Getting closer - An ATI graphics guide

[spamyoutodeath]

Thanks to all for your work on this! It took me ages just to get proper resolution on my mobility 9000 without this. the pack works fine here, i just wish i'd seen it sooner!

[tyre777]

im using a 9600 M goodtime, and using callisto 3, umm we get severe window tearing with qe and ci at 1400x1050, is there a way to get rid of that too?

[imjustabill]

Worked flawlessly on My Averatec 6130 laptop with a 9600 Mobility. THanks!!!

[Racko]

Does this driver work on X300 graphic card?

[Guest goodtime]

im using a 9600 M goodtime, and using callisto 3, umm we get severe window tearing with qe and ci at 1400x1050, is there a way to get rid of that too?

 

 

Which driver pack are you using? CompaqX1000.zip or the Radeon9200pack.zip?

 

If it's the CompaqX1000 try this one:

 

http://art5dog.com/Radeon9200pack.zip

 

Also, you I may need to create a separate pack the 9600 M users. The OpenGL drivers are removed for 9200 users because the driver would crash. If you have QE/CI then OpenGL drivers should work. I can add it back in for 9600 users and you might have everything working.

[Guest goodtime]

im using a 9600 M goodtime, and using callisto 3, umm we get severe window tearing with qe and ci at 1400x1050, is there a way to get rid of that too?

 

Upgrade to AGPGart 2.0 should fix the problem. It is in my latest driver pack:

 

http://art5dog.com/Radeon9200pack.zip

 

If you have some custom extensions, you can install AGPGart 2.0 manually.

[artic_fox]

Goodtime,

 

Thanks for releasing your release. I have a Dell D600 w/ Mobility 9000. I couldn't get the JaS Ati.kext to do 3D HW excelleration either. Same kload error as other people were seeing.

 

However, I did want to comment about your suggestion to turn Speedstep off in the BIOS. In my BIOS, for the Dell D600 at least there is a comment that turning off SpeedStep will will default to turning the CPU at lowest performance. Well It definately does wack out everything turning SpeedStep off. Everything slows down and very noticable in the graphics.

 

However although Xbench agrees with your assertions that turning off Speedstep will make things run faster, I don't by it.. At least on the Dell D600.

 

Here is what Xbench 1.3 report with SpeedStep turned off:

 

Results 67.77

System Info

Xbench Version 1.3

System Version 10.4.6 (8I1119)

Physical RAM 512 MB

Model ADP2,1

Drive Type HTS548040M9AT00

CPU Test 43.36

GCD Loop 80.97 4.27 Mops/sec

Floating Point Basic 36.12 858.11 Mflop/sec

vecLib FFT 31.38 1.04 Gflop/sec

Floating Point Library 49.14 8.56 Mops/sec

Thread Test 62.36

Computation 42.61 863.22 Kops/sec, 4 threads

Lock Contention 116.26 5.00 Mlocks/sec, 4 threads

Memory Test 81.97

System 80.77

Allocate 74.15 272.31 Kalloc/sec

Fill 98.95 4811.12 MB/sec

Copy 73.81 1524.55 MB/sec

Stream 83.21

Copy 75.45 1558.46 MB/sec

Scale 76.37 1577.84 MB/sec

Add 92.46 1969.54 MB/sec

Triad 91.66 1960.78 MB/sec

Quartz Graphics Test 62.30

Line 49.29 3.28 Klines/sec [50% alpha]

Rectangle 59.45 17.75 Krects/sec [50% alpha]

Circle 55.10 4.49 Kcircles/sec [50% alpha]

Bezier 77.02 1.94 Kbeziers/sec [50% alpha]

Text 83.26 5.21 Kchars/sec

User Interface Test 104.11

Elements 104.11 477.80 refresh/sec

Disk Test 86.33

Sequential 142.23

Uncached Write 148.65 91.27 MB/sec [4K blocks]

Uncached Write 145.67 82.42 MB/sec [256K blocks]

