Originally posted by jsolomon
This is probably a repetitive question, but I haven't otherwise been able to find the answer:
How much are the X11 calls like XDrawRectangles() and XCopyArea() accelerated with nvidia's linux drivers? I can see that when I run x11perf and compare results between the native display and a Xvnc server the native display is about 4x faster, but is this all due to hardware acceleration?
If hardware acceleration is in effect, is this due to nvidia's work or is it just a part of XFree86? On comparable graphics cards, how would nvidia's driver compare to the Summit Xserver from Xig? (Again, I only care about X11 calls. I don't give a crap about OpenGL for the the purposes of this question.)
Finally, to what extent is hardware acceleration done? Are pixmaps treated like textures and kept on the card? Is rasterization for X11 rectangles and circles done on the card?
Do you only get acceleration with workstation cards like Quadro, or do you get it with any nvidia card?
I don't have any technical details to give you, but I can report that I get 2D graphical acceleration from a Geforce2MX, so it's not just Quadros.
Quadros definitely have more driver options - they get nice things like Unified Back Buffer - but that's probably a hardware feature imho.