Am I drawing too much power or something?

Atheus

Diamond Member
Jun 7, 2005
7,313
2
0
Okay so this is my brand new machine - 2500k, 8GB, SSDs, etc etc... and a way too expensive 750W Corsair power supply. It seems to have decided to turn itself off at random intervals of about 12-24 hours. Nothing else happens. The surge protector (replacing this tomorrow just in case) does not trigger and continutes to power other things - the screen and the speakers. I simply have to wait until it decides to work again, which maybe several hours later, or the next day... at which point everything works perfectly - voltages and temps are good and it'll run prime95 as long as you want - until it decides to turn itself off...

What the hell is going on? Thanks for any help...
 

ElFenix

Elite Member
Super Moderator
Mar 20, 2000
102,395
8,558
126
i would think memory problem. run the windows memory tester (it's found errors that memtest hasn't) and see what it finds.
 

Atheus

Diamond Member
Jun 7, 2005
7,313
2
0
I haven't overcocked yet but I did set the memory voltages and timings manually to the standards for that kit... Okay good idea I'll run the memory test - as soon as it decides to work again...

Maybe I should clear the CMOS? But you'd think with a memory problem I'd at least get lights and fans for a second or two right? I don't even get that.
 

Atheus

Diamond Member
Jun 7, 2005
7,313
2
0
I'm gonna get a multimeter and see what the 12v reads when the machine won't power on...
 

Atheus

Diamond Member
Jun 7, 2005
7,313
2
0
Extra molex plugged into a spare port on the PSU reads 0.13v on multimeter - still no lights and fans - just the light on the surge protector... the low voltage must be leftovers in the caps... trying another power point elsewhere in the house...
 

Atheus

Diamond Member
Jun 7, 2005
7,313
2
0
Different power socket, no surge protector, no change... :(

Need this machine for work...
 

westom

Senior member
Apr 25, 2009
517
0
71
Extra molex plugged into a spare port on the PSU reads 0.13v on multimeter - still no lights and fans -
Trying a different power point would be wasted time. First understand what you have.

A power controller determines when a CPU can execute and access memory. That power controller has many inputs that also decide to power off the machine and if the machine can power back on. The only way to know what is happening is numbers from that meter. So do this:

Set the meter to 20 VDC. Connect the black probe to the chassis. Find the purple wire that connects PSU at the motherboard. Probe that wire to read something about 5 VDC. Record that number to three digits.

Next touch the green and gray wires. Record each before and 'when the power switch is pressed'. Again, numbers to three digits.

Finally monitor any one red, orange, and yellow wire both 'when the power switch is pressed' and after voltages stabilize. Report those numbers.

The next reply will report on all components of the power 'system' - power controller, PSU, and other parts.

In your case, those same numbers should be taken both when the machine does boot and when it does not. The reply will contain knowledge and facts you have never heard before. Without numbers, your only replies can only be wild speculation. First learn what the power controller is doing before considering anything else.
 
Last edited:

Atheus

Diamond Member
Jun 7, 2005
7,313
2
0
Westom thanks for the detailed reply - I guess you're an EE? I only ever took one EE class and it was all about ICs... guess I've got some reading to do here...

After seeing you both mention memory I reset the CMOS to default, and it did turn on, but only lasted a couple of hours. I ran the Win 7 memory diagnostic during that time. No errors. I'll try to take the VDC measurements you suggest when I get a chance. I'll have took up the pinouts - Corsair seems to think it's a good idea to make all the wires black so you don't know what you're doing :/
 

westom

Senior member
Apr 25, 2009
517
0
71
I'll try to take the VDC measurements you suggest when I get a chance. I'll have took up the pinouts
ATX standard requires all wires to be standard colors. A power supply is not ATX standard if color standards are violated. What else did they short the supply of to increase profits? Since the supply would not have one of the most important requirements in its numeric specs - conforms to ATX standards. Most any power supply, missing essential functions required by the ATX standards, can still boot and run a computer. Very profitable when so many buy on dollars, watts, or a name. And ignore what is only important – spec numbers.

Apparently you need this:
http://www.smpspowersupply.com/connectors-pinouts.html
 

stahlhart

Super Moderator Graphics Cards
Dec 21, 2010
4,273
77
91
Which Corsair 750W supply? I've got the HX, and the ATX and EPS lines for the motherboard are color-coded; they're just bundle-sleeved in something like Techflex. The modular lines for the peripherals are insulated all black, though.
 

Atheus

Diamond Member
Jun 7, 2005
7,313
2
0
@stahlhart - the AX750 'Gold'- it was well reviewed. All the lines are black.

I haven't taken those VDC measurements yet; I decided to try some other (easier) stuff first, including a BIOS flash, since which it has not turned itself off... however it did screw up a memtest... I ran it for 24 hours while I was out for a night, and when I came back it was a mess of randmon unicode characters :/ - the #passes and #errors was covered up by the mess.

That's a step in the right direction though, and I'm now not sure this is a PSU problem, so I'm starting a new thread here...

http://forums.anandtech.com/showthread.php?p=32861511#post32861511
 

westom

Senior member
Apr 25, 2009
517
0
71
... and I'm now not sure this is a PSU problem, so I'm starting a new thread here...
A marginal power system can cause anything else to behave strangely. If the foundation under a house is collapsing, do you blame humidity for making doors stick? Until the foundation of your computer is known, then all other testing may only be chasing ghosts or fixing things that are not defective - ie memory.

MemTst is useful when a pattern is observed in its error messages. Strange and random errors suggest failure elsewhere. Just because you saw an error message does not say where a problem might exist. First learn what the numbers are saying.

From information in both posts, you have no reason to make those conclusions. Most of what you have posted (including the latest information) all implies a power system problem as a most likely suspect.
 
Last edited:

Atheus

Diamond Member
Jun 7, 2005
7,313
2
0
Okay I'm going to take the measurements now and try to rule out the PSU and mobo if I can - quick question though - are you asking me to probe these wires while everything is connected? How do you recommend I do this? Connect a paperclip to the meter probe and stick it down the header? Cut some insulation off the wires?