Ideally, I'd like to squeeze out as much performance as I can from this card, since the system is otherwise unused (file server, used sporadically).
I've also got my windows machine running 2x6850's (not crossfire) with cgminer, averaging about 170MH/s per card - which also seems quite low for these cards. I've got these cards configured with the following settings:
"vectors" : "2,2",
"worksize" : "128,128",
"lookup-gap" : "0,2",
"thread-concurrency" : "7680,4096",
"shaders" : "960,960",
"gpu-fan" : "10-98,10-98",
"temp-cutoff" : "95",
"temp-overheat" : "85,90",
"temp-target" : "72,72",
"api-port" : "4028",
"auto-fan" : true,
"auto-gpu" : true,
"expiry" : "120",
"gpu-dyninterval" : "4",
"gpu-threads" : "1",
"log" : "5",
"queue" : "0",
"scan-time" : "60",
"temp-hysteresis" : "3",
"shares" : "0",
I've tried setting the intensity to anything other than dynamic, and the system lags terribly. Since I use this system, dynamic is really the only choice I have for these cards. Setting gpu-threads to anything other than 1 seems to also halve the hash rate. The other issue is that I cannot seem to generate a new kernel, possibly due to some unknown problem with the 13.3 beta catalyst drivers and the included AMD APP sdk - I've tried all the known workarounds to solving this problem, aside from reinstalling Windows, but I cannot seem to get any version of cgminer to cooperate with this system. As such, the kernel I'm stuck using was compiled when I installed cgminer originally (v2.11.3, currently inserted into the 2.11.4 installation, with the above config).
I'm confuzzled by all this. Is there something I'm really missing regarding the Mac side of things?
Is there a way to fix my openCL support on my Windows box - Is this the only issue preventing my cards from hashing better (aside from the intensity, that is)?
HELP!
PS: I'm a computer tech by trade, I'm somewhat versed in shell scripting on *nix and Windows, understand how to compile stuff on *nix and Windows; In any case, any help would be great