MP 1,1-5,1 IOH Diode?

[paulcons]

Had some flakiness (boot drive SSD array disappeared!), thought it might be heat related, posted Macs Fan screen shot, engineer buddy said get my machine to a shop fast, the Northbridge was getting too hot (about 150ºF/66ºC). No broken spring clips on the heatsink, everything cleaned out, new thermals... Northbridge NOW runs between 95ºF/35ºC and 110ºF/43ºC (today's ambient here is going to get at least 5º hotter soon). I AM concerned about the IOH Diode though. They put a temp sensor on it for a reason I would guess. I've seen similar cMPs of my vintage running way cooler.

No fans, the diode runs ~190ºF/88ºC, with fans 168ºF/76ºC.

What temp range SHOULD this item live in? What temp signals a critical condition? Should I have it replaced? When should I start freaking out <lol>? And can someone explain what it does in there?

 

[Slash-2CPU]

Easy fix. Common, known issue. The rivets/pins that hold the IOH heatsink on break.

https://forums.macrumors.com/threads/mac-pro-heatsink-northbridge-rivet-loud-fan.1755982/

 

[amedias]

This bit:

the Northbridge was getting too hot (about 150ºF/66ºC). ... Northbridge NOW runs between 95ºF/35ºC and 110ºF/43ºC

seems to contradict this bit?

No fans, the diode runs ~190ºF/88ºC, with fans 168ºF/76ºC.

Or was the first set of temps with the fans at max? I can't for the life of me work out what temps you're actually seeing and under what kind of use. What temps does you NB run at under normal use?

Either way that looks fine, in a dual processor 4,1/5,1 70-80ºC is 'normal' with 80 being at the warm end. For a single CPU 4,1/5,1 temps are often 10 deg cooler but as always it depends on the load on the machine, how long it's been running under that load, and what the ambient temps are. It's been covered in a number of other threads but can be hard to find the info.

Some people like them to be as cool as possible and actively go out of their way to run fans faster or add additional fans, others are happy to let the default Apple SMC profile manage it, and those that don't use temp monitors are obviously just letting the SMC handle it (and probably running 70-80ºC). If your buddy was panicking at 66ºC he seems a little easy to spook ;-)

FWIW, I've got 6 4,1/5,1 here in a mix of dual CPU and single CPU and the only time I ever see a NB temp < 40ºC is in the first 2-3 mins after booting them in a cold room. They're all > 50ºC within minutes and under normal use between 60ºC and 85ºC depending on the machine, use and ambient room temp, and that's always with fans at default idle speeds.

 

[paulcons]

Contradict? NB WAS running a bit hot, someone told me too hot, he mentioned the issues with the plastic clips holding it's heatsink on (if that WAS an issue, and I did nothing, I COULD have fried the NB) had it looked at, clips were fine, it got cleaned out and new thermals applied, came back and the NB was 10ºC cooler. The NB is NOT the issue that concerns me.

As for the SMC, that MAY prevent a total meltdown, BUT I had good evidence that the NB was NOT performing its job exactly correctly. Running around 66º it failed to "see" my PCI card SSD array. The chip MAY have been within its limits, but one of it's jobs was not getting done. THAT is what started this whole thing off. AND another factor was that for the few weeks I have been setting up a win10 drive, so I have been hitting the boot picker a LOT more than I ever did previously.

It's a coolish morning here, ambient around my computer is 27ºC (ugh, I am SO used to F, but many of youse talk C!) and my NB is running 34.5º. Fans are running (1200ish In & Ex, 1700 Boost), but inaudible. Life is good EXCEPT that diode is running WAY WAY hotter than anything else in the machine. I am TRYING to learn a lot more about it to assess how much attention I need to pay to it. i.e. what is its function, what are it's thermal limits, at what point is it nearing its upper limit on heat.. stuff like that.

 

[amedias]

Contradict? NB WAS running a bit hot, someone told me too hot, he mentioned the issues with the plastic clips holding it's heatsink on (if that WAS an issue, and I did nothing, I COULD have fried the NB) had it looked at, clips were fine, it got cleaned out and new thermals applied, came back and the NB was 10ºC cooler. The NB is NOT the issue that concerns me.

It was the bit where you went from saying after a re-paste you NB was down to 43, to then saying...."with fans 168ºF/76ºC." that got me confused. Broken clips would likely see you well north of 80ºC under even moderate load, but I've seen machines running with broken clips and NB at 100ºC+ before, they'd still been running fine and stable, they'd likely been that way for months and months as the broken clips were only picked up during other work.

