Xeon X5680s Now $45... October 2018

[BobK]

Here's somethin'.

My hot-rodded 2009 Mac Pro 4,1 -> 5,1 has been having problems the past several days, or maybe weeks. It would stop responding to the (wired) keyboard connected via a hub to one of the front USB ports. And/or, my Magic Trackpad would get disconnected.

I thought there was a problem with the USB hub, or maybe with Bluetooth (which has often been flaky). Or both. (Bluetooth adapter is plugged into the hub.)

Today, I was typing in Apple Notes when the computer put itself to sleep and then shut down.

Pressing the power button, I'd hear disks and fans spinning, and the white LED on the front would light up, but there was no startup chime.

On my laptop, I did some web searches and opened the Apple Service Manual, which I found online last year. Based on some forum posts, I thought it might be the PSU, and did some tests (had to remove almost everything from the case except the motherboard and PSU).

Then I removed one of my de-lidded 5680s - still no chime. At this point, I should have put that one back in and removed the other, but I got impatient.

Instead, I replaced both 5680 CPUs with the original 2.26 GHz CPUs. It started up fine.

So..one of the 5680s appears to be bad [EDIT: Maybe not. See Aug 2 post below]. I'm running low on thermal paste, but when I get more, I'll see which of the 5680s is bad and maybe replace it. The question is whether to de-lid the next one myself, or buy a de-lidded one. There's no way of knowing whether my de-lidding shortened the CPU's life, though.

This is the second CPU failure I've had. The first one was DOA.

Anybody else had CPU failures several months after installing?

[HCMarkus]

I've never had a CPU failure on any of the Macs I have updated, but I have never done 4,1 dual CPU job, meaning I've not had to deal with the de-lidding thing.

Updating the thread title, used X5680s are now running under $60 each on eBay.

[mikehalloran]

There's always that old culprit—too much dust in the heat sink fins. It's always assumed one has looked at that but...

[BobK]

There's always that old culprit—too much dust in the heat sink fins. It's always assumed one has looked at that but...

I didn't check for dust this time, but I'm pretty sure I blew out the dust when I did the upgrade last winter.

If dust in the heat sink is the problem, wouldn't it affect both the 5680s and the original (cooler-running) 2.26GHz CPUs?

[Michael Canavan]

Right away my first thought is too little thermal paste or one of the heatsinks wasn't tightened down enough. When I first installed my de-lidded CPUs I was far too cautious with tightening down the heatsinks, and that caused kernel panics. It's very possible the same thing happened with your situation.

BTW ther are two small apps you can get for your mac that IMO are a big help with any modified computer: smcFanControl allows you to override the lower fan rates of your mac, I have mine set to about 30% higher rates than normal, and you can of course do the opposite and slow the fans down if you're recording quiet parts with the mac in the same room etc.

iStat Menus covers pretty much everything else, CPU heat levels etc. it's about $15.

All told I really haven't had any issues at all with my de-lidded 12 core, except maybe the fact I did the de-lidding myself, and it's somewhat scary to do.

[BobK]

Right away my first thought is too little thermal paste or one of the heatsinks wasn't tightened down enough.

I don't think it was either of those. But if/when I test the 5680s again, I'll check bolt tightness.

I have iStat menus. Didn't notice any changes to the temperatures in the menu bar, but I didn't dive in any further. I thought maybe the Northbridge heat sink plastic rivets had broken again, but they were fine.

Which de-lidding method did you use, Michael?

I did the razor blades + clothes iron. If I do it again, I'll try the vice method.

Thanks!

[Michael Canavan]

I did the razor blades + clothes iron. If I do it again, I'll try the vice method.

razor blade iron method. Personally the vice method is more than suspect to me. Too easy to really damage the chips. They hit 170F all the time so I can't see how the iron method would damage the chips.

[BobK]

I bought more thermal grease and did some tests.

To recap, the original processors work fine.

The computer will boot with either of the 5680s in slot A and nothing in slot B.

But it will NOT boot with both 5680s (with one exception). I tried both combinations (CPU 1 in slot A, CPU 2 in slot B and vice versa).

I really doubt it's matter of thermal grease or bolt tightening. I tried tightening the screws tighter than the service manual directs, and it didn't help. (Apple says "finger tight plus 1/4 turn".)

The exception is that the first time I tried both 5680s today, I forgot to put on the plastic retaining frame on the processor in slot B and it booted without issue. I shut down, put the retaining frame back, and it wouldn't boot. (I didn't try this additional times to see if it's repeatable.)

Any ideas?

[Michael Canavan]

I bought more thermal grease and did some tests.

To recap, the original processors work fine.

The computer will boot with either of the 5680s in slot A and nothing in slot B.