Uncached Read 124.52 36.44 MB/sec [4K blocks]

Uncached Read 153.86 77.33 MB/sec [256K blocks]

Random 61.97

Uncached Write 21.43 2.27 MB/sec [4K blocks]

Uncached Write 153.36 49.09 MB/sec [256K blocks]

Uncached Read 152.71 1.08 MB/sec [4K blocks]

Uncached Read 207.99 38.59 MB/sec [256K blocks]

 

Here is what Xbench reports with SpeedStep turned back on:

Results 37.42

System Info

Xbench Version 1.3

System Version 10.4.6 (8I1119)

Physical RAM 512 MB

Model ADP2,1

Drive Type HTS548040M9AT00

CPU Test 40.92

GCD Loop 67.72 3.57 Mops/sec

Floating Point Basic 40.71 967.38 Mflop/sec

vecLib FFT 27.91 920.78 Mflop/sec

Floating Point Library 44.27 7.71 Mops/sec

Thread Test 57.36

Computation 38.75 784.96 Kops/sec, 4 threads

Lock Contention 110.35 4.75 Mlocks/sec, 4 threads

Memory Test 37.24

System 42.46

Allocate 56.46 207.36 Kalloc/sec

Fill 50.15 2438.33 MB/sec

Copy 30.29 625.70 MB/sec

Stream 33.16

Copy 29.71 613.75 MB/sec

Scale 30.32 626.31 MB/sec

Add 37.06 789.47 MB/sec

Triad 37.02 791.90 MB/sec

Quartz Graphics Test 30.86

Line 28.38 1.89 Klines/sec [50% alpha]

Rectangle 20.49 6.12 Krects/sec [50% alpha]

Circle 23.21 1.89 Kcircles/sec [50% alpha]

Bezier 59.16 1.49 Kbeziers/sec [50% alpha]

Text 55.64 3.48 Kchars/sec

User Interface Test 32.87

Elements 32.87 150.85 refresh/sec

Disk Test 34.72

Sequential 64.67

Uncached Write 66.50 40.83 MB/sec [4K blocks]

Uncached Write 58.67 33.19 MB/sec [256K blocks]

Uncached Read 73.29 21.45 MB/sec [4K blocks]

Uncached Read 62.03 31.18 MB/sec [256K blocks]

Random 23.73

Uncached Write 7.98 0.85 MB/sec [4K blocks]

Uncached Write 63.37 20.29 MB/sec [256K blocks]

Uncached Read 64.26 0.46 MB/sec [4K blocks]

Uncached Read 83.58 15.51 MB/sec [256K blocks]

 

 

Although the Xbench results are much lower with SpeedStep turned ON, they are much more believable.

With SpeedStep turn OFF, how can my laptop's 5400 RPM 2.5" HDD throughput exceed that of the Xbench baseline score of 100 which is a Dual G5 with SATA 7200 RPM drives? And twice the memory throughput??? A Score of almost 70 on a 1.4 Ghz Pentium M? no way. Plus when in this configuration everything runs like a dog. iTunes struggles to play MP3s, when I turn on the visualizer I get about 1 FPS. When I turn SpeedStep back ON (per the BIOS default) everything is back to its snappy self. MP3's play fine, and turning on the iTunes visualizer give me about 8 fps

 

So turning OFF SpeedStep in the BIOS will not always give you better performance. In my case, on my Dell D600 my performance is aweful with SpeedStep turned OFF, and I don't care what Xbench reports.. I don't believe it one bit.

 

Thanks for your hard work on the Mobility 9000 graphics. I am excited by what you have achieved thus far and look forward reading your progress in the future.

 

BTW, I didn't benchmark fps with your Radeon9200 release. This was with the JaS Ati.kext. But I did run Xbench on both your Radeon9200 release and the JaS Ati.kext and results are equiv. Both with SpeedSpeed on and off. ON is much better for me, OFF is a dog.

I will report the fps in iTunes visualizer with your Radeon9200 release in my next post.

