Wont boot, blinking power LED

matrim

Distinguished
Sep 16, 2010
5
0
18,510
I recently resuscitated my old computer who's graphics card had died. I installed windows 7 32 bit and it was working great for a couple weeks. Today, i was playing Mass Effect 2 in 3D and the computer locked up. I pressed the reset button and it booted back to windows, but it locked up again and this time it would not boot up. Now, when I press the power button, the fans spin up, the HDD spins up and all the lights turn on, but nothing comes up on the screen, but the power LED flashes rapidly. There are no beeping sounds or anything of that nature either.

My build is as follows:
DFI lanparty UT nf4 Ultra-D
EVGA geforce 9600 GSO
Enermax 500 watt power supply
socket 939 3200+ winchester core
2 2 gigs Mushkin (I know I can't use it all in 32 bit, but I was planning on going back to 64 bit.)

Please help, I have no idea whats wrong.
 
Work through our standard checklist and troubleshooting thread:
http://www.tomshardware.com/forum/261145-31-read-posting-boot-problems
I mean work through, not just read over it. We spent a lot of time on this. It should find most of the problems.

If not, continue.

I have tested the following beep patterns on Gigabyte, eVGA, and ECS motherboards. Other BIOS' may be different.

Breadboard - that will help isolate any kind of case problem you might have.
http://www.tomshardware.com/forum/262730-31-breadboarding

Breadboard with just motherboard, CPU & HSF, case speaker, and PSU. You do have a case speaker installed, right? If not, you really, really need one. If your case or motherboard didn't come with a system speaker, you can buy one here:
http://www.cwc-group.com/casp.html

Make sure you plug the CPU power cable in. The system will not boot without it.

I always breadboard a new build. It takes only a few minutes, and you know you are putting good parts in the case.

Resurrecting an old, dead computer pretty much qualifies as a new build. In this case, it ran a little while then, for some reason, died.

You can turn on the PC by momentarily shorting the two pins that the case power switch goes to. You should hear a series of long, single beeps indicating memory problems. Silence indicates a problem with (in most likely order) the PSU, motherboard, or CPU. Remember, at this time, you do not have a graphics card installed so the load on your PSU will be reduced.

If no beeps:
Running fans and drives and motherboard LED's do not necessarily indicate a good PSU. In the absence of a single short beep, they also do not indicate that the system is booting.

At this point, you can sort of check the PSU. Try to borrow a known good PSU of around 550 - 600 watts. That will power just about any system with a single GPU. If you cannot do that, use a DMM to measure the voltages. Measure between the colored wires and either chassis ground or the black wires. Yellow wires should be 12 volts. Red wires: +5 volts, orange wires: +3.3 volts, blue wire : -12 volts, violet wire: 5 volts always on. Tolerances are +/- 5% except for the -12 volts which is +/- 10%.

The gray wire is really important. It should go from 0 to +5 volts when you turn the PSU on with the case switch. CPU needs this signal to boot.

You can turn on the PSU by completely disconnecting the PSU and using a paperclip or jumper wire to short the green wire to one of the neighboring black wires.
http://www.youtube.com/watch?v=5FWXgQSokF4&feature=youtube_gdata

This checks the PSU under no load conditions, so it is not completely reliable. But if it can not pass this, it is dead. Then repeat the checks with the PSU plugged into the computer to put a load on the PSU.

If the system beeps:
If it looks like the PSU is good, install a memory stick. Boot. Beep pattern should change to one long and several short beeps indicating a missing graphics card.

Silence or long single beeps indicate a problem with the memory.

Insert the video card and connect any necessary PCIe power connectors. Boot. At this point, the system should POST successfully (a single short beep). Notice that you do not need keyboard, mouse, monitor, or drives to successfully POST.

Now start connecting the rest of the devices starting with the monitor, then keyboard and mouse, then the rest of the devices, testing after each step. It's possible that you can pass the POST with a defective video card. The POST routines can only check the video interface. It cannot check the internal parts of the video card.
 

matrim

Distinguished
Sep 16, 2010
5
0
18,510
Thanks a lot.
I took one stick of ram out and it booted up, which is weird because I took the same stick out a couple days ago and when I tried to boot after that it wouldnt. Anyways, just for shits and giggles, I tried putting the stick back in and it still booted fine. I ran mem test right before I put this computer together and both sticks passed. I'm running it again but am rather perplexed as to whats actually wrong. Maybe bad connection? Faulty ram? I guess we'll see when memtest finishes.
 
I have seen this in other systems where it has been a symptom of the bad capacitor problem on the motherboard. Google "bad capacitors" to get some images of what they look like. Usually they will be bulging, often with leaked goo around them. If you are skillful with a soldering iron, you may be able to "re-cap" your board. Considering its age, however, you would be better served by rebuilding it. If your budget is tight, take a look at the $400 SBM for a good starting point, but use a better PSU. The 500W Enermax you have is probably re-usable, unless it is really old.
 
Another possible culprit is the PSU, with bad caps there another possible cause too. A S939 system is very easily at the end of its useful life by now. Those nForce chipsets ran hot, which isn't a happy place for electronics.
Considering that it's working now, essentially after what you did was to re-seat your RAM, other connectors may need re-seating also.
IS there a speaker attached to your mobo? Without it, you would not hear the beeps; he's got a link in his post where you can get one cheap.
 

harleyspawn

Distinguished
Nov 16, 2011
3
1
18,510


I've run into this quite often while repairing computers. This can happen when the contacts on your ram and it's socket get a little faded. An easy way to confirm this is to boot to your bios put it on a page where you can see the clock tick by, and then start rocking your memory sticks back and forth in their sockets. Either your clock will stop ticking or the video signal goes all glitchy (if you're running on-board video) when you hit a memory stick with dirty contacts. If you can vigorously rock your memory sticks back and forth and the computer does not come to a stop, then you may have bad ram. Memtest doesn't always detect bad ram, but it's mainly good at detecting ram that will fail at high loads, or ram that is susceptible to overheating.

If you rock your ram back and forth and the screen does glitch out, or the clock stops, then what you want to do is take that ram out, and run a pencil eraser over the contacts until they're all shiny. Then re-insert them into your computer, rock the hell out of them before turning it back on, and then boot back to the bios. This time, when you rock your ram back and forth, your computer should continue to run. If not, just keep rocking that ram back and forth - what this does is cleans the contacts, as well as the memory socket's pins. it wears a little groove into the copper on the ram, and cleans the pins of the memory socket.

This little fix is something I use all the time at my computer shop. When I get a PC that boots up "sometimes," or has random page faults even though the RAM tests good, this has fixed the problems 95% of the time. One of my trusty tools is now a Pink Pet eraser, and it's saved my customers $100s.