Can some of you guys with 2010/2009 Mac Pro's give me your Max load temp readouts...

[JollyJoeJoe]

I have a 12-core Mac Pro which I just upgraded with 2x X5690 3.4GHz
and I am just trying to find out if one of my chips is running at a safe temp or not.

Using Prime95 http://mersenne.org/freesoft/
Blend test (Lots of everything) to hit max temps.

CPU B full load: 63c , CPU A is the culprit at 73c


For temp reading I am using iStat and would appreciate if you could give me the max temp reading it gives you after running the prime test for 10 min.
As I understand this application uses the apple temperature sensors on the heatsinks, so the readings are probably lower than the actual core temperatures, by how much I have no idea?

 

[derbothaus]

My W3680 never get's over 50ºC. Doing anything. How's your thermal paste? Also, you may have higher temps because the the coolers were built for 2x 95W processor and you just fit in 2x 130W processors. Just a guess. I don't know anyone here that has those in their Mac as they are impressively expensive. I hope it kills though.

 

[Transporteur]

As I understand this application uses the apple temperature sensors on the heatsinks, so the readings are probably lower than the actual core temperatures, by how much I have no idea?

iStat reads both the processors core temperature and the heatsink. iStatmenus reads only heatsink.
Harwaremonitor even reads the temperature of each core.

The temperature difference between core and heatsink can easily reach 15° under load. In idle, 5° is the average on my machine.

 

[DanielCoffey]

Assuming your thermal paste is good, one reason for heatsink temp differences is that one is first in the cool air flow and the other is downwind so gets warmer air.

 

[JollyJoeJoe]

My W3680 never get's over 50ºC. Doing anything. How's your thermal paste? Also, you may have higher temps because the the coolers were built for 2x 95W processor and you just fit in 2x 130W processors. Just a guess. I don't know anyone here that has those in their Mac as they are impressively expensive. I hope it kills though.

Have you benched it with Prime95? Because loads like HD video editing, encdoding etc. don't compare. Prime will add a good 10c above the highest load you would usually see with intensive applications.

 

[JollyJoeJoe]

Assuming your thermal paste is good, one reason for heatsink temp differences is that one is first in the cool air flow and the other is downwind so gets warmer air.

That's another thing I am trying to get at, in a 12-core Mac Pro is a 5c+ temperature difference between CPUB and CPUA a standard occurance because one heatsink is deeper and closer to the back of the case (For CPUA which is hotter) and the other closer to the door and front of the case.

 

[Knollwood7E]

MAC PRO X5670 12 core heat report

Attached is a monitor report I made this morning in response to your request. We have a 2 x 2.93 GHz 6 core 16BG MAC PRO. I used the Prime95 default stress test.

Update: I noticed that the upload of the pdf report failed. I will give a few numbers here.

CPU A Heatsink went from 109F to 151F
CPU B Heatsink went from 95F to 154F
Power Supply 1 from 68F to 93F
Power Supply 2 from90F to 102F
CPUCore 1 from 113F to 176F

Please let me know if you need additional info.

 

[Inconsequential]

My W3680 never get's over 50ºC. Doing anything. How's your thermal paste? Also, you may have higher temps because the the coolers were built for 2x 95W processor and you just fit in 2x 130W processors. Just a guess. I don't know anyone here that has those in their Mac as they are impressively expensive. I hope it kills though.

That's not right.

Unless you live in a room thats something like 5C then you won't be able to get a 6-core 3.33GHz CPU to 50C load.

My 2.66Ghz Quad *idles* at ~35C in a 25C room, and loads to 70C with fans ramping up *early*.

Even in deep winter and it was 10C in my room, it loaded to 60C!

 

[beto2k7]

63C and 67C respectively. Ambient 23C. Run time: 2 hours. Mac pro in sig.

 

[derbothaus]

That's not right.

I don't care. My Apt stays pretty cold and my hardwood floor leaks air straight into the intake
But I should probably run the actual test, huh? OK smart guys.

 

[JollyJoeJoe]

Thank you very much for all your replies.

