Originally Posted by CloudX
donkrx, you are a smart guy, but I think you are stressing the "hard" FFTs too much. Can you pass 15 mins of each?
I haven't bothered too much with the 1792 but it always seems easier to get through for me than 1344. These FFTs are included in the blend test, so I can either run them alone and find out in 15 minutes or less, or run it in the blend test and find out in 3 hours. I passed 10 hours in version 25.11 (which does not have 1344), added .010v, then switched to version 26.6 and failed at the 1344 FFT 3 hours into the blend test.
So I certainly wasn't going for overkill stability, my singular goal was to get thru 12 hours of blend. Seeing as I failed at the 1344 in the blend test it seemed like it would be helpful to test that FFT on its own.
That being said, these FFTs are indeed really fickle and I'm not sure how useful they are in the grand scheme of things. Being able to get through 12h+ in blend (without the 1344) but not 2 minutes of 1344 on its own, seems a bit suspect...... however people in this club are still passing the version 26 blend with 1344 so I would like to as well.
As it turned out my very last attempt (of changing VTT and PLL) before throwing in the towel worked out. I have now passed 1344 three times in a row for 15 minutes... I havent tried any more because this is enough for me to prove it wasn't a fluke. I'm ready to go for the blend test as soon as the AC is fixed.
One thing to consider is that it might be easier to pass 1344/1792 in the middle of a blend test compared to running it on its own. Maybe the gradual staggered loading in the blend test is easier on the CPU than starting 4 threads synchronously instantly from idle? (after a couple hours of blend testing the individual workers will be a few minutes apart in progress, this could help)