Your card is locking up because it's not getting enough volts for your core. Unsync them and have individual overclock profiles per core. I run 950core/1000mem at 1.1625v and get roughly 715 Mhash/s. I do not have the cores synced as one core is crappier than the other. So core crappy is @ 940, while the other one is at 960. I can also run 1.10v with core crappy at 900 and the other at 920. Unsyncing them will definitely help you balance core speeds with voltage.
I haven't tried unsynching the cores yet, I'll give that a shot and play around with it.
One thing to note though, that during testing, (synched of course at the time,) I set voltage to 1.1625 and left the core speed @ 725mhz. Even with 100% fan speed I was approaching the critical temperatures at 725mhz. Would unsynching make a difference then here?
You do not need to run at 1.1625v on the cores for 725mhz. Stock volts would run that. Unsyncing would help with stabilizing your system at a particular voltage, and optimizing core clock for each core, as each core is slightly different and one will almost always be better than the other.
- I would start at 1.1625v @ 100% fan. I use a combination of Sapphire Trixx and MSI Afterburner. Use Trixx to set the voltage and sync the core speed to a core clock you think may work. Test this configuration and make sure m0mchill's client isn't throwing you errors. Scroll up in the prompt to check your logs. This will take several tries if you don't already know the cores’ limit.
- Using GPU-Z, check your VRM temps at your synced core speed and voltage and make sure it’s not throttling you at this speed; throttling kicks in around 120C for the VRM. If it is throttling, lower your clock speed and voltage. Repeat the previous step until it is not throttling you anymore. When it stops throttling, this particular voltage is very close to the most voltage you will be able to use. I let my VRM temps hang out around 110C so that I can account for room temperature fluctuations throughout the day.
- Once you figure out the maximum voltage and a core speed that works for both cores, use afterburner to unsync the cores and raise one of the core speeds up 5-10mhz, check that GPU’s log for errors. If this is the bad core, your system might crash or start throwing you errors in the log. If it is throwing you errors, revert back to the original core speed for this gpu and make a note of which one it was, GPU1 or GPU2. This is the bad core.
Note: Afterburner messes up my fan speed, I don’t know if it will do it to yours, but if it does, you will need to go back into Trixx to fix it.
- If it is not throwing you errors, then most likely this is the better core and you can keep bumping up the core speed in small increments and checking the log for errors. When the first error comes up or your system hangs, move down 5mhz and retry that core.
- Once you’ve determined the max clocks for each GPU @ 1.xxxx volts, recheck your VRM temps as this higher clock rate as a higher clock rate at the same voltage will dump a little more heat.
Good luck with overclocking! Do it at your own risk!