dead chip, dead motherboard, or . . .?

[DrMrLordX]

While relentlessly fiddling with my system, I seem to have killed my CPU while (re)lapping it to get some gallium-based TIM out of the IHS. Or maybe I static-killed it during handling, or maybe it was just a bunch of NB overclocks/overvolts that finally caught up to me. Who knows.

Anyway, the system now refuses to post, and instead the harddrive light sticks on (power light does not come on). All board elements power up normally, though the board (MSI 790FX-GD70) does not report any error codes (not even FF).

Something's dead, and I'm pretty sure it's the CPU, though it may be the board. Any thoughts on which it might be? Is there something obvious I'm missing here that is not board/CPU related?

I didn't get any of the offending TIM on the motherboard, or any water/ipa/etc. either, and while touching board components I stayed fairly-well grounded'; the CPU, however, is a different story altogether. I did not stay grounded while handling it, which is par for the course when lapping. For whatever reason, it didn't survive this round, though it was not in a good state of health before re-lapping anyway (tons of memory errors, presumably due to an inadequate TIM application causing the region of the IHS over the NB/IMC to receive little to no cooling).

 

[jvroig]

There are also no beeps? (assuming you have an internal speaker connected; both of my boxes don't, so I have to do without beeps during the fortunately rare hardware debugging times).

Last time something like that happened to me due to excessive (or is it "normal" for us hardware types?) fiddling (screwing and unscrewing the motherboard, reapplying TIM and reseating aftermarket coolers with different backplates, etc), I almost went crazy, going through everything like reseating the coolers, making sure there is no metal contact at the backside of the motherboard, etc... after almost giving up, I tried the easiest thing to do (which I should have done first since it was practically effortless) and reseated the RAM. Finally, the monitor finally responded after turning on the PC (previously, the box would light up completely, and all fans - CPU and aux fans - would power up, but the monitor just won't power up, no signal).

Is this your Propus whose life is under question?

 

[DrMrLordX]

Yeah, it's the 635. I have no PC speaker hooked up, so there are no beeps.

Reseated the RAM? Hmm. I wonder if I've bumped it loose.

edit: reseating the RAM and trying each DIMM independently did not produce positive results. I suppose I could try booting without RAM at all to see if it will produce different errors, but that will have to wait until later.

HDD light still sticks on without any errors popping up on the GD70.

 

[DrMrLordX]

I seem to have found the culprit. Copper sediment (presumably from the lapping) has deposited itself between many of the outer pins of the CPU, almost assuredly leading to a short somewhere. Not sure if I can resurrect the chip by blowing it off with canned air . . . I'll try that later. Hopefully the socket didn't get full of that crap.

 

[jvroig]

After blowing the bits off the CPU and the socket with canned air, you could try it again. Unless there are obvious burn marks like those that happened in faulty foxconn sockets with early lynnfield boards, I think it would probably be revived if that is in fact the culprit.

Do you have another AM3 proc/mobo, can't do a switch to see which one between your board and cpu is actually causing trouble? (What unit are you using now that your Propus is currently in life support?)

 

[Gillbot]

I'd try cleansing the chip with 92+% rubbing alcohol then letting it dry.

 

[BD231]

If you can power up, the motherboard should be just fine; no sounds would lead me to think CPU as well. A good example is pushing your overclock to high to post, resulting in a powered on system, black screen and no use of the CPU.

Why lap your AII though? That's not going to get you any headroom ...

 

[DrMrLordX]

After blowing the bits off the CPU and the socket with canned air, you could try it again. Unless there are obvious burn marks like those that happened in faulty foxconn sockets with early lynnfield boards, I think it would probably be revived if that is in fact the culprit.

The canned air I have right now is insufficient to remove the copper sediment, so I will get a new can or try to find someone with a compressor or something. There are no burn marks so there is still hope.

Do you have another AM3 proc/mobo, can't do a switch to see which one between your board and cpu is actually causing trouble? (What unit are you using now that your Propus is currently in life support?)

Nothing else is currently available. I'm posting from my old x2 which . . . I don't think would work in the GD70, and is my wife's current machine anyway, so no tearing it down.

I'd try cleansing the chip with 92+% rubbing alcohol then letting it dry.

I used 91% which didn't seem to remove much. Gonna try some better air plus possible additional cleansing with the ipa.

Why lap your AII though? That's not going to get you any headroom ...

Mostly to reduce the amount of TIM that is necessary to cool the chip. The x23-7783D application I had on it originally (before I started dicking around with Liquid Metal Ultra) used maybe 1/4 the TIM people normally use, and the temps were quite good. Besides, copper is purty. That is, until it infests the pins on the bottom of the CPU anyway.

Also, I had to re-lap the CPU to get Liquid Metal Ultra off of it. That stuff leeches into copper like nobody's business, and apparently, you can't really get the best temps applying Ultra onto a surface that has already been . . . altered by Liquid Metal Ultra. Or so it is said. I don't know how much I really want to lap the chip + HSF again, so if I do revive it, I might try applying paste to the altered surface anyway.

 

[BD231]

I'd just see if you can't use a really small sewing needle and scratch the stuff off the base of the pin area. It would fit perfectly between the pins and the sharp tip would pop/loosen those little specs up easily. Then you can just wash it away with the alcohol. Even if you got it on the pins you'd be alright leaving it on as long as the base of the pins are clear of that stuff.

I've never handled a CPU while I was grounded, nor have I seen a chip die and there's a lot of carpet in my house. I'd say you have a very good chance at getting that thing back with a little work.

