neat trick: Powering up my computer with NO CPU HEATSINK

roid · Post by **roid** » Sun Jun 25, 2006 8:19 am

ok, today i've been trying to diagnose my pc's booting problems. it has been getting increasingly temperamental when booting up: generally it would boot up fine, but sometimes it would give memory or video beep error codes (fixed by giving the box a wack, and REbooting - yes the wack DID fix it). But eventually it just stopped booting at all - not even a POST code. Which is where it is now, i turn it on, all the fans startup, the power LED is on, the HDDs even spinup... but nothing happens - the screen does not activate and i don't even get a POST beep code.

Anyway after doing some reading i think my Motherboard is stuffed, more specificly i think i have a hairline fracture somewhere between the Power lead connector and the CPU's frequency timer - so that the \"Power_Good\" signal from the PSU (which i checked at the plug, the signal is strong) is not being recieved by the CPU timer, therefore the CPU timer keeps the reset button pushed down on the CPU. Basically it's the computer's inbuilt defence mechanism against bad power. The timer circuit does not allow the CPU to go through ANY clock cycles until it is assured that the PSU is supplying a stable high quality source of power. My PSU is fine (i have 3).

I suspect a hairline fracture, as the machine would especially play up when the weather changed (ie: thermal expansion/contracting), and also if i had twisted the case in some way like to remove or replace the case cover. The \"bonks on the head\" i would give to the machine to get it working again (which didn't always work first time either) were likely bringing the fractured segments together - and the increasing heat from the computer running was enough to thermally expand them to keep them firmly pressed together while it ran.

Ok, here's the neat trick part:
to test the theory that the CPU was not going through any cycles, i removed both it's fan and heatsink - so the die is bare. Then i turn on the computer, my finger on the CPU die. The fans and LEDs are going, the motherboard has power, yet the CPU die remains cool to the touch and therefore (afaik) completely undamaged by the process.

and that's my trick.

oh, since i'm posting here i may as well ask if anyone has had any experience with such things, and/or can spare any advice either way?

SuperSheep · Post by **SuperSheep** » Sun Jun 25, 2006 10:53 am

There could be a plethora of reasons your computer is not working properly. There are many failsafes built into any modern CPU that can inhibit its operation. You could have a PSU that is out of whack, memory thats corrupted, components, cards not properly seated or bad. You could have short from your motherboard to ground (the case). Generally speaking, PCB's (motherboard) are generally the last thing that is suspect anytime electronics fail. In almost all cases a component fails causing weird failures that generally look or act like something else.

fliptw · Post by **fliptw** » Sun Jun 25, 2006 2:42 pm

bad memory most likely. if there was an issue with the power supply end, you wouldn't get any lights or spinning fans at all.

Sir Sam II · Post by **Sir Sam II** » Mon Jun 26, 2006 8:46 am

I had a bad power supply that would restart the computer randomly, cause it to do reboot loops, & sometimes I had to bang it for it to work at all. Needless to say lights worked fine, & sometimes even the PSU would act just fine, alas replacing it fixed problem. Really it has to be a process of elimination in hardware related issues. Take everything out except whats required obviously mobo, psu, cpu, ram, & if not onboard video, a video card. Work with just those until problem is fixed, and save mobo for last.

Krom · Post by **Krom** » Mon Jun 26, 2006 9:36 am

I have seen a couple bad power supplies here that could power up and run a computer for 20-30 minutes, and then fry the motherboard.

Immortal Lobster · Post by **Immortal Lobster** » Mon Jun 26, 2006 10:21 am

there could be bad capacitors in the voltage regulation unit, so the CPU would not get clean power, and therefore shut down, much as exhibited. look for any bad capacitors in that cluster near the CPU.

You can also underclock an opteron to around 1ghz and a voltage of around 1.15 or so, it will run without a heatsink as well, except...it will actually do stuff

roid · Post by **roid** » Wed Jun 28, 2006 7:47 am

SuperSheep wrote:There could be a plethora of reasons your computer is not working properly. There are many failsafes built into any modern CPU that can inhibit its operation. You could have a PSU that is out of whack, memory thats corrupted, components, cards not properly seated or bad. You could have short from your motherboard to ground (the case). Generally speaking, PCB's (motherboard) are generally the last thing that is suspect anytime electronics fail. In almost all cases a component fails causing weird failures that generally look or act like something else.

fliptw wrote:bad memory most likely. if there was an issue with the power supply end, you wouldn't get any lights or spinning fans at all.