So far from all the info I got I can deduct the following:

* For 95W TDP 1366 Processors Max Load Heatsink temperatures hit approximately 63° - 66° (154F), core temperatures should be a good 10 - 15° higher, hence similar to a PC using a stock Intel heatsink they hit high 70°'s and for some lower 80°'s

* That CPUA does not differ much in temperature to CPUB, at least not by how much I am seeing, being 10-12c° at full load.

In my case I can assume that even though I am using 130W TDP processors, CPUB runs at acceptable temps, whilst CPUA goes higher than is desirable. One important thing I noticed is before I even upgraded the processors, with the old stock X5670 quad cores the CPUA heatsink temp read a good 6c° higher than CPUB at load. So after upgrading this effect amplified! It's also interesting to note than even though it's a 130W TDP processor CPUB Heatsink temps hit a max of 63c° which is roughly the same as 95W TDP chips.

I have now remounted CPUA heatsink and re-applied MX-4 thermal paste, this however did not help. Also as I applied thermal paste properly the first time I didn't expect miracles. One thing I did notice is a small superficial 6mm or so scratch on the polished contact surface of Heatsink A. It's not deep but I can still feel it with the tip of my finger and I am guessing it might be responsible for the higher temps on CPUA before and after my upgrade so the heatsink was scratched when the Mac was built, this is upsetting but not that I can claim warranty on such thing. However I am still not convinced a slightly scratched heatsink surface could hike up the temps so much as thermal paste would have filled those gaps. Who thinks it's because of the scratch or because the CPUA is just hotter running ?

 

[nanofrog]

One thing I did notice is a small superficial 6mm or so scratch on the polished contact surface of Heatsink A.

The surface may have more than that scratch going on (highs/lows on the contact surface, which isn't uncommon as a means of reducing their production costs - applies to any cooler manufacturer).

So you'd probably need to go ahead and lap the cooler's copper surface in question (there's resources on the web that offer a How To - it's not expensive either, but will take you a couple of hours).

 

[Tom Sawyer]

I have a 12-core Mac Pro which I just upgraded with 2x X5690 3.4GHz
and I am just trying to find out if one of my chips is running at a safe temp or not.

Using Prime95 http://mersenne.org/freesoft/
Blend test (Lots of everything) to hit max temps.

CPU B full load: 63c , CPU A is the culprit at 73c


For temp reading I am using iStat and would appreciate if you could give me the max temp reading it gives you after running the prime test for 10 min.
As I understand this application uses the apple temperature sensors on the heatsinks, so the readings are probably lower than the actual core temperatures, by how much I have no idea?

I see very similar numbers with my 2010 MP running dual W5580 procs (3.2ghz quad core 120w chips) under load. It was originally a dual quad 2.4ghz machine. FWIW, I saw mid 60's under full load on the westmere chips.

 

[JollyJoeJoe]

The surface may have more than that scratch going on (highs/lows on the contact surface, which isn't uncommon as a means of reducing their production costs - applies to any cooler manufacturer).

So you'd probably need to go ahead and lap the cooler's copper surface in question (there's resources on the web that offer a How To - it's not expensive either, but will take you a couple of hours).

These highs and lows are they likely to be evident if I placed a metallic ruler on the heatsink contact surface ? Or are they very minute and not perceivable to the eye ?

I have lapped copper heatsinks like the TRUE120 before to a mirrored finish. However I just put some 800, 1000, 1200 grit sand paper on flat hard glass surface and just slid the cooler a long the paper without pressing down much and just letting it do the work with its own weight and got a fairly good result. Would something like this suffice in this case?

The layer to sand off would also have to be very small, I am guessing like 1/4 of a 1mm or so as not reduce tension from the spring loaded screws that are used to screw down the heatsink.


But before I get to lapping anything I suppose the smartest thing to do would be to switch the CPU's around so I can tell if it's the heatsink at fault or the processor that is hot.

 

[nanofrog]

These highs and lows are they likely to be evident if I placed a metallic ruler on the heatsink contact surface ? Or are they very minute and not perceivable to the eye ?

This will depend on how it was made (i.e. in too much of a hurry or not). They always have imperfections as it's considered too expensive to get them within high tolerances.

But there have been surfaces I've seen that were so bad, you could see it with the naked eye, and others that would take a depth gauge to measure the variances.

But if the ruler is truly flat, you could give it a try and see.

