MP 1,1-5,1 CPU B temperature spiking, 15-20c hotter than heatsink w/fan at max

[AppleDrank]

I've tried re-applying thermal paste 2x.

See attached files for spikes. CPU B is running 15-20c hotter witch CPU B fan at max, and seeing spikes over 90c at startup and when I tried to run Geekbench!

What should I do?

MacPro 5,1. 2 x 3.33 GHz. Booting Mojave off PCIe SSD. RX580. Boot ROM 141.0.0.0.0. 24GB RAM.
[doublepost=1561152493][/doublepost]

I've tried re-applying thermal paste 2x.

See attached files for spikes. CPU B is running 15-20c hotter witch CPU B fan at max, and seeing spikes over 90c at startup and when I tried to run Geekbench!

What should I do?

MacPro 5,1. 2 x 3.33 GHz. Booting Mojave off PCIe SSD. RX580. Boot ROM 141.0.0.0.0. 24GB RAM.

[doublepost=1561152671][/doublepost]More showing CPU B temperature volatility w/max fan on CPU B - no load - not working on this computer.

 

[Slash-2CPU]

90°C under load with fan at what %?

Don’t even look at idle temps. The sensors are horribly inaccurate below 60-65°C.

I don’t remember if the heat sinks are interchangeable between A & B, but that would be my next try.

I’d then swap CPU’s between sockets.

Screwed up VRM, leaked-out heat pipe in heatsink, bad solder joint under CPU heat spreader lid, bent or warped CPU lid or heatsink base, or (unlikely if all memory channels work) improperly seated CPU.

If same CPU overheats like that with different cooler in different socket, it may be cracked solder between CPU and lid. Reheat lid to soldering temperature to fix it. Or just buy a new CPU.

 

[AppleDrank]

I've tried re-applying thermal paste 2x.

See attached files for spikes. CPU B is running 15-20c hotter witch CPU B fan at max, and seeing spikes over 90c at startup and when I tried to run Geekbench!

What should I do?

MacPro 5,1. 2 x 3.33 GHz. Booting Mojave off PCIe SSD. RX580. Boot ROM 141.0.0.0.0. 24GB RAM.
[doublepost=1561152493][/doublepost]
[doublepost=1561152671][/doublepost]More showing CPU B temperature volatility w/max fan on CPU B - no load - not working on this computer.

View attachment 844320 View attachment 844321

Voltage Sensors:

[doublepost=1561154827][/doublepost]

Voltage Sensors:

OK, I wonder if this is something. There is usually more like a .1 difference in voltage (not .05 as in the post above). However, on my cMP 1,1, which is a 2 x 2.66, the voltages are always the same, or within .01 of each other (e.g., 1.14 vs. 1.15).
[doublepost=1561155469][/doublepost]Hi Slash, thanks for the reply. Per this article on AnandTech, the heatsinks aren't interchangeable.

1. I'm going to need to hold off on switching the processors until I can get some more Noctua cleaning wipes, so I'm going to hold off on heating the CPU lid. Never done that before btw. I did run a Dremel with a felt polishing wheel lightly on the top of CPU B processor to clean of residue, and the temp issue seems a little worse now. Was that bad?

2. What is VRM?

3. What is a "leaked-out heat pipe in heatsink?"

Oh, and do you think it has to do with the ~.1 core voltage difference between the processors?

Thanks!

90°C under load with fan at what %?

Don’t even look at idle temps. The sensors are horribly inaccurate below 60-65°C.

I don’t remember if the heat sinks are interchangeable between A & B, but that would be my next try.

I’d then swap CPU’s between sockets.

Screwed up VRM, leaked-out heat pipe in heatsink, bad solder joint under CPU heat spreader lid, bent or warped CPU lid or heatsink base, or (unlikely if all memory channels work) improperly seated CPU.

If same CPU overheats like that with different cooler in different socket, it may be cracked solder between CPU and lid. Reheat lid to soldering temperature to fix it. Or just buy a new CPU.