As for the SMC, that MAY prevent a total meltdown, BUT I had good evidence that the NB was NOT performing its job exactly correctly. Running around 66º it failed to "see" my PCI card SSD array. The chip MAY have been within its limits, but one of it's jobs was not getting done. THAT is what started this whole thing off. AND another factor was that for the few weeks I have been setting up a win10 drive, so I have been hitting the boot picker a LOT more than I ever did previously.

I personally don't think your disappearing SSD array is due to it running at 66ºC as that's an entirely normal temp that thousands of us run at (or above) 24x7.
I'm not suggesting your issues aren't real, just that they are unlikely to be caused by your NB temps alone as they're not overly high. There's likely other factors or other components that are causing the issue.

It's a coolish morning here.... my NB is running 34.5º.

That's considerably cooler than most people see for a NB temp.

Fans are running (1200ish In & Ex, 1700 Boost), but inaudible.

Did you manually increase them or is that what the SMC is running them at?
As a point of reference, also 27ºC ambient here and all my fans are running at default/min

PSU - 500rpm
PCI - 800rpm
Exhaust - 600rpm
Intake - 600rpm
Boost (A+B) - 917rpm (yep, that seems to be the min for this machine)

Temps are all in check, NB is ~68-78 depending on load.

Life is good EXCEPT that diode is running WAY WAY hotter than anything else in the machine.

Now I'm confused again, you just said your NB is running at 34.5, but that "diode is running WAY WAY hotter than anything else in the machine" is every other temp in your machine below 34.5ºC ? Are you actually talking about two different readings here, maybe the NB Heatsink and NB Diode?

Lower temps are always a laudable goal, but I think people sometimes get too hung up on them, it's worth remembering that these machines were designed to run 'warm but within spec' and that's OK, anything within spec and stable is acceptable, you might want cooler but in the grand scheme of things it won't make much difference, if things aren't actually overheating then it's a mostly theoretical longevity argument. After all, the vast majority of Mac Pros that are out there have been running for about a decade with nothing other than the SMC defaults, it's only us geeks who like to tinker who ever both actually checking on temps, those thousands of machines have done OK on the default SMC so far, and they're still here...

Obviously you need to be vigilant for actual overheating issues, and broken clips etc, but beyond that 'running warm' is OK.

 

[thornslack]

The NB is generally the hottest thing in the machine. This is normal. Your temps are normal. The NB has no bearing on your SSD array being recognized. It’s a processor interconnect and lots of data pumps through it but it’s not really doing anything. You may want to read up a bit, your assumptions seem a little off. But you’re in the right place to get help, good luck.

 

[Ludacrisvp]

Keep in mind the standard of the world when computer temps are discussed is C not F.
Sure cooler is always better for longevity and potential improved reliability of "flaky" devices.

 

[paulcons]

Initially I had THOUGHT the IOH Diode temp WAS the NB chip, was told by an engineer the NB chips temperature comes from a thermistor that is part of it. Here it is called the Heatsink, is that not correct (seems logical to read the temp of the heatsink to see where the chip is at)? Temps on the "Heatsink/NB" definitely came down with the cleaning and new thermal. The IOH Diode is my concern, it actually went a bit UP in temp when I put my cleaned up tray back in the machine. Off the bat it was reading around 88º. I have played around with MFC to have fans adjusted based on the diode (In, Ex, Boost) temp and under normal load, 30º ambient EVERYTHING is 32-38º except the diode at 76º. I CAN hear the fans but they aren't obnoxious. Under heavy load (MMO game) the fans DO rev up and are pretty audible. I am trying to better understand this item... which is why I asked the questions about it at the top. I am trying to focus on the Diode... if it IS designed to run up to 80º normal with a limit around 100º then I get MUCH less concerned.

I have made inquiries about this part... thinking it not something like the NB chip itself which MAY no longer be made, it may be a standard part and can be replaced by itself, IF NEEDED.

 

[thornslack]

The diode is the NB. It has its own heatsink. Both have thermal sensors. The diode is always warmer than it’s associated heat sink because you can’t have perfect thermal transmission, generally to the tune of about ~5-10C. Many people replace their processors in the cMP but often the NB gets neglected. As the thermal paste ages it is less effective, and apples plastic rivets have a habit of shearing off at the bottom. This ceases to hold the NBs heatsink down properly. 88C definitely is sitting towards the top end of what you would want to see and could indicate bad rivets, thermal paste, heatsink positioning or some combination of the above.

Also you can’t replace the NB to my knowledge. It’s a soldered chip on the processor daughter board. Replacing it would require a super specialist and likely cost more than just replacing the whole board.

 

[paulcons]

In point of fact I JUST had it looked at by a tech. No broken plastic at all, fresh thermals applied, all dust in the big heat sink blown out. I am just trying to learn about this stuff, no idea how a IC like the Northbridge is called a diode! AND in some of the sensor recordings a triode. I do have a decent idea about heat transfer, but (right now) 76.5º to 36.9º? That DOES seem like a rather huge drop. OR that de-dusting and replace 10 year old stick apple paste with a brand new pad would make its temp actually go up?

