MP 5,1: What are your CPU temps?

[jls095]

Hi all,

Getting a new MacPro 5,1 with the 3.2 GHz Quad-Core Intel Xeon (W3565) processor. What are you all seeing as far as CPU operating temps? I’m asking just because I’m just generally curious.

I know the standard answer is "don't worry, your Mac will control its temps with fans", but as I said, just curious what you guys are seeing.

Looking forward to using it in my lab for some heavy modeling.

 

[macmaniac75]

for mine, a 4,1 updated to 5,1, with a W3690 at 3,46 GHz, fully loaded (6 SSDs + 4 HDDs, 2 velocity solo x2 cards + a GTX670 + a maxpower esata card) the temperatures reported by istat are :
- cpu in idle (light load) is around 45 °C for an ambient temperature at 33°C
- it can goes up to 85°C max under a full load of 100% for the 6 cores (under handbrake for instance)

 

[hfg]

for mine, a 4,1 updated to 5,1, with a W3690 at 3,46 GHz, fully loaded (6 SSDs + 4 HDDs, 2 velocity solo x2 cards + a GTX670 + a maxpower esata card) the temperatures reported by istat are :
- cpu in idle (light load) is around 45 °C for an ambient temperature at 33°C
- it can goes up to 85°C max under a full load of 100% for the 6 cores (under handbrake for instance)

Curious what your Northbridge temp is showing, and the ambient at the time?

With an ambient of 30C, my Northbridge is 60C, while the CPUs are 34C and 37C while basically idling (dual 3.33GHz 12-cores in a 2012 5,1).

 

[Studio K]

I'm glad you posted this question. I"ve been examining my own CPU temps. I have Mac Pro 4,1 (now 5,1) quad-core w3570 at 3.2 Ghz.

IStat reports for the CPU T diode a temperature of 43C at idle and up to 84C when using Handbrake (all cores at 100%).

The Temperature Monitor app also shows the temps for your cores. Mine go up to 96C under Handbrake. They idle in the upper 40's.

Using SMCFanControl to reset your minimum fan speeds can lower things a bit under load. I boosted Intake, Exhaust, and BOOSTA fans to 1250 rpm's each, and this lowered my maximum Tdiode temp to 74~75C. The cores peaked at 88~90C.

 

[jls095]

Wow, surprisingly hotter than I thought under load. What's it like during browsing/flash activity?

Edit: I see that, theoretically, the W3565 should run cooler under load than the W3570 and W3690 (both mentioned above) since it operates at lower max voltage - 1.225 vs. 1.375.

 

[Macsonic]

Can anyone confirm that higher clock CPUs generate more heat than lower clock CPUs since faster CPUs use more energy? As long as we are comparing the same architecture, same group of models. Like example a 2.26ghz generates less heat than a 3.0ghz?

 

[Studio K]

Can anyone confirm that higher clock CPUs generate more heat than lower clock CPUs since faster CPUs use more energy? As long as we are comparing the same architecture, same group of models. Like example a 2.26ghz generates less heat than a 3.0ghz?

I would like to know this, as well.

Some people (in other threads) post much lower CPU temps than mine. And some have higher clock speeds than I do. I am not clear if they are using fan manipulation to achieve those lower temps. I suspect they must be, because Apple seems to set the RPM's very low to keep things quiet.

Whatever the case may be, it appears that, using Apple's default fan speed settings, the fans kick in at the last possible moment to stabilize the CPU temps so that they don't exceed 95~100C.

I hope that more people share their temps on this thread.

 

[TzunamiOSX]

MacPro 5,1 12x3,33

Idle
36-42°C

all Cores 100% (Cinema4D)
87°C and 68°C

CPU 1 is 10 degrees warmer then CPU 2

 

[Studio K]

Wow, surprisingly hotter than I thought under load. What's it like during browsing/flash activity?

Edit: I see that, theoretically, the W3565 should run cooler under load than the W3570 and W3690 (both mentioned above) since it operates at lower max voltage - 1.225 vs. 1.375.

IT is hot (under stress). But browsing activity does not produce high temperatures. My CPU rarely heats up at all during 'normal' usage. The T diode stays between 42 and 46C. The 4 cores will hover at 46~52C.

Only when stress-testing or 'Handbrak-ing' does it sizzle.

