right CPU core. So if u base ur compilation on .7 version there shouldnt be any problem.

I'll check sources. It might be interesting.

Versions 10.4.3 -.5 -.6 .-7 are the same.

Thank for the information.

[yorkie]

I correct topic with new link. Look it!

 

Thanks, I've installed but get black screen. I've booted in safe mode. Output of dmesg | grep AGP is

AGP: Found VIA  AGP bridge for AMD
AGPBridge buses: pri 0 secBus 1 subBus 1 BridgeDev 0
AGPgart: BridgeMemoryRange [f8000000, faffffff]
AGPgart: BridgePrefetchableMemoryRange [e0000000, efffffff]
AGPgart: BridgeIORange [0000f000, 00000fff]
AGP: saveBridgeState OK
AGP create nub for 004510de bus=1 cap=68
AGPGart: publishNub display

 

Attached is output of ioreg -l -x -w 1024 >ioreg.txt

 

is this any use?

ioreg.txt.zip

[Slice]

Thanks, I've installed but get black screen. I've booted in safe mode. Output of dmesg | grep AGP is

AGP: Found VIA  AGP bridge for AMD
AGPBridge buses: pri 0 secBus 1 subBus 1 BridgeDev 0
AGPgart: BridgeMemoryRange [f8000000, faffffff]
AGPgart: BridgePrefetchableMemoryRange [e0000000, efffffff]
AGPgart: BridgeIORange [0000f000, 00000fff]
AGP: saveBridgeState OK
AGP create nub for 004510de bus=1 cap=68
AGPGart: publishNub display

It is dmesg from initialization. But where is starting of AGPGart?

 

Attached is output of ioreg -l -x -w 1024 >ioreg.txt

is this any use?

I see you use AGP_Base=0xe0000000. For your case it is equal to VRAM_Address. Not good.

It is better for you AGP_Base=0xf0000000.

 

I don't see any AGPTarget in your system. What version of IOPCIFamily do you use?

Check your BIOS settings about graphics. Aperture=128Mb, Primary VGA=AGP and so on.

[yorkie]

Sorry, that was the log from the boot in safe mode. I've tried again, with IOPCIFamily 2.4.4, and a base of 0xf0000000. Here are the log messages:

AGP: Found VIA  AGP bridge for AMD
AGPBridge buses: pri 0 secBus 1 subBus 255 BridgeDev 0
AGPGart subBus corrected to 1
AGPgart: BridgeMemoryRange [f8000000, faffffff]
AGPgart: BridgePrefetchableMemoryRange [e0000000, efffffff]
AGPgart: BridgeIORange [0000f000, 00000fff]
AGP: saveBridgeState OKAppleVIAATADriver: VIA 8237 SATA (CMD 0xe900, CTR 0xea00, IRQ 20, BM 0xe200)
AGP create nub for 004510de bus=1 cap=68
AGPGart: publishNub display
AGP: Coherence support: no
AGP: GART is 32 bit capable
AGP: Found an AGP 3.0 compliant device.
AGPAMD: 1 hammers found
AGPAMD apbase=f0000000
AGPAMD: aperture [f0000000, 08000000]
AGPVIA gart installed 04267000
AGPAMD trace hammer PCI space
AGPAMD trace host PCI space
AGP: Setting 08 data rate
AGP: command written target=00000b02 master=1f00e302
AGP memory 1b862000 length 00001000 offset 00000000 cnt 00000000
AGP memory 1b862000 length 00001000 offset 00001000 cnt 00000001
AGP memory 1b862000 length 00001000 offset 00002000 cnt 00000002

 

I've attached another ioreg output.

ioreg.txt.zip

[Dan Druff]