I AM running fans now to try and keep it under 80º... no idea when ambient goes up and I have it under heavier load. Funny, the tech did say that it was soldered and thus not replaceable... funny, I come from the era where repairs WERE commonly done on the board level... hell, I used to do HeathKits way back in the days of my youth! I DID snag a replacement (slower CPU) which should be here next week...

 

[amedias]

IOH = input/output Hub = Northbridge

The diode is a temp sensor within the IOH package so when you see IOH diode temp it’s reporting (as near as can be) the internal temp of the IOH

The IOH heat sink temp is an additional external sensor attached to the heatsink, this is an indicative temp of how hot the heatsink is but is obviously subject to variation in both actual physical location and how well it’s attached.

The IOH diode temp is the one that normally operates in the 50-80 def C range depending on single Vs dual CPU, load on the machine and environmental variables such as ambient temp, fan speed and how much fluff is in the heatsinks.

There’s always a difference between them, it’s to be expected but you only really need to be co Verne’s when the temperatures are VERy different as that can indicate poor thermal contact between the IOH and it’s heatsink, ie: broken clips, bad paste or some other issue.

88 deg C for the NB is on the warm side and I would be investigating at that temp but not panicking yet...

(I now understand what you were saying, it was your referencing of the sensors that was confusing)
[doublepost=1562091855][/doublepost]I’m out right now but if I get home at a sensible hour I’ll post some screen shots from a bunch of single and duals I have to show you some indicative temps and differences, but there’s already a load around the forum from various users if you can dig them out

 

[paulcons]

Ah, things are starting to make more sense now. My friend, who I must have misunderstood, did say that he thought the diode was "shot." I was thinking of the kind of diode that passed current in only one direction, so why does my machine work? Totally makes sense to me, it being a temp sensor... the guy who did the work even showed me pictures of the before and after... I saw the junk he cleaned off the NB chip itself and the little blue thermal pad he put on it. He said the heatsink clips were fine. He also said he saw some brownish stuff that I should not have to worry about. It MAY very well be that the diode DID get damaged, that it kinda works as it returns a temp but that temp is way incorrect. Do you know how it is positioned compared to the heatsink? I scoured some threads but only have a good idea what the heatsink itself looks like. I now HAVE a T15 torx and should get some Noctua2 Friday... I need to lay eyes on things!

So my operating theory is that the diode got damaged enough to still kinda work, but give erroneous readings. How can there be almost 40º difference between it and the heatsink? BUT that also means I do NOT know exactly what the chip is running at. Next week I get a tray I bought on ebay (I saw a killer price and jumped with a very well rated vendor... HE has another piece up now for more than I paid). Think I have a decent general idea about application, but I have seen some videos that I THOUGHT put way too much on the chip... I get that it's supposed to end up being a THIN coating.

 

[thornslack]

I would think that using a pad instead of thermal compound is part of the problem. You shouldn’t see a delta of 40c. Also, I don’t think you can remove the NB heatsink and still use the same rivets. So the pad is maybe junk and the heatsink is maybe no longer attached properly. Is it just sitting on a thermal pad? Did your repair guy know how to work on these computers?

Also you always know the NB temp from its sensor reading. The difference between its temp and its heatsink is indicative of poor thermal transfer between the chip and its heatsink (read: thermal pad, broken rivets).

 

[paulcons]

Or that the diode is simply not reading the temp correctly... which apparently CAN be the case if it got overheated enough. Yes, it is possible the tech guy totally screwed me over and did a very sloppy job, but I tend to not think that, I questioned him pretty well and he was careful enough actually take before and after pictures of his work. Sometime next week I will have the opportunity to actually see if there is any issues like you describe may be the case... after I get the replacement tray I ordered.

I WOULD like to see a picture of this diode (and something like a part number/manufacturer), just to be able to have some comparison to my tray and the one I'll get next week as well as looking at it's specs. Easy to find shots of the NB heatsink itself, but I'd still like to know where this item is located.

 

[tsialex]

Or that the diode is simply not reading the temp correctly... which apparently CAN be the case if it got overheated enough. Yes, it is possible the tech guy totally screwed me over and did a very sloppy job, but I tend to not think that, I questioned him pretty well and he was careful enough actually take before and after pictures of his work. Sometime next week I will have the opportunity to actually see if there is any issues like you describe may be the case... after I get the replacement tray I ordered.

I WOULD like to see a picture of this diode (and something like a part number/manufacturer), just to be able to have some comparison to my tray and the one I'll get next week as well as looking at it's specs. Easy to find shots of the NB heatsink itself, but I'd still like to know where this item is located.