 

Thanks!

[artic_fox]

ok... just did my iTunes visualizer test with your 9200 release.

I get the same as with the JaS Ati.kext. 8 fps.

 

 

btw, if I run the visualizer in full screen mode then I get 24 fps. However, I want to point out the full screen appears to be 640x480 resolution. My 8fps result with running iTunes visualizer in a "large" window.

 

Perhaps this explains the 24 fps results reported eariler by captainsquash with JaS Ati.kext

[Guest goodtime]

Artic_Fox:

 

On my Compaq X1000 laptop, with Speedstep enabled it crawls. With Speedstep disabled it's normal.

 

I guess it varies from BIOS to BIOS. Thank you for pointing it out. As long as you have your BIOS settings under control, then your OSx86 UI experience will be much better.

 

This release does not solve the 3D Acceleration/OpenGL Problem. My G3 Powerbook blows any 9200 OpenGL test out of the water and the old Powerbook only has an 8MB ATI Rage card with no QE.

 

The Radeon 9200 doesn't have 3D/OpenGL Acceleration at the GPU yet, but I think we will get the get there after XRadeonFB and then things should be much better. Hardware OpenGL will be much better for iTunes, Chess, DVD Playback, and other 3D operations.

 

Thank you for trying it out and I appreciate the feedback. One day we will will figure out OpenGL and QE.

 

Goodtime

[Guest m41k]

Just a heads up, Aperture 1.5 needs 10.4.8.

 

Yep, but it only depends on the QuartzCore.framework from 10.4.8

So everyone who got 10.4.5/6/7 might just copy that framework from .4.8

[Guest goodtime]

Yep, but it only depends on the QuartzCore.framework from 10.4.8

So everyone who got 10.4.5/6/7 might just copy that framework from .4.8

 

m41k,

 

Is there any speed benefit from 10.4.8's QuartzCore.framework? Just curious if you noticed anything.

 

GT

[perfect0]

Upgrade to AGPGart 2.0 should fix the problem. It is in my latest driver pack:

 

http://art5dog.com/Radeon9200pack.zip

 

If you have some custom extensions, you can install AGPGart 2.0 manually.

 

Just wondering if I can install your AGPGart.kext. I have an Insprion 9200 that has various resolutions and QE and CI supported thanks to the Callisto tweaks although I still have artifacts and mouse tearing. I have an ATI Mobility Radeon 9700, would you think the AGPGart.kext will work for me? If so, into where do I move the file? Thanks in advance...

[Guest goodtime]

Just wondering if I can install your AGPGart.kext. I have an Insprion 9200 that has various resolutions and QE and CI supported thanks to the Callisto tweaks although I still have artifacts and mouse tearing. I have an ATI Mobility Radeon 9700, would you think the AGPGart.kext will work for me? If so, into where do I move the file? Thanks in advance...

 

No, I don't think the driver pack will help tearing with QE/CI. It only fixes a resolution problem with a 2D Enabled 9200 Radeon Card. It basically removes the out of spec 2D Acceleration in favor of better stability.

 

Btw, it's not my AGPGart, it's joblo10's. I only tested it and added it to my driver pack.

 

GT

[Guest goodtime]

Btw, if you want to know if you have 2D Acceleration.

 

Open the terminal:

 

ioreg -l

 

Find Radeon8500

 

and look for this:

 

|

| | | +-o ATIRadeon8500 <class ATIRadeon8500, registered, matched, a$

| | | | {

| | | | "AccelCaps" = 0

| | | | "IOGLBundleName" = "ATIRadeon8500GLDriver"

| | | | "IOSourceVersion" = "1.4.20.28"

| | | | "CFBundleIdentifier" = "com.apple.ATIRadeon8500"

| | | | "IOClass" = "ATIRadeon8500"

| | | | "IODVDBundleName" = "ATIRadeon8500VADriver"

| | | | "IOProbeScore" = 0

| | | | "IOMatchCategory" = "IOAccelerator"

| | | | "IOPCIMatch" = "0x514C1002 0x516C1002 0x4C661002 0x496610$

| | | | "IOAccelRevision" = 2

| | | | "IOCFPlugInTypes" = {"ACCF0000-0000-0000-0000-000a2789904$

| | | | "IOProviderClass" = "IOPCIDevice"

| | | | "PerformanceStatistics" = {"vramFreeBytes"=60012544,"surf$

| | | | }

| | | |

| | | +-o ATIR2002DContext <class ATIR2002DContext, !registered, !$

| | | | {

| | | | }

| | | |

| | | +-o ATIR200Surface <class ATIR200Surface, !registered, !matc$

| | | {

| | | "IOUserClientCrossEndianCompatible" = Yes

| | | }

 

 

ATI R200 2D Context is loaded

ATI R200 Surface is loaded

 

These are what make up 2D Accel. on an ATI Video Card.

 

Also there are no R200 GL Drivers loaded. It doesn't show them listed.

 

GT

[t0ffluss]

Just wondering if I can install your AGPGart.kext. I have an Insprion 9200 that has various resolutions and QE and CI supported thanks to the Callisto tweaks although I still have artifacts and mouse tearing. I have an ATI Mobility Radeon 9700, would you think the AGPGart.kext will work for me? If so, into where do I move the file? Thanks in advance...

 

Did you ever try the AGPGart.kext file?.. it probably wont work.. I too have 9700 mobility, running at 1400x1050) QE/CI supported, get alot of Mouse tearing and heavy artifacts when scrolling. (pressing F11 twice fixes the problem)

 

 

Richard

[kasakks]

my ati mob x600 was working fine on 10.4.3 on 1400 x 900 now i installed 10.4.6 and callisto v8 crashed before booting... v6 boots into it.. when i try to cahnge resolution it crashed .... v3 is not working at all.. any idea?

[t0ffluss]

Thought I´d share a video, to illustrate the graphics issues I have.. not a big problem, new windows needs to go through a "dual F11 key press", new windows shows alot of artifacts when moving them/ scrolling. But whats weird, pressing F11 twice removes this problem, and it NEVER happens again (with that window), new windows have the problem, again, pressing F11 fixes it.

 

video (3GP), sorry for bad quality, captured it on my W810i mobil phone, but it basicly shows what Im experiencing:

 

Video

 

 

Richard

[Guest m41k]

m41k,

 

Is there any speed benefit from 10.4.8's QuartzCore.framework? Just curious if you noticed anything.

 

GT

 

 

Goodtime,

 

i didn't noticed any speed benefit.

I was just playing around with 4.8 update to find out what files required to get Aperture running on earlier versions like 10.4.5.

Without it i wasn't able to use the menu or view/import any pics etc.

 

greets

[perfect0]

Sorry that I haven't updated my post here, I no longer have mouse tearing. I followed a post on installing something for the pointer that does not tear.

[Guest goodtime]

Sorry that I haven't updated my post here, I no longer have mouse tearing. I followed a post on installing something for the pointer that does not tear.

 

I think Mouse Tearing will become a thing of the past once 10.4.8 Kernel kinks are worked out for SSE2, SSE3 and AMD.

[nastyhome]

Will we see better ATI drivers (for 9200) on the new 10.4.8?

[GoMac]

Goodtime,

 

I have a Radeon 9250:

 

Adapter description......... ASUS A9250

Adapter device ID........... 0x59601002

Display settings............ 1280x1024, 32bpp

 

Do you think your 9200 pack could work?

 

I can get it to 1280x1024x32 max. I tried Callisto b8 yesterday, it loaded but I got a black screen afterwards. I guess I have to use the installation disc and start a terminal to delete the Callisto kext to get my system back.

 

Any advice is very much appreciated!

 

GoMac

[joe75]

I think Mouse Tearing will become a thing of the past once 10.4.8 Kernel kinks are worked out for SSE2, SSE3 and AMD.