 

[Slash-2CPU]

Voltage Sensors:

1. I'm going to need to hold off on switching the processors until I can get some more Noctua cleaning wipes, so I'm going to hold off on heating the CPU lid. Never done that before btw. I did run a Dremel with a felt polishing wheel lightly on the top of CPU B processor to clean of residue, and the temp issue seems a little worse now. Was that bad?

2. What is VRM?

3. What is a "leaked-out heat pipe in heatsink?"

Oh, and do you think it has to do with the ~.1 core voltage difference between the processors?

Thanks!

Do not assume the voltage sensors are telling the truth on a system that's erratically running hot. They're probably correct, but if it's a voltage regulation problem, it can throw the sensor off as well. Even when they are correct, the margin of error on them is (i think) around 0.02V.

Don't draw comparisons between 1,1 and 5,1 CPU voltages. They are very different designs. The CPU's in 4,1/5,1 are almost guaranteed to always run at different voltages unless at true idle or running all cores 100% load.

1. You don't need Noctua cleaning wipes. That's marketing fluff. Paper towel, toilet tissue, or coffee filters and isopropyl alcohol is more than good enough if you're not doing crazy overclocking or chilled cooling. Blow the dust off after cleaning. Polishing usually doesn't help much for a stock CPU on a stock heatsink(again, insane overclock or chilled vapor cooling is different). If you don't clean all the polishing compound off, then it may hurt a little.

2. Voltage regulator module. The logic and transistors that regulate voltage that feeds the CPU. It's all soldered onto the CPU tray, so not a true "module." Module is a dated term I guess. Old Pentium server boards had actual modules that plugged in that regulated the CPU voltage. https://en.wikipedia.org/wiki/Voltage_regulator_module

3. Those tubes you see inside the heatsink are heat pipes. They contain a fluid(usually water, alcohol, or a mix of both) that boils/vaporizes over the hot CPU and rapidly re-condenses in the heatsink fins before being wicked by the rough surface inside the pipe back to the CPU area. If the heat pipe has any hole or crack, the fluid evaporates, and the heat pipe will not cool the CPU at all. It will technically carry a tiny amount of heat by conduction, but practically nothing. I've only ever seen two fail, not Apple ones. A Dell laptop and an old Radeon GPU.

Very high chance if you've never attempted it, or dont have much experience with soldering/hot work/reflow that re-heating the solder under the lid will murder the CPU.

Swap the CPU's. If it still overheats, then the CPU is bad.
If when you swap the other CPU overheats, then it's either the CPU tray failing or a bad heatsink.

If I had to guess, I'd guess bad CPU. It's just a guess based on what I've seen before.

 

[AppleDrank]

Cool, thanks. I will swap them ASAP and post results.

Voltage Sensors:
View attachment 844324
[doublepost=1561154827][/doublepost]

OK, I wonder if this is something. There is usually more like a .1 difference in voltage (not .05 as in the post above). However, on my cMP 1,1, which is a 2 x 2.66, the voltages are always the same, or within .01 of each other (e.g., 1.14 vs. 1.15).
[doublepost=1561155469][/doublepost]Hi Slash, thanks for the reply. Per this article on AnandTech, the heatsinks aren't interchangeable.

1. I'm going to need to hold off on switching the processors until I can get some more Noctua cleaning wipes, so I'm going to hold off on heating the CPU lid. Never done that before btw. I did run a Dremel with a felt polishing wheel lightly on the top of CPU B processor to clean of residue, and the temp issue seems a little worse now. Was that bad?

2. What is VRM?

3. What is a "leaked-out heat pipe in heatsink?"

Oh, and do you think it has to do with the ~.1 core voltage difference between the processors?

Thanks!

 

[MIKX]

AppleDrank

https://forums.macrumors.com/thread...mp-reduction-in-dual-cpu-cmp-4-1-5-1.2179729/

Your Booster 2 fan speed is much too high. Download MacFanControl


EDIT
Get some canned air and a new soft, non-conductive brush. Power down, remove the CPU tray, thoroughly de-dust ( from intake & outtake sides ) the whole area of the CPU tray, especially the CPU heatsink fins.

