I am testing lolMiner 1.22 (linux) with 40 GPU's (all 480/580/570 4G)
Resuts:
1) All speeds are always better than TRM
2) There are no rule to get better hashrates some cards go to 15MHz, some to 13MHZ, one goes to 25Mhz and other identical card on identical Mobo goes to 13MHz...
I think it will be nice to have a special instructions detailed for us to get the better hash on 4G GPU's.
PS: Enable 3Gen PCI not always meaning more hash on my tests
One of the important things is the PCIx8 or PCIx16... so that is also a bottleneck, lot of cards will give also a bottleneck if all the PCILanes are used... These is the reason why sometimes only 1 card is giving the max perfomance too...
The autotune made by the developer is great, but normally is hardware bottleneck.