Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

CPU upgrade Success and Failure

RAMtheSSD

RAMtheSSD

macrumors regular
Original poster
Mar 23, 2013
126
22
High on a Mountain Looking for Wisdom
Thinking that I could learn a lot and not spend a fortune in a disaster, I bought some used 5620s and got all set to work on my 2009 cMP while dreams of 36k gerkbench scores and 5590s danced in my head. The install was problematic only in so far as the thermal pad material in that it is messy, not adhesive, and 5mm is not enough to replace (even double wasn't enough!) the original material. Yesterday, after resetting the smc and pram, it turned on and saw one processor. I even got to print the system report and then...WHAM! Out like a light. So this morning, I redid the thermal pad (processor B's did not look like it was making contact) and smc and pram and I even got to geekbench! The score was an improvement of more than 1000 pts! Then, WHAM! Again, out like a light! The fans are both working and responded to SMCFanControl so, thankfully, it isn't the heat sink fans. It sees both processors today so it isn't the sockets --I was worried because pin damage is the source of many horror stories.

I thought I had the tiger by the tail but something is wrong but what? I get that the shutdown is protective but what is it protecting? The fans are working and the temperature is less than 100F (boost a) when the shutdown happens. Has anyone had this happen to their cMP?

Update: As of 1226 mst she won't boot to the chime so I can't boot to a hardware test.

Update: 1316 mst and still no boot to chime. Anybody think it could just be a coincidental failure of the power supply? If so, how would I know or how could I test it?
 
Last edited:
I doubt it is your power supply. There are too many stories from dual-CPU upgrades having to do with CPU positioning and heatsink tightening for me to think its anything other than that.

I'll assume you didn't delid since you got additional thermal padding, so how did you tighten the heatsinks? Did you use the washer method? Careful tightening method? Or just tightened away with no care in the world?
 
ActionableMango said:
I doubt it is your power supply. There are too many stories from dual-CPU upgrades having to do with CPU positioning and heatsink tightening for me to think its anything other than that.

I'll assume you didn't delid since you got additional thermal padding, so how did you tighten the heatsinks? Did you use the washer method? Careful tightening method? Or just tightened away with no care in the world?
Click to expand...

After looking at every piece of data a I could find, M4 washers 0.8mm thick and 3 per bolt for a total of 12 washers per processor appeared to be what was used by the majority of people reporting success. Two 1.1mm thick washers was the size next most often used but the plug and play kits come with the M4 and 0.8 washers. Originally, I was concerned that a conductive material would be problematic and I was going to use nylon washers but how many kits could someone sell if they burnt customer's computers? So brass and I tightened the screws back up by turning the wrench by the shaft between my index and thumb. I tightened the screws 1/4 turn each before moving to the diagonally opposed screw and I stopped when there was resistance. I counted the number of turns to remove the heat sinks and I expected to turn the screws a maximum of 4 full turns and no screw got 4 full turns. I reasoned that given resistance and less than 4 full turns would be tight enough to start and not seeing all the memory has been reported as indicating the heat sinks were too loose and I could always tighten the screws further: not like they are holding against a windstorm, right?!

Since the system reported all 32gb and two processors, I reasoned that everything was tight enough. Then, geekbench went fine and there was the sudden shutdown and there has been no chime since.
 
It sounds like you did a very good job. I think some trouble shooting is in order. Process of elimination can narrow down the problem.

Pull CPU B and see if it boots up. If it doesn't boot up, problem is with CPU A or CPU A socket. If it does boot up, problem is CPU B or socket B. Next would be to put CPU B in socket A and see if it boots up, etc.

When you have the CPUs out, might as well check for pin damage. It sounds like you prevented the possibility of that with the washers, but its worth checking anyway.
 
Something very much like this happened to me when I upgraded a MP 1,1 from 2x2 core to 2x4 core. I bought a pair of used CPUs on eBay.

I did the install, it booted . . . ran for a bit . . . then just stopped. I reinstalled the old CPUs, up it comes and runs with no problem.

The person who sold me the CPUs took them back, so I wasn't out any money. I got a new pair and they worked fine.

So consider the possibility that the used CPUs just died. It does happen.
 
Last edited:
RAMtheSSD said:
I will do as you suggest but if the processors or the socket were damaged, why would it work long enough to do geekbench?
Click to expand...

I don't know that a socket or a CPU is damaged. The testing steps are just to gather additional data which may shed light on the problem. There are other things to do as well.

