nMP PCIe Switch Diode very hot

[armut]

Hello people,

I have a problem with my nMP.
I recognized via iStat Menus that my PCIe Switch Diode is getting very hot.

To lower this I did NVRAM- and PRAM-Reset but nothing changed.

Do you know how to fix this or what the problem is?

Thanks...

 

[h9826790]

I don't have a nMP, but post the screen capture may help the others to give out better comment.

Also, quite a few of these hardware monitoring softwares (including iStat) are not optimised for Mac Pro (both new and old), they may give out useless reading, or use confusing terms which makes you believe that the computer has problem. In fact, in most case, the computer is indeed perfectly normal.

 

[armut]

Thank you for your replay.

Here is the screenshot.
Now its a little bit lesser than before because the temperature was about 87°.

I hope you can help me. I also have no Apple Protection Plan.

 

[h9826790]

I think now the others with the nMP may able to help you by comparing the numbers from their machine.

Anyway, will the temperature drops if you manually set the fan speed to something like 1200RPM?

 

[armut]

The temperature is going down to 72° when I set the fan speed to 1900 RPM but I have also to say that we have about 35° outside...

 

[DeltaMac]

That's just one temp (out of many), and not something that should worry you...
87 C is hot, but not dangerously so.
Why do you think that's a problem?
Are you having other troubles with your nMP?

 

[armut]

No, I do not have any problems but I recognized it because the PCIe Switch Diode was marked red.

 

[h9826790]

IStat turn the numbers in red by their own logic, there is no real relationship between red number and abnormal.

e.g. On cMP, there is a number called "Core X Relative to ProcHot". In fact, this number is the bigger the better, but iStat will turn it into red when it over certain level, which cause some users believe their very good Mac Pro is running hot.

 

[armut]

Ok, thank you very much for your replies

 

[AppleDroid]

Ok, thank you very much for your replies

Was just going to add my system with d300s also reports much higher Pcie switch temps as well. Seems to be a common thing with stat and these machines.

 

[theSeb]

Mine is currently at 73 Celsius with the fan on at the lowest speed. It seems to stay at around that point on my machine, looking at the history in iStat

 

[SoyCapitanSoyCapitan]

Ambient temperature. If your room is 20 degrees you should expect that chip to run around 68-75c depending on how much load it is under. For every degree the ambient temperature increases the Mac can't remove the same amount of heat, so it has to get hotter. Reduce the room temperature.

 

[filmak]

~74º C, fan @ 789 rpm

 

[Xde]

The PCIe switch is a PEX 8723 from Avago/Broadcom.
This part is not listed on Broadcom, maybe a part for special costumers.
A similar version is PEX 8724.
This chip has a metal case and is located on the new Mac Pro i/o board on the outer side, beside the mains connector.
Picture from „i fix it“: https://d3nevzfk7ii3be.cloudfront.net/igi/XhGBBKpAXc3SZwGp.large

When a Mac Pro is turned on from cold state, the chip has immediately around 60 degrees celsius.
After 10 minutes of idling or Safari usage, the temperature increases up to 75 degrees, more or less,
depending of the room temperature and workload. Can reach easy 85 degrees on some circumstances.

When I look into the PEX 8724 data sheet,
power dissipation is 5 watts and the temperature specs are 0 and 70 degrees celsius.

Related to these specs, this chip seems to be overloaded, with the given data of the temperature apps. All apps are showing same results, I am using MacsFanControl.

I imagine, if this values are real, Apple should have discovered this and would have used a
heatsink on the chips metal surface.

Has anyone some more insight into this temperature thing? I mean some more than "this is okay".
thanks everyone!

 

[hammo123]

Out of interest I checked my PCIe Switch Diode temp using Macs Fan Control, I don't have iStat menus.
I took screenshots of the temps both loading the CPU and then 5 minutes later at idle.
Seems to be pretty steady around 63-66C.

