max performance of pci-e gf6600gt?

i am rendering hundreds of thousands of lines. i must draw every fragment, so i can’t use depthtesting (or any other testing :slight_smile: ). i am using blending for accumulating the overdraw.

i am unable to improve performnace, there is no difference between IM vs. VA vs. VBO (with VA and VBO i am using simple vertex shaders, no precomputing on CPU). fragments shaders are also simple (one liners, just setting the color).

with gdebugger and instrumented driver i found out, that i am rasterization bound. the app is 99% of the time waiting for ROP.

now i am achieving 220-250 millions of fragments per second on a pci-e gf6600gt 128MB, a64 3000+, 2GB ram, win xp sp2, with 80 series drivers. is this the max performance of this card? or could it be better?

http://techreport.com/reviews/2004q3/geforce-6600gt/index.x?pg=3
http://www.extremetech.com/article2/0,1697,1643480,00.asp

those pages are contradictory, one says 2Gpixels/sec and other 4Gpixels/sec.

that is theoretical peak, which says, that my app could be 10 to 20 times faster. hahaha. how to achieve that???

Consumer cards have typically less performance with lines.
Do you use smooth lines ? multisampling ? more-than-1 width ? All these can slow you down.

no AA, no smooth lines, width just 1.

i tried rendering lines as degenerated quads, but performance was the same, but visual quality was bad