Bug#583653: radeon.agpmode=-1

[Arno Schuring <aelschuring@hotmail.com>]

I can confirm that switching to PCI mode alleviates the slowdowns for me
as well, on a Radeon 9550 (r300). But the overall 2d performance (as
measured with gtkperf) is still 50% lower than without KMS: 18 seconds,
as opposed to 12 seconds without KMS. Interestingly, the gtkperf tests
take 19 seconds with agpmode=-1.

Could anyone elaborate on the slowness, like what is causing it and why
PCI mode helps? Is this issue tracked in an upstream bug report, or is
someone actively trying to fix it?


Thanks,
Arno