I have lapped copper heatsinks like the TRUE120 before to a mirrored finish. However I just put some 800, 1000, 1200 grit sand paper on flat hard glass surface and just slid the cooler a long the paper without pressing down much and just letting it do the work with its own weight and got a fairly good result. Would something like this suffice in this case?

Yes.

You don't need to go find a lapidary machine and use 200k diamond grit on it.

The layer to sand off would also have to be very small, I am guessing like 1/4 of a 1mm or so as not reduce tension from the spring loaded screws that are used to screw down the heatsink.

You shouldn't have to remove much.

But before I get to lapping anything I suppose the smartest thing to do would be to switch the CPU's around so I can tell if it's the heatsink at fault or the processor that is hot.

Not a bad idea.

 

[JollyJoeJoe]

This will depend on how it was made (i.e. in too much of a hurry or not). They always have imperfections as it's considered too expensive to get them within high tolerances.

But there have been surfaces I've seen that were so bad, you could see it with the naked eye, and others that would take a depth gauge to measure the variances.

But if the ruler is truly flat, you could give it a try and see.


Yes.

You don't need to go find a lapidary machine and use 200k diamond grit on it.


You shouldn't have to remove much.


Not a bad idea.

So I lapped the heatsink a bit, at least to the degree that I could see it was more or less flat. Before I lapped it there was some minute irregularity on its surface and placing a credit cards edge on it you could just see it.

I have also swapped the processors around, unfortunately none of this seems to help. I can now conclude that it has nothing to do with the CPU, since either one under Heatsink A runs hot. So its the heatsink not doing its job well or just the positioning of heatsink A inside the system (but this does not appear to effect others much) so it's probably just a crappy heatsink.

What could be wrong with it is anyones guess. Either I didn't lap it properly, definitely made it flatter though or its not making firm contact with the CPU surface for a number of reasons, screws/mounting mechanism is faulty or maybe the fan is not working at full speed, however the RPM readings are fine. One thing that can't be denied is that it gets very hot to touch, heatsink B is just very warm, whilst heatsink A feels like you fry eggs on it. If it would be making poor contact would it get so hot ?

 

[mjsmke]

Ambient temp - 20C-26C
CPU Idle - 29C-32C
CPU Full Load - 62C-80C
CPU Full Load with SMC Fan Control* - 45C-55C

*Exhaust - 1500rpm
*Boosta - 1500rpm
*Intake - 1000rpm

 

[ndraves]

I have a 12-core Mac Pro which I just upgraded with 2x X5690 3.4GHz

Just out of interest what is your Geekbench score on that?

 

[nanofrog]

So I lapped the heatsink a bit, at least to the degree that I could see it was more or less flat. Before I lapped it there was some minute irregularity on its surface and placing a credit cards edge on it you could just see it.

I have also swapped the processors around, unfortunately none of this seems to help. I can now conclude that it has nothing to do with the CPU, since either one under Heatsink A runs hot. So its the heatsink not doing its job well or just the positioning of heatsink A inside the system (but this does not appear to effect others much) so it's probably just a crappy heatsink.

What could be wrong with it is anyones guess. Either I didn't lap it properly, definitely made it flatter though or its not making firm contact with the CPU surface for a number of reasons, screws/mounting mechanism is faulty or maybe the fan is not working at full speed, however the RPM readings are fine. One thing that can't be denied is that it gets very hot to touch, heatsink B is just very warm, whilst heatsink A feels like you fry eggs on it. If it would be making poor contact would it get so hot ?

Just out of curiosity, have you tried both CPU's under the cooler you've just lapped?

I don't expect the CPU's to be that different from each other (~ same temp of either CPU under that particular socket + cooler), but performing this test would rule out CPU differences if there's no notable change.

 

[beto2k7]

Just out of curiosity, have you tried both CPU's under the cooler you've just lapped?

I don't expect the CPU's to be that different from each other (~ same temp of either CPU under that particular socket + cooler), but performing this test would rule out CPU differences if there's no notable change.

He said he did swapped processors

 

[nanofrog]

He said he did swapped processors

I recall that. I'm interested in the temp readings since the cooler's been lapped (guess I should have clarified this). But with the last post, I'm still not sure if the temp's within spec or not (presume the CPU thermal readings are ~ the same when run on the same socket and cooler combination, and it's the location and/or cooler that's the problem).