Once the PSU outputs it's "power_good" signal down the grey wire, the CPU timer (part of the mobo) begins outputting a frequency to the CPU which until now has not had any frequency. The first thing a CPU does once it gets it's clock cycle is goto the BIOS and begin to run the POST programs one by one.

My CPU is not going through clock cycles, it is remaining absolutely cool to the touch. AFAIK IT IS NOT EVEN BEGINNING THE POST CYCLE. Memory checks are part of the POST.
A computer should boot with no memory, giving a "★■◆●! no memory!" error code. Oui?

Sir Sam II wrote:...Really it has to be a process of elimination in hardware related issues. Take everything out except whats required obviously mobo, psu, cpu, ram, & if not onboard video, a video card. Work with just those until problem is fixed, and save mobo for last.

You cannot eliminate any more than currently is:
no memory
no videocard, or ANY cards
no hdds/fdd/cdroms
no CASE (it's on the bench)

it's just a motherboard with a CPU on it, hooked up to a PSU. It is also hooked up to a power switch (so it can be turned on), power LED, and a PC speaker.

It should be giving me a beep code telling me that i have a memory or video error (all part of the POST program). The POST will not run unless he CPU is running, as it's the CPU which runs POST programs from BIOS (even POST programs don't run themselves, they need a processor).
If the CPU is running, but has fallen over (crashed) or something then it will still be hot to the touch. Even idle and crashed CPUs are hot.

The fact that the CPU remains cold leads me to believe that the problem is with the "power_good" signal inside teh motherbaord (as i have tested the signal outside of the motherboard - it is fine). Or the CPU timer is busted and is receiving the "power_good" signal but for some reason still isn't supplying a frequency to the CPU. The CPU timer is located on the motherboard.

Immortal Lobster · Post by **Immortal Lobster** » Wed Jun 28, 2006 8:56 am

roid wrote: The fact that the CPU remains cold leads me to believe that the problem is with the "power_good" signal inside teh motherbaord (as i have tested the signal outside of the motherboard - it is fine). Or the CPU timer is busted and is receiving the "power_good" signal but for some reason still isn't supplying a frequency to the CPU. The CPU timer is located on the motherboard.

Immortal Lobster wrote: there could be bad capacitors in the voltage regulation unit, so the CPU would not get clean power, and therefore shut down, much as exhibited. look for any bad capacitors in that cluster near the CPU.

I Bet its the capacitors, becuase if bad, the capacitors will not supply clean power to the CPU, or theyll supply dirty power.

one thing I can be certain of, a crack in the PCB wont do it, and is very unlikely. PCB is actually rather flexible

roid · Post by **roid** » Thu Jun 29, 2006 9:10 am

ahh yes, i see your point. now that you mention it one of the capacitors near the cpu DOES have a deep scratch (even a gouge) on it - but i never thought more about it.

Krom · Post by **Krom** » Thu Jun 29, 2006 9:23 am

I wouldn't say the PCB is \"very flexible\" but I have done my fair share of installing dimms in new motherboards, they do flex enough that it shouldnt be something to worry about.

Duper · Post by **Duper** » Thu Jun 29, 2006 10:26 am

Roid,

a barrel cap or an smt cap?

a barrel cap won't mind a scratch on the outside. they are cased. but you might have a cracked solder through the lead hole. Or the lead itself might not be connected properly in the cap itself. This is rather rare, but the fact that wacking the computer (that just sounds wrong) helped, that could be a solution.

If it's a smt cap, your screwed. Go get another MB.

yea, i know smt caps can be replaced (I do it all the time here at work) but good luck determining the value of the component. Even with a fluke meter.

Immortal Lobster · Post by **Immortal Lobster** » Thu Jun 29, 2006 12:21 pm

I say flexable in relation to other PC parts,

Roid, whats your motherboard, make and model?

roid · Post by **roid** » Thu Jun 29, 2006 7:15 pm

it's an Abit KT7. It's a scratched barrel cap.
i just did some reading and found out that the caps on these boards are commonly faulty :-/

think i'll upgrade anyway (a 900mhz duron just doesn't cut the mustard anymore)

Duper · Post by **Duper** » Thu Jun 29, 2006 8:26 pm

the top of the Caps aren't cracked are they? Or buldging?