3 Burned Processors - any ideas???

[Gudhjem]

I'm stumped.

I'm putting together a new system for a friend and have now burned out THREE Athlon 1600+ (1400Mhz) cpus. All three times the system ran OK for between 20 seconds and 2 hours, then went black. I know it?s the cpu because the fans spin at power up but no post and a cpu from another machine (Athlon 2000+) will post.

Troubleshooting
The heatsink and fan (PAL6035) is squarely seated, and, yes, I've used heatsink compound sparingly. I can tell it was seated correctly from the square pattern created in the paste when I remove it. This fan and heatsink worked fine for a year on a 1.33 Ghz.
The mobo is not shorting anywhere on the case. I've quadruple-checked
In two of the burn-outs, nothing was connected but video card and memory (in the 2 hour run, all was connected and Windows XP installed)
I've tried a new power supply (Sparkle 300w) after 1st burn out
I've tried a new mobo after second burn out (1st was a gigabyte GA-7ZXE, 2d an MSI K7E turbo 2, both via KT133a boards).
I've tried a fast-clamping surge protector after second burn out.

Third burn-out still happened, and this time after ony about 20 seconds.

Any ideas as to what could be causing this would be appreciated.

 

[Buz2b]

The mobo is not shorting anywhere on the case. I've quadruple-checked

Just to get the discussion started, have you tried booting with the board outside of the case? Even though you "quadruple-checked" that (booting outside the case) would be a more definative test/result. The other (obvious) things I would check would be the CPU installation. Did the lever go down easily and completely to a locking position? Have you tested the PS yet, whether on another system or by using an ATX PS testor. Antec makes one that is sold at many stores for under $14.00. The only other (longshot) idea would be if you have any high capacity fans connected to the MB. Also, when you say "burned out", what exactly do you mean? Can you describe the condition of the CPU and what made you conclude that they were burned out?
Anyway, that's a starting point. Good Luck!

 

[NyCxSpyder]

You didn't tell us anything about how to place the fan and heatsink. Are you sure you did it correctly? The one flaw of AMD processor is that they leave their core open unlike P4's where Intel included a metal shield on top that protects the core and aids a little in cooling.
Are you sure you didn't crack it? (Yes, all 3x) You should know this but as a reminder, when you put your heatsink on top of your processor, use the SLIGHTEST angle possible because you do not want to have a lot of pressure on one side of the processor.

If it wasn't that, I suggest you experience with booting up the computer with the case open and a house fan on the side blowing inside the computer case. Obviously, to me, it sounds like a heat problem. Maybe the heatsink/fan isn't recommended for the AMD processor you have. But first, I suggest trying with the house fan. If all is successful, then I would say your problem was a heat issue and that you need more fans and a better heatsink. Good Luck.

 

[LiLithTecH]

I think you have to start looking for the common denominator.
What was it that you used on ALL the builds?
Ram? Video? Case?

Are they all OEM CPU's? From the same vendor?

I think it is doubtful that it is caused by the PSU. The even the cheapest Sparkle is more than adequate.

What settings are you using to configure the FSB.
Was it the same on ALL 3 builds?

Even with a clamping surge protector, I would try another powering from another location.

Or it can be just a BAD run of CPU's. Is the vendor warranting them?

Just a side note.
Based on the Website Spec's and manual, both the gigabyte GA-7ZXE, 2d an MSI K7E turbo 2
are shown to go to AMD Athlon 1.5 ghz + (which more than likely requires a BIOS update to support beyond 1.5)

 

[Gudhjem]

Thanks for the feedback. Some more info in response to the issues raised above:

1. I concluded the cpus are burnt out because a) power up spins the fans and gives no POST, b) a cpu from a different machine does boot c) the burnt cpus won?t POST in a different, working machine.
2. I did boot with the mobo outside the case for the 1st cpu. Everything was fine. It was also fine inside the case for about a half hour when all locked up and went black during Windows install. None of the cpus will work outside the case after having gone black, nor will they work in a different and working system.
3. The ZIFF socket lever went down easily, felt fine.
4. It is of course conceivable that the h/s was not properly installed, though I have built at least 5 Socket-A Athlon systems and have never had this problem. I even used this very same heatsink on another board for about a year (it?s a hefty PAL-6035 copper insert h/s that I used to use for overclocking). The only notable thing is that the clip for this h/s seems VERY stiff and it takes an amount of force to get it in place that makes me uncomfortable. It does get in place though, and remember that 2 of the cpus worked for a little while before burning. I install the h/s by first putting a very thin layer of h/s compound on both the h/s and cpu, then hooking the off-side of the clip over the retainers on the socket, then squaring the h/s over the four pads, then, while placing pressure on the offside of he h/s, I use a screwdriver to hook the clip over the near side of the cpu. Yes, the h/s is oriented so that the indentations match the raised edge of the socket and the lever side of the socket.
5. I have not tested the Power supply, though this is the second new power supply I?ve tried. I bought the Sparkle after the 1st cpu went, since that seemed the most likely explanation at the time.
6. I have fans in the case, but the only fan connected (other than for the cpu that took a couple of hours to burn) was the cpu fan.
7. The case was always open. Can?t imagine ambient temp could be an issue, and certainly not for the one that burned in 20 seconds.
8. The only common denominators include: case (a new in-win 500a); ram (2 sticks of 128mb PC133 SDRAM that ran fine on another machine for a year); video card (an old TNT1); and monitor. Even the CD-ROM and floppy were not hooked up yet for the 20 second burnout. Eliminating common denominators has been my approach, hence the new cpus, new mobo, and new power supply.
9. fsb has been jumpered to both 100MHz and 133Mhz settings in first two burnouts, only time for 133 in last one. In all 3 burnouts, the bios correctly recognized the cpu as 1600+ @1400MHz.
10. Different vendor for the cpus (2 from newegg, one from M-Wave).
11. For the cpu that lasted 2 hours, I was able to run a bios update (on the GA-7ZXE). Cpu still burnt out about an hour later.