It's on the die of the northbridge and not a external sensor.

 

[tsialex]

 

[paulcons]

It's on the die of the northbridge and not a external sensor.

Ah... definitely not replaceable then. Here's a snapshot I just took. I'd find it hard to believe the tech would leave the heatsink virtually unattached to the NB. I don't want to pop off the heatsink until I get some Noctua2... and my second tray. Still seems to me the most likely thing is the diode is damaged and not reading correctly.

 

[tsialex]

Ah... definitely not replaceable then. Here's a snapshot I just took. I'd find it hard to believe the tech would leave the heatsink virtually unattached to the NB. I don't want to pop off the heatsink until I get some Noctua2... and my second tray. Still seems to me the most likely thing is the diode is damaged and not reading correctly.

View attachment 846487

If the heatsink was unattached, it would be around 100ºC.

The usual idle temperatures for the internal sensor is around 73º to 77ºC for a place with 30º to 35ºC ambient temperature. It's winter down here, so my idle temp is down to 66º to 71ºC.

 

[paulcons]

If was unattached, it would be around 100ºC. The usual idle temperatures for the internal sensor is around 74~77ºC for a place with 30 to 35ºC ambient temperature.

Really? Even with such a disparity between it and the heatsink reading? This is last night, I shut down MFC for ~10 minutes or so to get back to default fan speeds and... same 40º delta. AND clearly needing some fan beyond default.

 

[tsialex]

Really? Even with such a disparity between it and the heatsink reading? This is last night, I shut down MFC for ~10 minutes or so to get back to default fan speeds and... same 40º delta. AND clearly needing some fan beyond default.

View attachment 846494

You probably still have some problem, like a concave chipset or a convex heatsink or some bad thermal conductivity somewhere, but you won't get the idle temperature below 70º something without active cooling the heatsink.

The normal idle temperature for a warm climate is around 75ºC idle.

 

[paulcons]

Thanks so much for chipping in, I am a nooge to know the whys and how comes! Looks like my fan profile is keeping it in check until I can see what the replacement does (under a game load, fans spin faster (louder) but keep that diode around 75-6º).

 

[thornslack]

My money is on the thermal pad and possibly the heatsink not being clamped down properly. Pads are not as effective as actual thermal compound.

 

[h9826790]

I also agree that the problem is "thermal pad". Thermal paste should be used. And all the temperature reading points to poor thermal transfer but not sensor's error.

If OP really want / prefer thermal pad, only the graphite pad should be use, but not those normal thermal pad.

 

[paulcons]

Point taken... I have seen a lot of reviews that good quality thermal pads are only a few degrees below paste in terms of heat dispersion. So let’s say paste brings the diode down 10C, that still leaves a rather large delta between it and the heat sink. You’d think that such a great heat generator would affect other items in the area as well. At least I now have a load of real world readings as to exactly how my machine is running. In any case, soon as I get my replacement board I can open this stuff up and see directly for myself.

 

[h9826790]

Point taken... I have seen a lot of reviews that good quality thermal pads are only a few degrees below paste in terms of heat dispersion. So let’s say paste brings the diode down 10C, that still leaves a rather large delta between it and the heat sink. You’d think that such a great heat generator would affect other items in the area as well. At least I now have a load of real world readings as to exactly how my machine is running. In any case, soon as I get my replacement board I can open this stuff up and see directly for myself.

I don't know which review you are talking about. If you use thermal paste to replace thermal pad, may be only few degree improvement. Because "use thermal pad originally" already imply "the thermal demand is low". In lots of case, even the thermal pad is completely missing (e.g. for a VRM), the setup can still work, but just with very high temperature (in fact, many graphic card or motherboard has no cooling for VRM).

e.g. (Case 1) No thermal pad = 70C, with thermal pad (and cooler) = 50C, with thermal paste (and cooler) = 40C

No matter how good the thermal paste is. It can't give you too much improvement, because there is not much room to improve at the very beginning (e.g. assume OAT = 30C, then there are only 40C difference between the VRM and the air)

However, if something need thermal paste as the very beginning (e.g. CPU), you can't just replace the thermal paste by thermal pad. The difference won't be "just few degrees", it will be a lot worst. There are much more heat need to be transfer, and even the conductivity of the thermal pad isn't that bad, the thickness is deadly.

e.g. (Case 2) with thermal paste =55C, with thermal pad = 85C, no thermal pad (also imply no cooler) = 130C

The difference will be much much bigger than Case 1. Because the same OAT = 30, there is 100C difference between the chip and the air temperature. That performance difference between thermal pad and thermal paste is not a fixed "degree C", but a function of "how much heat need to transfer". The larger the demand, the bigger the difference.