Use the soft brush to ( gently ) go over the entire CPU tray area then go over the CPU tray a final time

Have a close look at the rear, CPU tray exhaust fan for dust build up - the cheesegrater holes get clogged which greatly hinders hot air extraction- I removed a handful of dust from my 4,1>5,1 in April this year and got immediate positive results.

 

[AppleDrank]

Do not assume the voltage sensors are telling the truth on a system that's erratically running hot. They're probably correct, but if it's a voltage regulation problem, it can throw the sensor off as well. Even when they are correct, the margin of error on them is (i think) around 0.02V.

Don't draw comparisons between 1,1 and 5,1 CPU voltages. They are very different designs. The CPU's in 4,1/5,1 are almost guaranteed to always run at different voltages unless at true idle or running all cores 100% load.

1. You don't need Noctua cleaning wipes. That's marketing fluff. Paper towel, toilet tissue, or coffee filters and isopropyl alcohol is more than good enough if you're not doing crazy overclocking or chilled cooling. Blow the dust off after cleaning. Polishing usually doesn't help much for a stock CPU on a stock heatsink(again, insane overclock or chilled vapor cooling is different). If you don't clean all the polishing compound off, then it may hurt a little.

2. Voltage regulator module. The logic and transistors that regulate voltage that feeds the CPU. It's all soldered onto the CPU tray, so not a true "module." Module is a dated term I guess. Old Pentium server boards had actual modules that plugged in that regulated the CPU voltage. https://en.wikipedia.org/wiki/Voltage_regulator_module

3. Those tubes you see inside the heatsink are heat pipes. They contain a fluid(usually water, alcohol, or a mix of both) that boils/vaporizes over the hot CPU and rapidly re-condenses in the heatsink fins before being wicked by the rough surface inside the pipe back to the CPU area. If the heat pipe has any hole or crack, the fluid evaporates, and the heat pipe will not cool the CPU at all. It will technically carry a tiny amount of heat by conduction, but practically nothing. I've only ever seen two fail, not Apple ones. A Dell laptop and an old Radeon GPU.

Very high chance if you've never attempted it, or dont have much experience with soldering/hot work/reflow that re-heating the solder under the lid will murder the CPU.

Swap the CPU's. If it still overheats, then the CPU is bad.
If when you swap the other CPU overheats, then it's either the CPU tray failing or a bad heatsink.

If I had to guess, I'd guess bad CPU. It's just a guess based on what I've seen before.

Slash,

The machine would not boot after I swapped the CPUs. The white light by the power button blinked repeatedly.

I re-swapped the CPUs, and it powered right up.



MIKX,

Everything is clean. The CPU B booster is running so fast because I set it to maximum in MacsFanControl.

 

[Slash-2CPU]

Slash,

The machine would not boot after I swapped the CPUs. The white light by the power button blinked repeatedly.

I re-swapped the CPUs, and it powered right up.



MIKX,

Everything is clean. The CPU B booster is running so fast because I set it to maximum in MacsFanControl.

Bad CPU. That is really unusual. Those original, or eBay or China?

Either run that CPU until it dies or swap it now if it's unstable or if it's really bothering you that badly. Good chance you could run it like that for years and it would be ok.

Very, very unlikely that it will damage anything when it fails, but you will be down until you replace it then.

 

[AppleDrank]

Thanks, Slash. I was looking forward to your explanation. The fact that it works one way but not other seems very unusual indeed.

I took a few photos of the processors. CPU B has the darker coloration overall. Unfortunately, I bought this machine used, so I don't know the history of the processors. Is China making knock offs?

CPU A looks like these.

Here's a photo of CPU B (the bad one). Unlike CPU A, the pads in the middle are a dark yellow color. I don't know what that is.

Bad CPU. That is really unusual. Those original, or eBay or China?

Either run that CPU until it dies or swap it now if it's unstable or if it's really bothering you that badly. Good chance you could run it like that for years and it would be ok.

