My friend's R9 280X can have 3100/3500 (default/OC) cpm on clpts-v0.2.2, and my gtx690 has 2200 cpm in total. I believe this means AMD now is much faster in SHA 512, single chip beats my two. clpts also asks for 2 threads for each chip like your miner, apparently, it is much more efficient and optimized.
Check this post https://bitsharestalk.org/index.php?topic=2598.0
Thanks for the information you provided. I don't have a R9 280x just now, on my HD7850 2GB, clpts-v0.2.2 is ~15% faster displayed, ~10% faster net than PtsGPUz v0.4c, which net performance is 840 cpm.
===================
Preliminary test result, only on HD6670 1GB:
AMD HD6670 1GB, test using AMD Catalyst 13.12, Windows 8 x64, G1610 CPU:
1. ominer 0.8, 425 cpm, fixed to upcpu, pool fee 2%, net performance user gets:
417 cpm;
2. PtsGPUz 0.4c, 380 cpm, pool fee 3% at ptspool.com, net performance user gets:
368 cpm;
3. ominer 0.9 preview 2, 336 cpm, fixed to upcpu, pool fee 2%, net performance user gets:
329 cpm;
4. clpts 0.2.2, 218 cpm, fixed to ypool.net, miner fee 5%, pool fee 5%, net performance user gets:
197 cpm.
Note: clpts 0.2.2 seem to be the fastest on AMD cards with 2GB or more video memory, but slowest on cards with 1GB memory.