perplexing FAH hard locks

[rommelrommel]

Been trying to get a second computer worth folding on and am using:

Q9450 w/TRUE120
evga 780i mobo, using NB fan, have remounted NB/SB cooler
corsair hx520 ps, verified rails to be well within spec
4gb OCZ pc-8500 DDR2

I can get the computer stable at 3.4 without too much trouble, temps are alright, will run prime blend for days. Previous to getting the 9450 I had a E2140 @ 3.0 completely stable.

But, FAH won't remain stable at anything more than 2.8 regardless of any settings. It hard locks the computer, and the logs just show the last completed frame, never any errors or shutdown messages. It usually locks within 2-12 hours.

I have some experience with the 780i, 425 FSB is completely stable, the memory is memtest86 stable, the power supply rails are well within spec, I'm just not getting it.

Anyone else had a similar problem? I'm almost ready to sell it all and shell out the extra bucks to build another 2500k/2600k rig.

[chrisk]

Is there safely room for additional volts for the cpu/ram/NB?

[Rison]

Welcome to folding stable. There is benchmark stable and folding stable. Two entirely different beasts for some reason.

Check your voltage on the CPU as ChrisK mentioned, but check your memory speeds and volts also, perhaps something is going sideways.

Use CPUZ to see what your clock is at (if using windows) and CPUZ has a system temp program too, make sure it's not overheating on you.

[stoanee]

Originally Posted by Rison

Welcome to folding stable. There is benchmark stable and folding stable. Two entirely different beasts for some reason.

Interesting that you mention that. When I overclock I always use folding as my primary stress test.

[supaflyx3]

I've had WU's fail multiple times on both GPU & SMP folding due to not being FAH stable but fine with prime/furmark.
Try increasing your voltages a smidge and see how that works.

[Tim_H]

I'd agree with what everyone is saying so far, I've had this issue before a few times. I had to do some combination of lower the overclock and up the voltage a smidge. I started with the voltage, if that doesn't work I lower the clock. Folding isn't meant for it but it's a great stress tool!! If I remember some people have "trapped" wu's for this purpose, but I don't remember how it's done.

[rommelrommel]

Yeah, I had thought it was memory at first because it was undervolted, fixed up that and the timings and it didn't help, I already was running the memory a lot slower than it was spec'd for.

I've been running CPU-Z and realtemp the whole time, and nothing is happening that way, I did the vdroop mod so things are pretty stable voltage wise. I've taken all the voltages to the limit of what I consider safe with no change... like the computer is stable around 1.25 vcore but even a test run at 1.4 vcore which is pushing it on 45nm didn't give stability. I may have just found the limit at 2.8ghz but sucks to leave another 10-20% on the table.

I used to have a little farm in the tbred b/barton overclocking days and I have had to back off a little for folding stable but never nearly this much.

[Tim_H]

NB voltage?

[rommelrommel]

Everywhere from maxed to stock, no change.

[chrisk]

Strange. Have you tried memtest for 2-3 passes? Might be a bad section of your ram that prime is not hitting, but f@h is.