2) Dan Druff : the file was in 10.4.7 COMBO i dont know why u couldnt find it maybe it's the same as .5 i may be wrong...Anyway ive already attached the file in the X1600 AGP Topic u can easily find it. But mainly what I was saying is X1600AGP can work without artifacts etc with ATINDRV 10.5.2 and original 10.5.2 X1600 drivers. U wont get AGP enabled and accelerated AGP at the moment, that's why Slice is working like a log But ATINDRV 10.5.2 is more optimized u'll get a little more perf. Enjoy

My PCI config from post 404 is configured like that, all 10.5.2 w/ graph update and I get image tearing. I know it's the graph update that did that, but once installed, it's quite hard to remove it without reinstalling. But even with the graph update installed, I can achieve setups w/o image tearing (see my previous posts) and I still have the OpenGL updated engine from the Graph update. Beside that, the image tearing isn't very annoying. It happens often, but any image update (i.e. closing a window, updating the image or anything that reupdate windows server) do the trick. Anyway thanks for the tip, it may help others.

1) to SLice : IONDRVSUpport 10.4.5 and 10.4.7 are the only version working on X1600 cause it does seek for 71C2 (first cpu on the card) to send the signal to the outup (VGA/DVI). When u use IONDRVSupport 10.4.11 for example it will result in a blue screen : no video signal is send to the right CPU core. So if u base ur compilation on .7 version there shouldnt be any problem.

I was re-reading your post and I got an idea, if we could just add 0x71E2 (the Dev ID of the second GPU) to the ATIRadeonX1000.kext, data could be properly processed by the second GPU. Once I get home, I'll try that and give my results.

[Slice]

Sorry, that was the log from the boot in safe mode. I've tried again, with IOPCIFamily 2.4.4, and a base of 0xf0000000. Here are the log messages:

AGP: Found VIA  AGP bridge for AMD
AGP: command written target=00000b02 master=1f00e302
AGP memory 1b862000 length 00001000 offset 00000000 cnt 00000000
AGP memory 1b862000 length 00001000 offset 00001000 cnt 00000001
AGP memory 1b862000 length 00001000 offset 00002000 cnt 00000002

Yes, It means graphics data go through AGP bus. And where is it? I don't know. Common Sense with the same config were successful in Tiger and have black screen in Leo. He didn't solve the problem. I don't know Leo at all. May be some kext replacement with Tiger versions?

 

I was re-reading your post and I got an idea, if we could just add 0x71E2 (the Dev ID of the second GPU) to the ATIRadeonX1000.kext, data could be properly processed by the second GPU. Once I get home, I'll try that and give my results.

The device 71e2 have class code 0x038000 so it is not VGA device. I don't know.

Did you try to change IOGraphicsFamily.kext?

And about your tests. You have in Windows

Shading (OpenGL Standard) : 2082 CB-GFX

And in MacOSX 1636. What do you want else?

I have in Windows score=2500 and 1600 in MacOSX with perfect(?) NVidia PCIE driver.

Do you expect that AGPGart increase your score to 5000?

 

2 CycloneFR

You have bad assignment for gfx string about timings

Your value

"ATY,SCLK" =

"ATY,RefCLK" absent

Izmael, Dan Druff,

"ATY,SCLK" =

"ATY,RefCLK" =

Any other values?

[Dan Druff]

And in MacOSX 1636. What do you want else?

I have in Windows score=2500 and 1600 in MacOSX with perfect(?) NVidia PCIE driver.

Do you expect that AGPGart increase your score to 5000?

 

I don't really understand the question... First, I did get 1636, but now I'm capped to ~1000.

 

After each test I replaced my kext with 10.5.2 kexts as I noted all changes, so after this test,

i did the same. But now I'm unable to get again these scores. There must have been a change that I didn't logged, a shame...

 

Also it should be noted that even if CB was high, GioFx wasn't improved at all.

 

Then to answer to your question, I don't want a functionnal OSX, I know how to make one, I want full AGP support on my Hack. Why? For the fun of it, to make something not Mac work as a Mac. This isn't for games or stuff like that, I have Windoze. And as I'm not able to make this work by myself, I try to help you to do so. So my final answer to your question is this one: Make it works!

