Stress Testing (Haswell) and Getting Crashes Rather than Computation Errors?

[flexy]

My usual way of testing overclocks is OCCT or Prime95.

For some time, a "failure" in overclocking for me is not so much that I get a computation error that OCCT or Prime would say the result is wrong but the PC just crashes and restarts.

Now I am slowly and gradually wondering whether this could be a PSU/Power issue?

I am using a very old Toughpower 750W which otherwise is excellent and never had a problem... I read somewhere that Haswell might need 25A-30A rails, and this PSU is a 4-rail PSU with 18A on each rail (60A total on the 12V). Would sudden crashing/restarting be indicative that the PSU is a problem here?

 

[BonzaiDuck]

Actually, there's a CHANCE it could be the PSU, but your narrative might suggest that you expect the stress-test software to simply end the test and "tell you" of an error.

Back in the day when us old-timers were over-clocking Sandy Bridge -- way, way back in 2011 -- I'd asked that question myself.

You should expect BSODs and resets. Even in my most recent experience, OCCT -- to a lesser extent Prime95 -- will report test-termination or show a thread stopped only when you're very close to a stable clock setting.

If you get BSODs, the stop-code errors can tell you what to do next; they will be due to either a memory/memory-controller failure (09C), the CPU (101), or a combination ("general hardware") with code 124. There is a list available matching some ten stop-codes with their indications.

If you only get resets, that's a nut a little harder to crack. And it could still be the PSU. Do you have a PSU tester?

 

[Burpo]

You didn't say what board or CPU.. or if it's overclocked..

 

[BonzaiDuck]

You didn't say what board or CPU.. or if it's overclocked..

And . . . "+1" to that . . too . . .

 

[Puffnstuff]

Well the fact that this is happening during an OC might tell us something about the cause.

 

[DrMrLordX]

This topic sort of brings up the old multiple rails vs. single big rail debate (PC P&C anyone?). I don't see how your PSU's rail layout would matter if you've got an 8-pin ATX or EPS 12V connection to the motherboard. You should be delivering power to the board from (at least) two separate rails, and the EPS 12V spec allows up to 28a on its own.

If there is a problem with your PSU, it probably isn't due to the rail layout.