 

[DrMrLordX]

You must by psychic. I was thinking of getting some needles as well . . . so I got some. Will try to scrape the sediment out when I get the time.

 

[DrMrLordX]

Good news and bad news.

The good news is that the HDD no longer sticks on when I try to boot with the processor installed; I went to work with a needle and a lot of 91% ipa to clear out the copper sediment that was (apparently) shorting numerous pins. The bad news is that now I'm getting error code FF, which is basically "CPU can't initialize" or something along those lines. Might be bent pins or problems with the motherboard, but at this time I can't really tell. I guess I'll have to reseat the memory and keep trying to get it to boot, or tear it down again and make sure all the pins are lined up properly, because some of them DID get bent a wee bit as I was trying to get out the copper sediment (dammit).

 

[jvroig]

I think you now have to buy a Thuban earlier than you expected.

 

[VirtualLarry]

FF as a post code means "boot sucessful". At least it does on my DFI X48 board with LED readout.

 

[DrMrLordX]

I think you now have to buy a Thuban earlier than you expected.

Ha ha! Yeah, maybe so.

FF as a post code means "boot sucessful". At least it does on my DFI X48 board with LED readout.

Well, it does if it moves on to other post codes after that. If it sticks on FF, it means the board recognizes the CPU but that the CPU can't or won't do anything. I think this is what happens if you try to boot with no memory, for example.

 

[Duvie]

so is there a chance some more copper debris is in there and the pin is shorting out?

Take the mobo out and hold it upside and lightly tap the back of the motherboard to see if you can pop some of the pieces loose if there are soem remaining.

You may need to test the cpu on another board just to see if it is the socket or the cpu itself now...

 

[DrMrLordX]

It is possible that there is more debris shorting pins, either clinging to the underside of the CPU package or in the socket itself.

I've got a Sempron 140 in the mail to serve as a test unit to see whether or not the board is having problems. I'll be sure to see if I can shake any debris loose from the socket itself after the next teardown.

 

[DrMrLordX]

Sempron 140 engaged. Now to overclock it . . . mua ha ha!

(yes, the board still works. Yay)

 

[jvroig]

Make it so, Mr. DrMrLordX.

(And try an unlock as well?)

 

[tweakboy]

You can try this test.

Run a prime or everest stability test and let it run for 1 hour. If comp still working then your fine, touch the cpu or hsf and see how hot it is. Also go to BIOS and reset everesthing to default values. Then let us know whats going on.. thx

 

[deimos3428]

I've got a Sempron 140 in the mail to serve as a test unit to see whether or not the board is having problems. I'll be sure to see if I can shake any debris loose from the socket itself after the next teardown.

I don't have any answers for you other than "check for bent pins", but I'm convinced -- a Sempron 140 is an excellent and economical backup CPU. Getting one today.

 

[DrMrLordX]

Make it so, Mr. DrMrLordX.

(And try an unlock as well?)

I'll be making a thread about that eventually . . . but for now:

25C idle, 27C load? At 1.35v vcore? This chip looks like fun. That was without the Honeywell room fan blowing on it, too. Firing that up brought load temps down to 26C and idle temp down to 24C. It won't run so cool when I go above stock and/or unlock it, but whatever.

I still need to figure out if my beta BIOS supports Thuban already or whether or not I should update to the release version, but that won't stop me from playing with the chip a bit in the meantime.

You can try this test.

Run a prime or everest stability test and let it run for 1 hour. If comp still working then your fine, touch the cpu or hsf and see how hot it is. Also go to BIOS and reset everesthing to default values. Then let us know whats going on.. thx

I ran Prime95 blend for about 4 hours last night. Rock solid, no errors. It works. Nothing is hot, and I don't think this chip actually produces any heat . . . or not enough to register with the nh-d14 cooling it.

I don't have any answers for you other than "check for bent pins", but I'm convinced -- a Sempron 140 is an excellent and economical backup CPU. Getting one today.

Tell me about it. These chips will be useful as backups for a long time I suspect. Also look out for the new Sempron 145 (C3 stepping) that has a higher multiplier and should be just as unlockable. They aren't for sale yet anywhere that I can find, but they should be out there soon-ish.

 

[DrMrLordX]

Well the stupid thing doesn't seem to want to unlock. Unlocked at stock clockspeeds (2.7 ghz) all it does is reboot while loading windows, repeatedly, no matter how much extra voltage I give it. I stopped at 1.5v.

As a single-core CPU, looks like it tops out at 3850 mhz, even though it doesn't break 31C while priming, and that's with 1.48v vcore. Oh well, can't complain for $25 I guess!

This chip hardly merits its own thread unless I can find some way to make it dance . . .

 

[DrMrLordX]

double post blah

 

[jvroig]

As a single-core CPU, looks like it tops out at 3850 mhz

Would that be more or less similar to the single-threaded performance of a Deneb core at 3.4? Lower or higher?

 

[DrMrLordX]

Hard to say for sure; Sargas (Sempron 140/145) has no L3 but 1 meg of L2 per core. Deneb can contribute its entire L3 cache to a single core when running a single-threaded app, but it will only have use of 512k of L2. So, it's probably a wash given identical NB speeds and memory speeds/timings.

The Sargas might win out at high memory/NB speeds + tight timings in tasks that have a large enough working set that cache has less impact on performance (SuperPi 32M, for example). I was already able to beat Deneb with Propus in SuperPi 32M, so Sargas should do even better.