On Tue, 14 Jan 2003, Tim Roberts wrote:
> On Tue, 14 Jan 2003 07:51:10 -0500 (EST), Vladimir Dergachev wrote:
> >
> >One thing I don't understand about this is that this is an integrated
> >graphics chipset, right ? I.e. it uses part of system RAM for video RAM.
> >There is no code in radeon driver to deal with this yet.
>
> And there doesn't need to be. The allocation of video RAM from system RAM is
> done by the BIOS very early in boot time. The physical address of the video
> RAM gets stuffed into the PCI BAR register. The driver doesn't know or care
> whether that address is physically on an AGP bus, or in system RAM.
>
> The Savage driver is an example. The Savage/IX uses shared memory, whereas the
> Savage/MX has its own separate VRAM. The driver doesn't care at all.
Thanks for the explanation and example. It is possible that Radeon driver
messes with PCI and/or memory controller registers (GATOS driver certainly
does). It could be that PLL block is slightly different as well (ATI
webpage mentions something about integrated master clock).
>
> Interesting side note: I had a fellow this week report to me that his disk
> performance goes down by HALF when he switches from 1024x768 60Hz depth 8 to
> 1280x1024 85Hz depth 24. He has a shared memory system, and the simple act of
> refreshing the screen at the higher resolution puts a constant 450MB/sec load
> on his processor bus. Depending on the speed of your front-side bus, that
> could be from 1/3 to 3/4 of your north bridge throughput.
Yep, a good reason to stay from integrated graphics.. unless perfomance is
not an issue.
best
Vladimir Dergachev
>
> --
> - Tim Roberts, [EMAIL PROTECTED]
> Providenza & Boekelheide, Inc.
>
>
> _______________________________________________
> Devel mailing list
> [EMAIL PROTECTED]
> http://XFree86.Org/mailman/listinfo/devel
>
_______________________________________________
Devel mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/devel