[Dan Druff]

Here I will make a review of my latest setups

 

#1

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport 10.4.5

atindrv 10.5.2

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

From my OpenGL results from Cinebench are crappy (with tearing in bonus!) but GioFX runs better than ever (in fact better than with IONDRVsupport 10.5.2).But, the real new point is that my card is seen as AGP with those results onGioFX, never able to achieve that before!

 

 

#2

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport -> From Slice (V1)

atindrv 10.5.2

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

No boot to GUI, no that's not true. I've got to GUI, resolution was 1400x1050, I had mouse and blue screen... but nothing else. Two rebbots and nothing. So the IONDRVsupport from Slice is incompatible with ATINDRV from 10.5.2.

 

 

#3

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport -> From Slice (V1)

atindrv 10.4.5

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

No boot to GUI either. Again, I've got to GUI, resolution was 1400x1050, I had mouse and blue screen... but nothing else. Two rebbots and nothing. So the IONDRVsupport from Slice seems incompatible with ATINDRV of any version.

 

 

#4

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport 10.4.5

atindrv 10.4.5

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

As you see, My setup is identical to the setup in my post 452 but my results are less shiny. There's a change that I've lost, and it seems critical... I'll try to find out what change is it.

 

#5

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport 10.4.5

atindrv 10.4.7

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

In this last setup, I've got image tearing, like with 10.5.2, in all not a good combination.

 

 

I've made another boot with the 71E2 mod I was proposing...

 

#6

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport -> From Slice (V1)

atindrv 10.4.5

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 + 71E2 key added

 

No boot (2 trials) and no data.

 

 

I've made two other tests

 

#7

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport -> From Slice (V2)

atindrv 10.4.5

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

Boot to GUI and no problem (i.e. no tearing)

Shading (OpenGL Standard): 1611 CB-GFX (idem to performance under Vista, which is really a piece of junk!)

GioFX: ~445

 

This is a setup with high end OpenGL like my "lost" setup. Good to know I managed to recreate it!

 

Sreenshot   

 

 

#8

Leo 10.5.2

AGPGart 2.6.5.b2 -> AGP_Base_0xf0000000

IONDRVsupport -> From Slice (V2)

atindrv 10.4.5

IOGraphicsFamily -> From Slice

Graph kext -> Graph update 1.0 with mod to ATIRadeonX1000.kext with 71C2 key added

 

Boot w/o problems, no tearing.

Shading (OpenGL Standard): 1418

GioFX: ~443

 

This setup has a less shiny OpenGL output from CB, but in all it's a good setup. It seems that IOGraphicFamily has an impact upon the OpenGL from CB. From 1611 to 1418 with only this kext being different.

 

Sreenshot   

dmesg_7.txt

ioreg_7.txt

dmesg_8.txt

ioreg_8.txt

[cyclonefr]

slice i'll resend u my ioreg since i now use PCEFI V8 harcoded with original X1600 ioreg from mbp ok? u'll let me know if my settings are better

anyway since AGP doesnt work it wont change a lot lol

[topcatsteve]

Apature setting

 

I have a laptop (M6Ne ASUS) with an ATI 9600 M10 64Mb Apature I am trying to get to work with Leo4All and have had some basic sucess (1024x786x32) and AGPGart 2.n. I would like to get the full resolution of (1280x800x32) and belive a update of the AGPGart might do it. So I replased my with 2.6.

 

But on the verbose startup/boot screens all the right things came up until I gote a worning about setting my Apature in my BIOS. Opps my BIOS dous not have a setting for the Apature! Is there any way of setting it in the .plist or something? The AGPGart gave a Apature address of D0000000:DFFFFFFF where as I need E0000000:EFFFFFFF can I manualy set this?

 

Thanks Steve

 

Sorry about the spelling I am dislex....