Originally Posted by NvFuchs
I don't have KDE4 at hand and can't be arsed to emerge the
whole thing, but when I tried it on a live system (!) it was
fast enough to work with, even though the system ran
from RAM / CD-R.
If tried it live then you were probably using either vesa or nv, right?
well it should be faster with nv (dunno about vesa) than it is with nvidia. The fact that it is faster under xephyr or xnest points to problem with acceleration. Mind you I only tried this with kde4.0.x and not with trunk (which is much improved in some ways, but not in others).
In terms of speed I'm merely referring to Qt painting performance, overall kde4 is much faster esp. startup times.
I noticed that you have pixmap placement at 1 and glyph cache enabled. I thought that pixmaps had to be created on gfx mem for the glyph cache to work? how does it compare to ipp at 2.
P.S. I noticed that your vesa scores are quite similar to your accelerated scores. I suggest you try xrenderbenchmark
it uses qt4 to test xrender. Obviously it is a synthetic benchmark but I haven't seen any results of mobile and non beefy gfx cards.
I would be really interested in your scores.
P.P.S. It can take a long time to complete.