Cliff's:
System in sig performs power-off without warning between heavy load and idle. PSU shows good voltages on multimeter. PSU is Antec TP 480 20-pin into 24-pin M/B. Cause?
Full text:
By "hard shutdown" I mean that the behavior is like turning off the PSU power switch.
By "PSU OK" I mean I tested all rails to be +50mV with a DMM yesterday.
By "medium load" I mean I can run dual Orthos for hours on end without there being any sort of error and I can start SuperPI on top of that and get accurate results, but then the system will perform one of those hard shutdowns at such moments as when I click "OK" in the SuperPI completion dialog.
I'm sure there are some here who know what I'm talking about when I say that I sort of ... feel when the system is going to hard crash, like an anticipation just before it does, but the funny thing is that it appears unrelated to heavy loads.
I figured it was the PSU, an Antec TruePower 480 with a 20 pin connector in a 24-pin slot, but the EZ-Plug molex and the 4-pin P4 connector are compensating. I took digital multimeter measurements last night and I have +50 millivolts on all rails. I measured +5 and +12 on molex, +3.3 on the "odd" connector, so at least undervoltage isn't the problem. Stability still may be, of course. Vcore I didn't measure with the DMM, but even software claims it's between 1.33 and 1.38 when set to 1.35 (stock for 4800+). Speedfan hasn't detected any large voltage fluctuations, and reading with the DMM, I didn't detect any fluctuations at all. +5.05 and +12.05 V rock solid stable throughout measurement.
I don't think it's a memory problem. I have tried both underclocked and stock clocked on the RAM, I have tried 1T command and 2T command, I have tried auto VDimm and 2.8Vdimm and I have still been able to provoke the error.
What I do to provoke it is partially to run dual orthos with several measurement applications running aswell, and then just... fiddle around. I run a few SuperPI measurements on top, which usually LOWERS temps because it distracts the system from full Orthos I suppose, and then all of a sudden, no warning, it will hard shutdown.
Temps at the time are never above 55C, and it can crash at 51C. While overclocking, a core can reach 57C, but still not cause any Orthos or Pi errors. Taking measurements with CoreTemp, I see a temperature differential between the two cores of up to 5C, which seems like one heckuvalot, but I figured it may be a calibration error - AMD state they do not calibrate these at all and up to a 14C error is possible.
What I am considering, a cheap purchase, would be a 20-to-24 pin motherboard power adaptor, but would that really make any kind of difference? It's not as if there's going to be any more current, is it.
Thank you for reading my first thread ever, and thank you all for all the threads that have helped me ever since I first discovered the forum. This submarine is surfacing, and I hope someone will find the will to assist. All suggestions welcome!
System in sig performs power-off without warning between heavy load and idle. PSU shows good voltages on multimeter. PSU is Antec TP 480 20-pin into 24-pin M/B. Cause?
Full text:
By "hard shutdown" I mean that the behavior is like turning off the PSU power switch.
By "PSU OK" I mean I tested all rails to be +50mV with a DMM yesterday.
By "medium load" I mean I can run dual Orthos for hours on end without there being any sort of error and I can start SuperPI on top of that and get accurate results, but then the system will perform one of those hard shutdowns at such moments as when I click "OK" in the SuperPI completion dialog.
I'm sure there are some here who know what I'm talking about when I say that I sort of ... feel when the system is going to hard crash, like an anticipation just before it does, but the funny thing is that it appears unrelated to heavy loads.
I figured it was the PSU, an Antec TruePower 480 with a 20 pin connector in a 24-pin slot, but the EZ-Plug molex and the 4-pin P4 connector are compensating. I took digital multimeter measurements last night and I have +50 millivolts on all rails. I measured +5 and +12 on molex, +3.3 on the "odd" connector, so at least undervoltage isn't the problem. Stability still may be, of course. Vcore I didn't measure with the DMM, but even software claims it's between 1.33 and 1.38 when set to 1.35 (stock for 4800+). Speedfan hasn't detected any large voltage fluctuations, and reading with the DMM, I didn't detect any fluctuations at all. +5.05 and +12.05 V rock solid stable throughout measurement.
I don't think it's a memory problem. I have tried both underclocked and stock clocked on the RAM, I have tried 1T command and 2T command, I have tried auto VDimm and 2.8Vdimm and I have still been able to provoke the error.
What I do to provoke it is partially to run dual orthos with several measurement applications running aswell, and then just... fiddle around. I run a few SuperPI measurements on top, which usually LOWERS temps because it distracts the system from full Orthos I suppose, and then all of a sudden, no warning, it will hard shutdown.
Temps at the time are never above 55C, and it can crash at 51C. While overclocking, a core can reach 57C, but still not cause any Orthos or Pi errors. Taking measurements with CoreTemp, I see a temperature differential between the two cores of up to 5C, which seems like one heckuvalot, but I figured it may be a calibration error - AMD state they do not calibrate these at all and up to a 14C error is possible.
What I am considering, a cheap purchase, would be a 20-to-24 pin motherboard power adaptor, but would that really make any kind of difference? It's not as if there's going to be any more current, is it.
Thank you for reading my first thread ever, and thank you all for all the threads that have helped me ever since I first discovered the forum. This submarine is surfacing, and I hope someone will find the will to assist. All suggestions welcome!