 

[Xde]

Right. But depends on the outside temperature. At 20 degrees your value might stay a while.
Startup temp is close to 60.
With 26 degrees outside temp (no aircon), value will climb over 70 degrees easily, nearly at idle.
Now in summertime, I have 70-75 degrees everyday, only looking some websites.
Maybe it is not an extreme, but compared to CPU/GPU wich are around 38 degrees, it seems
high to me. Just when the datasheet says 70 degrees max.
If I could reach this chip easily, I would have set an heatsink on top.

 

[Xde]

Edit from my post obove...
I thought, it is needed to highly disassemble the Mac Pro, including the flexcables, to reach the chip.

Sadly, the chip is deeply inside the Mac Pro.
But now I imagine, removing only the bottom case part of the Mac Pro could gives access to the chip.
I am not sure, if the circular logic board needs to be removed, but I imagine, the diameter of the board is smaller,
so that it may be possible to reach the location.

There is only little space in the gap between the i/o board and the connectors rear panel.
Additionally, just aside of the mains connector, there is the mains cable passing from the connector to the power supply.
So, there is not much room to add a heatsink, maybe a small one, with 5mm height or less or simply a slightly larger heat plate than the chip surface.
I am thinking about an "L" form heatplate, positioned below the mains connector while the L angle profile would touch
the chip surpace. Glewed then with isolation and coating on the PCB.

Only to think about. Maybe at a later moment, I will discover this more.

 

[Lennyvalentin]

Only to think about. Maybe at a later moment, I will discover this more.

Don't obsess over it. These software temperature readers aren't super reliable or accurate, nor is the hardware the software is sampling either (the diodes vary in sensitivity and need calibration for reliable temp readings; this may or may not have been done, who knows.)

If there was an issue with these chips it would have manifested itself by now. The thread is several years old, and so far nothing, from what I can tell. So just - err - chill...

 

[Xde]

I hope you're right. What we see is a chip that seems running out of specification.
If this values are real or not, nobody knows.

But I am not so confident, that Apple does everything as best they can do.
You can see that many people have problems with their nMP, and Apple is unable to service.
They have an unofficial repair program for the GPUs (why not official if here problems exists).
Nevertheless, people with these kind of GPU problems are left in the rain.

And if I could easily reach this chip, I would set a small heatsink on top.

 

[Xde]

Update:
modification

I must say, I had no problems at all with my nMP, but the high temps of the PCIe switch made me uncertain.
So I did this as an act of prophylaxis.
The max. working temp value from the Avago/Broadcom factory is set to 70 degrees Celsius.
In summertime, my PCIe switch went up to 70-75 degrees without load.

Okay, I disassembled all and figured out the possibility to set an heatsink up.

Unfortunately, there is very little space for a heatsink. Furthermore, there is the internal AC mains cable
coming from the mains filter, passing partly above the PCIe switch, decreases space here.

Its only possible to use a 5 mm hight heatsink, thats not much.
So, I got a 19 x 19 mm heatsink, which was slightly to wide and I cut off 1 rib from each side.
It has now around 13 x 19 mm, 5 mm hight.

To be able to remove the heatsink later for any reason, I did not use 2 components epoxy heat-conduction glue.
Instead I took a silicone, known as heatsink plaster.
The efficiency is not that good, but I wanted a chance to remove it anytime later.
(Btw: heatsink plaster also is not easy to remove. Using a razor blade, the gap is usually less then the razor blade thickness)

Finally...,
the modification does a reduction of about 5 degrees C. (See table)
Personally I am unsure, if it was worth, I was in hope to get around 10, but the limited space makes that dificould.
Maybe a heatsink with some more lenght (example 30 mm) and a better thermal glue could do that.

Now, while having room temps around 24 degrees, chip temp keeps everytime below 70 degrees.
The room temps were measured with an inexpensive instrument from the bay. So a tolerance is probably. PCIe switch temps are measured using MacsFanControl.

(There is another chip, the PCH "platform controller hub", which shows also slightly elevated temps, but not that
high than the PCIe switch. The PCH is located on top of the circular interconnect board.
With another little heatsink, temp went also down 4-6 degrees)

Now fotos and temp values, giving the room temp and the PCIe temp values before and after setting up the heatsink.
---------------------

After all, I would not recommend to anyone to disassemble the nMP.
This is really precision engineering, and the probability to damage something is everything else than low!
The mezzanine connectors are also something special, and the feeling was really, "will the nMP run again after all"?.