If you want to check the power supply, check the diagnostic LEDs. Press the DIAG button on the backplane and see if the "PSU PWROK" and "5V STBY" LEDs illuminate.

In fact, I don't know why I didn't start with this, but are any of the error LEDs illuminated? For example the red LED near socket B? Or any of the memory banks too.
 
So I resat the real time clock (switch by the battery) and got a boot. Even though everyone might feel like cheering, don't. The fans that had been running fine before (that is, this morning) are now going all out. Worse, it now thinks it is a single processor machine and that remains even after putting the other heat sink and processor back in and to top it off, 8gb vanished! The hardware test won't get past an smc error (reset or not) and when it sort of did (can't tell, never got past the probe) the hardware test just hanged! I am thinking of taking all the ram off and testing with one stick at a time while doing multiple smc resets (apparently, some people with iMacs have had a similar problem after upgrading the video card (news to me! I didn't know one could!) and somehow that worked.

Incidentally, the power supply led is green and the standby led lights up amber which is scary until you see photos saying that it is supposed to be amber --weird if true.

Earlier, when it booted with just CPU A, there was a red light but even when it said it was a single processor machine, the light wasn't on.
 
I am leaning very strongly toward a bad CPU or damaged socket. Did you examine socket B? Swap the two new processors with each other? Swap the two for your old processors?

Yes that LED is supposed to be amber. I believe the power supply is fine. Those LEDs are the check for the power supply suggested by the service manual.
 
I am going to leave it be for the weekend but Monday, I will go back to the old processors. I took photos of socket (at different times; seemed like a wise precaution since the pattern of the pins can play tricks even on the best eyes) B an there seems to be no difference even in black and white.

Thank you both for your help. I will start back up on Monday and I hope you will both be around then.
 
To the OP. When I upgraded my CPU's in my 2009>2010 cMP I installed one CPU at a time. I took my time tightening the screws on the heatsink until it showed all the ram for that CPU. I repeated the process with CPU 2. I didn't fool with thermal pads at all. System has been rock solid. Zero heat or fan problems.

Hope you get your cMP up & running :)
 
I have the opportunity to pick up a dual tray with everything intact, including the 2 CPU's for £120 on ebay. Unfortunately CPU B socket has some damage. Is it worth buying just for the ancillaries to form the basis of a future dual CPU build or not? Part of me says get it for the heatsinks alone, as they're pricey, and the processors could be sold for a few pounds on eBay. A new dual board is going to be around £400 I've found, and then I'd have to get a pair of Xeons later.

Worth the money and bother, or wait for a fully done dual tray at a later date?
 
ActionableMango said:
Kinda wondering what happened.
Click to expand...
I thought some days off would help me relax and give me some perspective because I was frustrated. I will get back to it after Big Bang today; it feels like an adventure again.

I'm thinking of putting the 5520 in socket b and leaving the 5620 in socket a (it detects it and seems to work so why change it?) and seeing if it boots unless someone here thinks that is a bad idea.
 
And thoughts on my query about the value in buying a dual cpu board with one cpu socket damaged? Thinking logically the metal tray and heat sinks are worth close to $300 if prices online are to be believed. That would leave me needing to buy a new CPU board and the CPU's themselves at a later date. Someone would probably buy the 2x 2.26GHz CPU's in the meantime to recoup a bit of the $150 for the currently damaged complete CPU tray I figured.

Then there's a slight chance that by some miracle I could fix the current board. If not, it'll be an interesting display in a picture frame on my cabin wall!

I've priced up rough dollar amounts, but I'm in the UK.
 
Stephen Valente said:
And thoughts on my query about the value in buying a dual cpu board with one cpu socket damaged? Thinking logically the metal tray and heat sinks are worth close to $300 if prices online are to be believed. That would leave me needing to buy a new CPU board and the CPU's themselves at a later date. Someone would probably buy the 2x 2.26GHz CPU's in the meantime to recoup a bit of the $150 for the currently damaged complete CPU tray I figured.

Then there's a slight chance that by some miracle I could fix the current board. If not, it'll be an interesting display in a picture frame on my cabin wall!
Click to expand...

When you look at Ebay for the market value of these parts, you need to look at "sold listings", not asking prices, and not completed items that didn't sell. People can ask for whatever they want--it doesn't mean it's selling for that price. This is what I do in order to ensure confidence (or discover that my parts have no value).

