Re: New sleep code for ATI M6, M7 and M9

[Michel Dänzer <daenzer@debian.org>]

On Mon, 2003-03-24 at 22:27, Jamie Wilkinson wrote:
> This one time, at band camp, Michel Dänzer wrote:
> >On Mon, 2003-03-24 at 09:01, Jamie Wilkinson wrote:
> >> This one time, at band camp, Benjamin Herrenschmidt wrote:
> >> >Can you all try with DRI and PCI GART instead of AGP GART ?
> >>=20
> >> Sure, can you tell me what I need to do to turn on PCI GART?
> >
> >Don't load the agpgart module or use Option "ForcePCIMode". Sleep worked
> >with PCI GART on the old TiBook with an M7.
> 
> Okay, I commented out Option "AGPMode" "4" and added Option "ForcePCIMode"
> "true".
> 
> /var/log/XFree86.0.log shows:
> (**) RADEON(0): Forced into PCI-only mode
> 
> Sleep and wakeup works here, too, glxinfo reports that direct rendering is
> being used.
> 
> What's the difference between using PCI and AGP GART?  Which is "better"?

AGP GART is faster (and as we all know, faster is better :) - in fact,
on newer TiBooks PCI GART is pathetically slow, even much slower than
with DRI disabled (for 2D, 3D might be slightly faster than software
rendering). My guess has been this is somehow related to the L3 cache,
what kind of performance do people get with PCI GART on an iBook with an
M7?


-- 
Earthling Michel Dänzer (MrCooper)/ Debian GNU/Linux (powerpc) developer
XFree86 and DRI project member   /  CS student, Free Software enthusiast