Decided to see if its worth overclocking some cards since it seems risky buying new GPUs.
I got a high quality ASUS 280X DC2T and decided how far I can take it. Had to un-undervolt it back to the stock bios.
At 1230/1800
@1.3V and +20 I got 24MH/s.
Will play around and try to lower the voltage as low as possible while it remains stable.
Comments
The VRM temps kept spiking over 100C. It ran very hot even when undervolted and underclocked.
Gigabyte has crap VRM also.
If you have any ASUS or MSI, try overclocking those.
Only for benching, not for mining.
Anyway, all my rigs are finally running stable. Haven't had any problems for days now. I think I shouldn't create more sleepless nights for myself. If it works, leave it alone
Will probably try and get it runing at 23.5Mh/s since at that speed the core only needs to be 1175 and requires less volts.
1. Higher watts - higher temps - higher chance to burn out VRM on your card.
2. Higher watts - less cards per one PSU.
Need to find balance. 1230 on GPU is too much for Tahiti.
Slow down to 1150-1180, some cards can run it on 1.1-1.12 Vgpu
But I changed the clock down to 1175 with the same memory clock, lets me run it at 1.2V and uses like 30 watt less per GPU.
I successfully replaced a line of them on VRM, when they start to burn. Card works again.
It was only one case of VRM damage on hundreds of mining Gigabyte 7970's that I supervise.
Will try undervolting next to get power down.
By the way, my config is also MURDERING X11 at 13.333MH/s !!!
ethminer -M -G --benchmark-warmup 20
( I think you're going to kick ass BTW)
Went back to ether, easier to mine and consume less power.
first change "DisableAMDTempControl": from false to true
then
add --gpu-fan to "ExtraLaunchParameters"
Seem like my normal random pick 280 in bunch outperform your good 280x one .
Now stable run testing .