 

 

I'll keep my fingers crossed

[Guest goodtime]

Will we see better ATI drivers (for 9200) on the new 10.4.8?

 

Yes and No. At least not until there is and SSE3 emulator in the 10.4.8 kernel. We'll have to wait and see. I am hoping the Radeon 8500 drivers will work without a hitch.

 

GT

[GoMac]

Success!

 

Following Goodtime's advice, I changed the device ID in of the CallistoHAL.kext in his Radeon9200Pack to my Radeon 9250 Device ID (0x59601002). After runnig the Applescript in the pack, I restarted and got all resolutions to choose from. So now I am enjoying my full 1600x1200x32. No QE / acceleration though.

 

Thanks for your help Goodtime!

 

 

Here my graphics card/display dump:

 

Monitor

Windows description......... SyncMaster 204T(Digital)

Manufacturer description.... SyncMaster

Manufacturer................ Samsung

————————————————————————————

Plug and Play ID............ SAM01AE

Serial number............... HCHY707169

EDID data source............ I2C bus (real-time)

————————————————————————————

Manufacture date............ 2005, ISO week 28

EDID revision............... 1.3

Display type and signal..... Digital

Sync input support.......... n/a

Screen size................. 410 x 310 mm (~22")

Power management............ Active off/sleep

 

Color characteristics

Display gamma............... 2.20

Red chromaticity............ Rx 0.640 - Ry 0.330

Green chromaticity.......... Gx 0.300 - Gy 0.600

Blue chromaticity........... Bx 0.150 - By 0.060

White point (default)....... Wx 0.313 - Wy 0.329

 

Timing characteristics

VESA GTF support............ Not supported

Horizontal scan range....... 30-81kHz

Vertical scan range......... 56-75Hz

Video bandwidth............. 170MHz

Extension blocks............ n/a

Timing recommendation #1.... 1600x1200 at 60Hz

Modeline................ "1600x1200" 162.000 1600 1664 1856 2160 1200 1201 1204 1250 +hsync +vsync

 

Standard timings supported

640 x 480 at 60Hz - IBM VGA

640 x 480 at 67Hz - Mac II

640 x 480 at 72Hz - VESA

640 x 480 at 75Hz - VESA

720 x 400 at 70Hz - IBM VGA

800 x 600 at 56Hz - VESA

800 x 600 at 60Hz - VESA

800 x 600 at 72Hz - VESA

800 x 600 at 75Hz - VESA

832 x 624 at 75Hz - Mac II

1024 x 768 at 60Hz - VESA

1024 x 768 at 70Hz - VESA

1024 x 768 at 75Hz - VESA

1152 x 864 at 75Hz - VESA

1152 x 870 at 75Hz - Mac II

1280 x 960 at 60Hz - VESA

1280 x 1024 at 60Hz - VESA

1280 x 1024 at 75Hz - VESA

1600 x 1200 at 60Hz - Samsung

1600 x 1200 at 60Hz - VESA

 

Raw EDID base

00: 00 FF FF FF FF FF FF 00 4C 2D AE 01 30 32 52 42

10: 1C 0F 01 03 80 29 1F 78 2A EE 95 A3 54 4C 99 26

20: 0F 50 54 BF EF 80 A9 40 81 80 81 40 71 4F 01 01

30: 01 01 01 01 01 01 48 3F 40 30 62 B0 32 40 40 C0

40: 13 00 98 32 11 00 00 1E 00 00 00 FD 00 38 4B 1E

50: 51 11 00 0A 20 20 20 20 20 20 00 00 00 FC 00 53

60: 79 6E 63 4D 61 73 74 65 72 0A 20 20 00 00 00 FF

70: 00 48 43 48 59 37 30 37 31 36 39 0A 20 20 00 23

 

Display adapter

Adapter description......... ASUS A9250

Adapter device ID........... 0x59601002

Display settings............ 1280x1024, 32bpp