FYI

Table shows room temp on the left. Following to the right by the PCIe switch temp after x minutes,
at least after additional 5 minutes HD video playback.
All temps in idle state until video playback and from a cold startup.
Values above the line are without heatsink, temps below the line are with heatsink added.

 

[filmak]

Update:
modification

I must say, I had no problems at all with my nMP, but the high temps of the PCIe switch made me uncertain.
So I did this as an act of prophylaxis.
The max. working temp value from the Avago/Broadcom factory is set to 70 degrees Celsius.
In summertime, my PCIe switch went up to 70-75 degrees without load.

Okay, I disassembled all and figured out the possibility to set an heatsink up.

Unfortunately, there is very little space for a heatsink. Furthermore, there is the internal AC mains cable
coming from the mains filter, passing partly above the PCIe switch, decreases space here.

Its only possible to use a 5 mm hight heatsink, thats not much.
So, I got a 19 x 19 mm heatsink, which was slightly to wide and I cut off 1 rib from each side.
It has now around 13 x 19 mm, 5 mm hight.

To be able to remove the heatsink later for any reason, I did not use 2 components epoxy heat-conduction glue.
Instead I took a silicone, known as heatsink plaster.
The efficiency is not that good, but I wanted a chance to remove it anytime later.
(Btw: heatsink plaster also is not easy to remove. Using a razor blade, the gap is usually less then the razor blade thickness)

Finally...,
the modification does a reduction of about 5 degrees C. (See table)
Personally I am unsure, if it was worth, I was in hope to get around 10, but the limited space makes that dificould.
Maybe a heatsink with some more lenght (example 30 mm) and a better thermal glue could do that.

Now, while having room temps around 24 degrees, chip temp keeps everytime below 70 degrees.
The room temps were measured with an inexpensive instrument from the bay. So a tolerance is probably. PCIe switch temps are measured using MacsFanControl.

(There is another chip, the PCH "platform controller hub", which shows also slightly elevated temps, but not that
high than the PCIe switch. The PCH is located on top of the circular interconnect board.
With another little heatsink, temp went also down 4-6 degrees)

Now fotos and temp values, giving the room temp and the PCIe temp values before and after setting up the heatsink.
---------------------

After all, I would not recommend to anyone to disassemble the nMP.
This is really precision engineering, and the probability to damage something is everything else than low!
The mezzanine connectors are also something special, and the feeling was really, "will the nMP run again after all"?.

FYI

Table shows room temp on the left. Following to the right by the PCIe switch temp after x minutes,
at least after additional 5 minutes HD video playback.
All temps in idle state until video playback and from a cold startup.
Values above the line are without heatsink, temps below the line are with heatsink added.

Thanks for sharing.

 

[Selsk]

I also noticed my PCIe switch running hot. So usually when doing general work on my nMp such as light photoshop stuff...I run Macs Fan Control app and base the controll of the PCIe switch diode. This keeps everything cool. Even in a hot office, my fan only ramps up to 1300-1400rpm and keeps the switch at 69-70 degrees.

 

[th0masp]

I run Macs Fan Control app and base the controll of the PCIe switch diode. This keeps everything cool. Even in a hot office, my fan only ramps up to 1300-1400rpm and keeps the switch at 69-70 degrees.

Same setup here. I did this right after receiving my machine and hearing about the potential issue here in the forum. Just like another poster further up I do not exactly trust Apple to always 'know best' and those default temp readings combined with the lazy fan profile did look like a problematic combination to me.

I mean I do appreciate that the thing is near inaudible out of the box - but it better not cook itself during use.

 

[vett93]

I found this old thread. I have the iStat Menu 6.4. The PCIe switch diode reads about 140 degrees F (which is about 60 degrees C) when I am doing casual work. The room temperature is about 72 degrees F (22 degrees C).

 

[Henninges]

FYI : i see a massive increase of the PCIe Switch temperatur while using an egpu on my 6,1...never decrease under 70°c even in idle.