Ultimately, my hope with the lapping process was to get the CPU to 67C or less under max load (meets Intel's thermal specs @ 100% duty cycle).

 

[JollyJoeJoe]

Just out of curiosity, have you tried both CPU's under the cooler you've just lapped?

I don't expect the CPU's to be that different from each other (~ same temp of either CPU under that particular socket + cooler), but performing this test would rule out CPU differences if there's no notable change.

Hi,

As I said I have swapped the processors around to see no real difference.
If you're asking have I tried using the CPU that was originally under heatsink A before it was lapped with heatsink A after it was lapped, the answer is no.

I just decided to eliminate as many factors as I could in one go. So I removed both heatsinks, swapped around the processors and the known cool running CPU in socketB/HeatsinkB was put into SocketA\Heatsink A and is it turned out the CPU's had nothing to do with it, both run at near same max (67c -72c diode temp) temperatures under Heatsink B and both run hot under heatsink A (diode temperature of 85c, heatsink temp of 73c or so as reported by hardware monitor)

 

[philipma1957]

Hi,

As I said I have swapped the processors around to see no real difference.
If you're asking have I tried using the CPU that was originally under heatsink A before it was lapped with heatsink A after it was lapped, the answer is no.

I just decided to eliminate as many factors as I could in one go. So I removed both heatsinks, swapped around the processors and the known cool running CPU in socketB/HeatsinkB was put into SocketA\Heatsink A and is it turned out the CPU's had nothing to do with it, both run at near same max (67c -72c diode temp) temperatures under Heatsink B and both run hot under heatsink A (diode temperature of 85c, heatsink temp of 73c or so as reported by hardware monitor)

What are you doing for fan control? You know that a full iStat program costs only 15 dollars. You get full and complete control of all the fans in your macpro. Six speeds can be set for each fan. When I added the hex core 3.2 to replace the quad core 2.8 I ran cpu test all 6 cores maxed at 100% for 2 days. My cpu became too warm just over 72c So I set iStat up with speed 6 speed 6 is all fans at full speed. Speed 1 is all fans at lowest setting 2 to 5 are all inbetween. Lots of people always chime in and say smcFancontrol is free. I say for a 3000 to 6000 dollar machine speed the 15 bucks and get a program with much better fan control and general info.

 

[JollyJoeJoe]

What are you doing for fan control? You know that a full iStat program costs only 15 dollars. You get full and complete control of all the fans in your macpro. Six speeds can be set for each fan. When I added the hex core 3.2 to replace the quad core 2.8 I ran cpu test all 6 cores maxed at 100% for 2 days. My cpu became too warm just over 72c So I set iStat up with speed 6 speed 6 is all fans at full speed. Speed 1 is all fans at lowest setting 2 to 5 are all inbetween. Lots of people always chime in and say smcFancontrol is free. I say for a 3000 to 6000 dollar machine speed the 15 bucks and get a program with much better fan control and general info.

I don't think fan control will help in my case, you see my processors are cool when just doing everyday things, the temp will not go over 50 or so.
The fans I would presume spin at max speed (1115rpm ?) when procs get hot. What is your max rpm for BOOSTA/BOOSTB? and what can you speed up the exhaust and intake to?

However introduce prime95 for full load testing and you get to see the temps the machine is likely to hit say after 8 hours of video encoding etc. which I'll be doing a lot of.

 

[philipma1957]

i am not on the pro at this moment. i will give the figures for each setting. in about 20 minutes. I took 3 screen shots first shows how my machine is set for surfing. speed 2222 that which I program look at the fan speeds. second shows how I just am about to change it to speed 6666. third screen shot shows full speed settings of 6666 the machine is really fast fans. I ran this machine at all 6 cores well 12 with the virtual cores at 100 % for 2 days my temps were at 55c with fast settings. I am just saying for 15 bucks this may be the right work around.


In fact I still have it at max speed and temps dropped even more. IN my hex 3.2 core mod I have a lot of test results. done with your needs in mind(heavy cpu users).
I tested the machine with all cores maxed at 100% for a reall long time and showed results. In fact the mod was done to help people that need a better machine for business at a better price.