I cannot speak for the UK, but here in the USA the base model CPUs for Mac Pros are basically worthless so I am surprised that you'll be able to recoup money from selling them. Also I can't believe there's that much of a market for people just buying the metal tray and heat sinks. Even if they will eventually fetch some money, you might be stuck trying to sell those for a while.
 
ActionableMango said:
When you look at Ebay for the market value of these parts, you need to look at "sold listings", not asking prices, and not completed items that didn't sell. People can ask for whatever they want--it doesn't mean it's selling for that price. This is what I do in order to ensure confidence (or discover that my parts have no value).

I cannot speak for the UK, but here in the USA the base model CPUs for Mac Pros are basically worthless so I am surprised that you'll be able to recoup money from selling them. Also I can't believe there's that much of a market for people just buying the metal tray and heat sinks. Even if they will eventually fetch some money, you might be stuck trying to sell those for a while.
Click to expand...

I was thinking more of the cost of buying heatsinks and the tray for the time when I was able to replace the CPU board and get two processors. Those parts aren't cheap on the US stores I've looked at, so figured that buying even a half broken circuit board with these items for around $150 would be cheaper than a pre-built dual CPU board at a later date.

That's my logic anyway.
 
Stephen Valente said:
I was thinking more of the cost of buying heatsinks and the tray for the time when I was able to replace the CPU board and get two processors. Those parts aren't cheap on the US stores I've looked at, so figured that buying even a half broken circuit board with these items for around $150 would be cheaper than a pre-built dual CPU board at a later date.

That's my logic anyway.
Click to expand...

Oh I see now, I misunderstood. Good luck with your project!
 
RAMtheSSD said:
I thought some days off would help me relax and give me some perspective because I was frustrated. I will get back to it after Big Bang today; it feels like an adventure again.

I'm thinking of putting the 5520 in socket b and leaving the 5620 in socket a (it detects it and seems to work so why change it?) and seeing if it boots unless someone here thinks that is a bad idea.
Click to expand...

I doubt having 2 different CPUs installed at the same time would be that great of an idea...not sure if that would be expected to boot or not...I would think not. You can test the new CPUs singly, in socket A, and that might tell you something.

FWIW, it seems a little scary at first, but you might think about removing the IHS using a vise (do a search for instructions and video), rather than delicately tightening down the heatsinks while trying to avoid ruining a socket. I went the IHS-removal route, and it actually was quite easy. You just have to be gentle and careful removing it, then carefully scrape-then-sand away the remaining solder. Once you've done it this way, you can toss all the washers and thermal tape aside and not worry about fouled heatsink geometry possibly causing your instability.
 
GREAT NEWS! The screws needed tightening! crazy thing! I thought "since it is dead anyway and I am going back to the two 2.26s..." tighten them and see what happens... Somehow, the screws must have loosened themselves because they were almost off! So, again, thinking dead processors, I went tight, rather than fingers on shaft, hands on the T bar, tight and put the tray back in and... SURPRISE! IT Booted right up! So I went right into the hardware test and weird video card warning but nothing else (turns out that the two memory sticks that I just got are dead --the Apple hardware test only sees 2 gb (entrecheck, somehow, sees 8 but the problems are there)-- and, so far, taking the two sticks out killed the warning. I need to do some thinking, some measuring, and some math (I am wondering how far the heatsink moves down per turn given the pitch of the screws and the pressure on the processor --the delided surface area is gobs smaller) but it seems that two 1.1 washers are not the same (in so far as the method to install and the hazards etc..) than three 0.8 washers. The extended hardware test is now running so I will let everyone know how it turns out but it does see two processors.

I started this whole thing because I could get the 5620s for 6 bucks and I think of it as a learning experience for the 5690s when the time comes. If you are buying the tray for any reason other than the heatsinks (I don't think you can pull the socket and put another one in but if you can, let everyone know, it will be great info) and the fans, I wouldn't.

I never did get to do the two different processors but, theoretically, why wouldn't they boot? The machine only uses the second one when forced (virtual machines can rock when you force it to run (someday I will figure out how to strip the OS to bare bones) with just the one program --I think that the virtual machine "sees" the six cores as being all in one processor so, ironically, things like iMovie or word run better. I don't have the "pro" stuff (Logic Pro or Final Cut) so I don't know if the effect would hold as AFAIK they are built to take advantage of the whole machine. I wonder how Geekbench would do?
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back