Very, very unlikely that it will damage anything when it fails, but you will be down until you replace it then.

 

[Slash-2CPU]

Ones bought from China aren't counterfeit or fake in the traditional sense of the word. In fact, most of them are just fine used chips. It's just a wider, deeper net and there's no telling what they might drag up.

It'll be salvaged from scrapped servers, relabeled engineering sample CPU's, something that failed QC testing but works good enough that some guy marked it destroyed, failed but went home with it in his pocket. I think a lot of it is systems that were 'recycled' and scrubbed of all data here in the US, bought for pennies per system, packed into 40' shipping containers, and stripped for parts in China with minimal testing before resale.

You'll see all sorts of color variations in the packaging substrate(green fiberglass). Doesn't mean much. Same with the pads.

Personally, I'd stick it in and run it til dead. I'd half-consider moving up to a pair of X5690's if they're cheap enough. The 4% speed boost would be 96% placebo though. You'll never be able to notice 4%. Maybe 4% is worth it if you're doing DAYS of video transcoding or a week-long dynamic failure analysis, but if you were you'd already have ordered the X5690's.

For curiosity's sake, put a straight edge on the CPU lid then on all four edges of the pads and see if the CPU is warped or bent. I have no idea how that happens, but I've seen ones that are fully convex. Would explain everything.

 

[AppleDrank]

The 4% speed boost would be 96% placebo though.

Hah!

For curiosity's sake, put a straight edge on the CPU lid then on all four edges of the pads

I am a little confused about what you mean by pads (and wondering if I misused the term). This is what I was referring to re yellow pads. Just want to make sure I should put a straight edge on those?

 

[Slash-2CPU]

Those are capacitors.

The gold pads that contact the pins. Straight edge along the 2 lines I drew.

 

[AppleDrank]

Thanks again, Slash. I'm just checking in. I'm shopping for a replacement 5680, and waiting until it comes in before I pull CPU B.

Though if it is convex, I wonder if I could flatten it somehow...like with a sanding block, or belt sander?

Those are capacitors.

The gold pads that contact the pins. Straight edge along the 2 lines I drew.

 

[Slash-2CPU]

Look up “IHS lapping.” It’s not too complicated. No power tools involved.

The IHS being ever so slightly high in the center is ok. That’s where 99% of the heat transfer occurs. What you don’t want is high corners, low center.

Any gouges or resulting high spots, like divots, in the IHS or the heatsink base are critical to sand flat. I have no idea how it happens, but I’ve seen some that look like they were hit with something really hard with a lot of force.

 

[AppleDrank]

I found that CPU B is concave, as if (as one might expect), the clip that presses down on the middle of the chip has bent the chip, over time. Meanwhile, the heatsink is convex, but with the opposite orientation.

I did some work on it. I didn't get it perfect, but I don't have a good sense of how far is too far.

The results reflect that, with small improvements in CPU B temperature volatility, and temperature relative to the heatsink. CPU B is still 5C hotter than CPU A, at 45C vs 43C, and hotter relative to the heatsink (9C vs. 3C).

However, the actual improvement may be more significant than these results reflect, due to my A/C unit failing, which has caused system ambient to increase from 30C to 40C.

All things considered, I am happy with the results so far. I will check in probably next week after the A/C is replaced and see if there is further improvement. Thanks for your help with this Slash!

Look up “IHS lapping.” It’s not too complicated. No power tools involved.

The IHS being ever so slightly high in the center is ok. That’s where 99% of the heat transfer occurs. What you don’t want is high corners, low center.

Any gouges or resulting high spots, like divots, in the IHS or the heatsink base are critical to sand flat. I have no idea how it happens, but I’ve seen some that look like they were hit with something really hard with a lot of force.

 

[MIKX]

AppleDrank


said . . ..

" CPU B is still 5C hotter than CPU A, at 45C vs 43C ".

These temps are more than acceptable if it is Summer where you live. Very cool in fact.

 

[Slash-2CPU]

Nice work!