But it will NOT boot with both 5680s (with one exception). I tried both combinations (CPU 1 in slot A, CPU 2 in slot B and vice versa).

I really doubt it's matter of thermal grease or bolt tightening. I tried tightening the screws tighter than the service manual directs, and it didn't help. (Apple says "finger tight plus 1/4 turn".)

The exception is that the first time I tried both 5680s today, I forgot to put on the plastic retaining frame on the processor in slot B and it booted without issue. I shut down, put the retaining frame back, and it wouldn't boot. (I didn't try this additional times to see if it's repeatable.)

Any ideas?

Ok so I have to state first off that I’m not there i the room with you and your dilemma so I cannot be held responsible for anything that happens because you read my story here.

I had one cpu fail when I mounted it at first too. IMO the dried thread lock left over after you remove the heat sinks is the culprit, at least it was in my case. It made it seem like I had tightened it all the way down when I hadn’t. Everything you mentioned points to that, that it worked without the retaining frame for instance. This all begs the question how do you get dried thread lock out of a thread hole??

Anyway tightening more worked for me, I suppose it’s possible to damage the CPU but obviously not as easy todo without the lid on it..

[BobK]

How/why would dried thread lock affect the 5680s and not the original CPUs?

[HCMarkus]

I haven't done a dual CPU 4,1 swap, so cannot speak from experience, but If it worked without the plastic retaining frame, might that not suggest it really is a tightening or alignment issue? Maybe MIke C has it right...

OTOH, you don't want to bend the CPU tray pins.

OTOOH (On The Other Other Hand), with a de-lidded CPU, it probably is almost impossible to bend the CPU tray pins.

Hope you get it sorted sooner rather than later.

[BobK]

I tried a few more things, none of which helped:

1. After reading forum posts mentioning dust, I removed the power supply this morning, then opened its case and vacuumed and blew out as much dust as I could. There wasn't that much dust, and I didn't see any obvious signs of fried parts.

2. I tried the 5680s again. Made sure the bolts were tight.

3. Removed the plastic frame from the one in slot B.

However, looking at the processors with a magnifying glass, I remembered that when de-lidding, I accidentally sliced off part of one of the small protruding silver rectangles. (I recently read on a forum that these are capacitors...)

But if that's causing the problem, why did it work fine for almost 9 months? And why does it still work when installed alone in slot A?

I'm considering replacing that CPU, since it's about a third of the price of a power supply.

[James Steele]

However, looking at the processors with a magnifying glass, I remembered that when de-lidding, I accidentally sliced off part of one of the small protruding silver rectangles. (I recently read on a forum that these are capacitors...)

But if that's causing the problem, why did it work fine for almost 9 months? And why does it still work when installed alone in slot A?

I'm considering replacing that CPU, since it's about a third of the price of a power supply.

That's what I'd do. My story about the upgrade of my 4,1 was pretty tragic. I saw a YouTube video where a guy wet-sanded the tops of the processors to get all the solder off of it and ended up screwing up those capacitors you're talking about. Blew a lot of money wrecking two processors and finally I sent mine off to a guy who did the de-lidding for me and they worked great.

I don't understand why it worked, but if you messed up one of those capacitors, I wouldn't even chance it and get another processor.

[BobK]

That's what I'd do. My story about the upgrade of my 4,1 was pretty tragic. I saw a YouTube video where a guy wet-sanded the tops of the processors to get all the solder off of it and ended up screwing up those capacitors you're talking about. Blew a lot of money wrecking two processors and finally I sent mine off to a guy who did the de-lidding for me and they worked great.

I don't understand why it worked, but if you messed up one of those capacitors, I wouldn't even chance it and get another processor.

Thanks, James! Gonna order one right now.

[BobK]

UPDATE:

I'm back in business. The twist is that the bad processor was NOT the one with the partially-shaved-off capacitor!

=====

The story:

I bought another 5680 on eBay. It arrived today and I de-lidded using the clothes-iron method.

Installed it in CPU slot A, with slot B empty, and the Mac booted without issue. So far, so good.

Then I added the old "good" 5680 - without the shaved capacitor - in CPU slot B. Same problem as before: It powered up but wouldn't boot.

I was afraid I might have one of those elusive power supply issues like I saw at the Mac Rumors forums. One poster said Apple replaced many of his Mac Pro's internals, but the shutdown problem remained. They gave up and refunded him the repair fees. Then he replaced the PSU, even though Apple hadn't found it to be a problem, and it resolved the issue.

Before contemplating that option - which is $180 to $300, depending on the vendor - I decided to try the 5680 with the shaved capacitor in slot B, and it booted! All RAM is recognized - everything seems fine. Fingers crossed.

This was a PITA, but I'm glad it worked.

Thanks for the help and encouragement, guys!