Code:
sudo apt-get install fglrx-updates fglrx-amdcccle-updates fglrx-updates-dev
sudo aticonfig --lsa
sudo aticonfig --adapter=all --initial
sudo reboot
sudo aticonfig --adapter=all --odgt
I start with a .conf file, which I will copy here when I get home and can access it.
Right now, I have my clock rates pretty close to stock, since I wasn't seeing any change in CGMiner (Leading me to believe it's not a tuning issue right now), but my two R9 280X (Not in Crossfire and the second card does not have a dummy card due to the temp suggesting it doesn't need one).
CGMiner displays 1.098MH/s. My pools, currently https://pool.coinyechain.info/ because why the fuck not, show almost 1600KH/s.
I don't think either one is right, since I should be getting anywhere from 1350-1450KH/s at these settings. I don't seem to be getting many stales, which is good. But I can't really tune my settings well without CGMiner being accurate to some degree and without a Kill-A-Watt (This part isn't your problem. I can do this).
I get the feeling this is a common issue, as I've heard tales of CGMiner giving the wrong hashrate before, and surely there's a fix floating around somewhere, right? It's an open-source program. But there's so much junk floating around that I can't find anything valuable.
Usually the blame is shifted on needing an optimized kernel and has a link to a vague discussion of the hypotheticals of writing an optimized kernel, but if anyone says anything like "great! But where is it?" they're yelled at and told to write it themselves, because anyone can compile an OpenCL kernel, amirite?
Other than that, I'm not sure where to even begin to look.
If there's a bounty set up for a fully scrypt-optimized CGMiner, I wouldn't mind chipping in.
On my miner, I have a large, circular fan on top of the milk crate holding my cards and mobo, blowing down.
This works very well, but when I tested the thermal cutoff and overheat marks by turning off the big fan and only using the GPU fans (Set to run at 100% at all times. No autofan), the temperature was marked at 80C to be "overheat" and 90C to shutoff, but both carts exceeded both temperatures and did not shut off.
I imagine this is part of the syntax that prevents it from registering. I'm sure it'll be an obvious once I can get the conf quoted, but my guess is that it should be
"temp-cutoff" : "90,40"
Instead of
"temp-cutoff" : "90"
Where the second value would be the temp at which the card turns back on, but this is speculation, obviously since it doesn't work for me. It might also make Card0 shut off at 90 and Card1 shut off at 40.
I'll have to get back to this when I get home and can get the .conf for reference.
I am willing to tip a fair amount, depending on the scale of the results.