I recently got an HD 5970 and had previously been using puddinpop's miner with a Nvidia so thought I'd stick with it because it's neat and small. I had problems getting full performance and had started another thread and got some helpful suggestions that pointed me here, I was using -gpu=0 and -gpu=1 to start two difference instances as in the readme.txt but for some reason it seemed to be ignoring them, so my second miner would halve the speed of the first. Then when I installed your software things were sweet and I was getting double the performance and about what I expected
.
The joy was short lived though because I quickly got the BSOD on subsequent occasions after it was running OK for a while. I know there are a 101 driver/hardware problems that could cause it but thought after playing around further I'd post my results to see if you could think why it might be happening, it seems to be related to how the process is launched:
If I use your application set for auto-start I get a freeze and/or BSOD straight away, I did that after originally manually starting it while setting up when all looked OK. Anyway doing it manually and leaving a delay between worked OK so that's a problem elsewhere no doubt.
However even though I got it started that way it wouldn't keep running long, which I'd put down to a motherboard / temperature issue but I thought to use my original command line file to start a miner on GPU 0, and your program to start it on GPU1 it all ran fine for 12 hours. I wanted to stop it for a while, closed the command line OK but pressing stop on yours froze a little while then gave the BSOD.
I noticed that poclbm.exe was using 25% CPU all the time on a quad core, I didn't know if that was normal or maybe related in some way.
In general while tweaking things I can stop and start the command line version with no ill effects, but any attempt to stop the instance created by your program causes grief. For the command-line version I've just been pressing the close "X". With yours if I start a single instance and stop it causes problems 100% of the time which seems odd, maybe the process is killed in a slightly different way?
I couldn't seem to spot the source code for this beta version, I was going to take a look at what command-line parameters were pased by default to launch the miner and I thought that might give me some clues as to possible reasons for the problems I was having. I thought if I could start them the same way that would isolate if it's something to do with the resources allocated or something else.