Re: iBook with ATI Rage Mobility 128 (AGP 2x)

1
>Hi Denis,
>
>I have an iBook that "MacTracker" describes as:
> iBook (14") "Son of Pismo" (introduced October 2001)
> It has a 600 MHz G3 processor and a 14.1" LCD screen.
> Graphics: "ATI Rage Mobility 128 (AGP 2x)"
>
>The Apple System Profiler tells me my graphics is:
> card name: ATY,RageM3p29s
> card model: ATY,RageM3
> card revision: 2
>
>So, that is a long way of answering "Yes",
>I think I have a Rage "Mobility" chip.
>
>However, when I run ScreenTest, I only get one line of info about CopyWindow.
>That is, there are no lines telling me about VRAM or AGP.
>This is while I am running Mac OSX, with MATLAB in Classic mode.
>
>I assume it is the same under Mac OS9.
>I'll check and let you know if it is different.
>
>
>-- David
>
>
>>i have a question for you. Is the following line in DescribeScreen correct?
>>
>>case 'ATY,RageM3p29s', fprintf('ATI Rage Mobility(?) in iBook\n'); %
>>David Jones G3 iBook/600.
>>
>>I.e. does it correctly describe your iBook's screen? Do you know
>>whether your iBook has a Rage or a Rage "Mobility" chip?


dear david

thanks. the driver is completely different under Classic, and tends
to be quite limited in services. i wouldn't try to extrapolate from
one to the other. however, i've got all the info i need for now.

what's "MacTracker"?

best

denis
2
Hi Denis,

You asked what "MacTracker" is ...

MacTracker is a program that runs under MacOS9 and MacOSX.
It is essentially a nice user interface on top of a database
that provides detailed information on every Apple computer ever made,
including processor speed, memory, graphics cards, etc.

http://members.shaw.ca/mactracker/


-- David
3
Hi Denis,

I know the new OpenOffScreenWindow with VRAM/AGP flags is experimental.

ScreenTest on my iBook in Classic mode
doesn't seem to detect the graphics card is AGP or has 8 MB of VRAM
and reports a CopyWindow speed of 43 MB/s.

ScreenTest in MacOS9 doesn't seem to detect AGP or VRAM
but reports a CopyWindow speed of 95 MB/s.

The factor of 2 difference seems suggestive
but I don't know if it has anything to do with the (AGP 2x).

-- David Jones
4
>Hi Denis,
>
>I know the new OpenOffScreenWindow with VRAM/AGP flags is experimental.
>
>ScreenTest on my iBook in Classic mode
>doesn't seem to detect the graphics card is AGP or has 8 MB of VRAM
>and reports a CopyWindow speed of 43 MB/s.
>
>ScreenTest in MacOS9 doesn't seem to detect AGP or VRAM
>but reports a CopyWindow speed of 95 MB/s.
>
>The factor of 2 difference seems suggestive
>but I don't know if it has anything to do with the (AGP 2x).
>
>-- David Jones
>

dear david

thanks for the interesting ScreenTest results.

1. the absence of AGP and VRAM scores merely indicates that
ScreenTest didn't succeed in getting results. at present it tries to
allocate quite a lot of memory and simply gives up if that doesn't
work. it's likely that your card is simply offering less than
ScreenTest is demanding. Feel free to enhance ScreenTest if this
interests you.

2. the difference in CopyBits speed is dramatic but not unusual.
We've been seeing dramatic differences in CopyBits speed with ATI
cards, and are still trying to track down the causes.
http://psychtoolbox.org/mac.html#drivers

3. XClassic: the setclut timing results under classic are a mess.
perhaps the normally accurate UpTime timing is being simulated,
within the os, by rotten Time Manager timing. our understanding of
setclut timing under Classic is still murky, but all the hard facts
are bad. it is my impression that using SetClut to wait for blanking
under classic will never work. one can wait for the interrupt, but it
is our impression that the interrupt is simulated, autonomous from
the actual time of blanking. the best hope for synching to blanking
under classic would be through OpenGL, which we haven't yet
investigated.

4. OS9: the SetClut timing results for the ATI Rage Mobility chip in
your iBook are quite disappointing. It appears that under Mac OS 9 it
always takes more than 1 frame to load the clut. So you synch at 30
Hz, not 60 Hz. This is lousy. You'll want to set the parameters in
PrepareScreen to wait for the interrupt instead of relying on SetClut:
Screen(screenNumber,'Preference','AskSetClutDriverToWaitForBlanking',0);
Screen(screenNumber,'Preference','SetClutCallsWaitBlanking',1);

5. OS9: However, before you do 4, please run a more directly diagnostic test:
ClutTimeTest

depending on the results, i'll modify PrepareScreen to set up your
card appropriately.

best

denis