Questions:
If I cracked (all 3) cpus, would I be able to see the crack? Nothing shows, they all look fine physically. Any reasonable chance they could run for awhile (as 2 of them did) before failing?

If not cracked, it also seems to me that heat must be the issue. The cpus behave just like overheated cpus, but I run out of ideas as to why that could be happening. I?m on the verge of buying a Dell (just kidding).

I appreciate the time taken to respond. Any ideas or theories are appreciated.

 

[Buz2b]

First off, you seem to have really "done your homework" in trying to remedy this. Second, I will just comment on a couple of points. You said, " did boot with the mobo outside the case for the 1st cpu. Everything was fine. It was also fine inside the case for about a half hour when all locked up and went black "
Since that time the MB(s) have been inside the case I presume, except AFTER they "burned out" or whatever went wrong. You then pulled the MB out and tested to see if they would run. Maybe the damage was already done. It would seem that there could be a short between the MB and the case. I know you said you "quadruple-checked" but the reality is that it still remains a viable theory. Possible yone of the standoffs is not making contact exactly where it should or a standoff is in the wrong place, or a screw head is larger than the insulated area around the install hole, etc (to mean all the other various assorted things that can short a MB). Just an idea, if I read that correctly. Just what you need to hear, but one way to find out is get another CPU and run it outside the case for a few hours at least to see what happens. Of course, NewEgg and MWave are going to get tired of RMA'ing CPU's too. J/K
Another minor point (which would not cause the failures you spoke of) that I would mention is that you said, " I install the h/s by first putting a very thin layer of h/s compound on both the h/s and cpu" Another possible communication error but you should only put a very thin layer of the compound on the CPU die, then you should take a small amount of compound and rub it onto the bottom of the HS (with a lint free cloth, not your finger) and then gently wipe it off. This is done on the HS only to fill in any microscopic imperfections or scratches in the surface. You should not leave a "layer", thin or not, on the bottom of the HS; it should be wiped off. If not, then that would essentialy mean that there would be two layers of compound between the HS and the CPU die. A definate "No-no". However, as I said, this would not cause the failures, only a bit more heat would be present.

As to the physical condition of the CPU's, if you have a good magnifying glass you can take a close look to see if there are any fine cracks in the CPU. Sometimes they can be literally a "hairline" crack that is almost invisible. Yes, they might work for a brief period it that was the case. You should also look closely at the area where the die is "connected" to the ceramic part of the CPU. A "squeezing" fracture could occur there.
As to being "burned", perhaps that should be further defined. If they are in fact burned due to excess heat, you may well be able to see an "off color" in the die. Hard to explain unless you've seen it but there would be a difference in color from a normal cpu. But, if they were "shorted" you might not see anything out of the ordinary in most cases.

 

[Gudhjem]

Thanks for all of the feedback.

Although nothing seems particularly obvious at this point, it seems the most likely explanation is that, despite my best efforts and the fact that I've never had this happen before, I've managed to damage all 3 cpus during installation.

I've ordered a fourth cpu (no, I didnt rma), and this time a new h/s as well. I got a Globalwin TAK68, because it uses an adjusting knob to tighten the clips, rather than clip tension. Seems like a more idiot-freindly install, which may be what I need...

I'll post if it works.

 

[Buz2b]

Good luck. Hope it all works out.

 

[Gudhjem]

Update:

I replaced the cpu (4th one . . .) and installed it with the new heatsink. As I mentioned, this one did not use clip tension to tighten, but instead an adjusting knob on top. The PC has now run fine for several days, and has been delivered to the happy end-user. My guess is that I must have cracked the first 3 cpus during install. Live and learn.

Thanks for the feedback.

 

[gonzo2k]

One CPU (4) per post (4)- a new record!

GL