The Tcase (Intel's recommended max temp for normal usage) for most of these CPU's is 67.9C. So my 'normal-use' temps fall well below that.
If you choose to stress the CPU (Handbrake runs it at 100%!!), you can fully expect to exceed the Tcase. Max-ing out all cores is not considered normal usage, I wouldn't think. I've read somewhere that anything at or below 95C is OK (for heavy use).

 

[jls095]

Whatever the case may be, it appears that, using Apple's default fan speed settings, the fans kick in at the last possible moment to stabilize the CPU temps so that they don't exceed 95~100C.

Those temps are concerning given that TCase (temp at the heat spreader on the chip) for processors in this class is just 67.9C.

Edit: Ah, it seems you've beat me to it!

 

[Studio K]

Here is an explanation of CPU temps and Tcase by someone claiming to be an Intel employee (from the Intel Support forum):

"The TCase is a number established by Intel® as a point of reference in order to understand what could be expected as per normal processor temperature.

Anything from the Tcase and below will be the expected temperature of the processor in normal use, anything that doesn't stress out the processor (watching movies, burning CDs, browsing the internet, creating documents, etc.) When the processor is stressed out meaning that you are running heavy processor applications that take control of the CPU or uses it at 100% the temperature will go beyond the Tcase. It can perfectly reach 90 to 95 degrees and the processor will still be OK. The cooling fan is in charge to keep that temperature there.

If the processor temperature reaches 100°C or more it will send a signal to the motherboard to shut down to prevent damage and most likely it won't be possible to turn the computer back in until it cools down."

 

[08380728]

Just dropped a W3690 into my 4,1 upgraded firmware to 5,1 (single CPU). Idle CPU Heat sink temp was 28C encoding with Handbrake as of now hasn't pushed past 64C. Northbridge chip is 59C currently, was about 30C Idle. Boosta fan is currently 852 RPM, PCI Fan 795 RPM all others below 600 RPM. Currently with case open and a single Samsung 840 Pro SSD.

Unfortunately I didn't check temps with the original W3520 CPU, though I did do a Handbrake encode before and will compare times and FPS in the saved logs.

 

[Studio K]

Just dropped a W3690 into my 4,1 upgraded firmware to 5,1 (single CPU). Idle CPU Heat sink temp was 28C encoding with Handbrake as of now hasn't pushed past 64C. Northbridge chip is 59C currently, was about 30C Idle. Boosta fan is currently 852 RPM, PCI Fan 795 RPM all others below 600 RPM. Currently with case open and a single Samsung 8

What is the peak temperature of your CPU diode when running Handbrake for more than 5-10 minutes? The longer you run Handbrake, the more opportunity the CPU has to heat up.
Do your fans ever speed up from the minimum rpm's?

 

[08380728]

What is the peak temperature of your CPU diode when running Handbrake for more than 5-10 minutes? The longer you run Handbrake, the more opportunity the CPU has to heat up.
Do your fans ever speed up from the minimum rpm's?

The details I posted were after 1 hour, of a current encode that lasted 1 hour 30 min encode, so max 64C and didn't go higher than that. I use Arctic Silver Céramique 2 as the thermal compound.

 

[Studio K]

The details I posted were after 1 hour, of a current encode that lasted 1 hour 30 min encode, so max 64C and didn't go higher than that. I use Arctic Silver Céramique 2 as the thermal compound.

64C was the heatsink temperature? What about CPU A Tdiode? That should be higher.
What do you use to monitor? iStat? Temperature Monitor?

Temp Monitor shows core temps, too. those are higher than diode.

I have a quad W3570 in my 4,1. Not quite as fast as your 3690. I think the hex-cores might be a bit cooler than the quads. I read something about them being more 'efficient'.

 

[08380728]

64C was the heatsink temperature? What about CPU A Tdiode? That should be higher.
What do you use to monitor? iStat? Temperature Monitor?

Temp Monitor shows core temps, too. those are higher than diode.

I have a quad W3570 in my 4,1. Not quite as fast as your 3690. I think the hex-cores might be a bit cooler than the quads. I read something about them being more 'efficient'.

Hi again. I used Macs Fan Control. The reading was for the CPU A Heatsink. There is another "CPU A Core from PCECI". I searched quickly but wasn't sure what it was. Currently the machine is idle and this reads 77C, BUT, and big BUT, I was concerned about this then I ran the Handbrake encode and this dropped way way down and the CPU A Heatsink was the one that read the highest at 64C.

I can run the encode again with a specific monitoring tool as per your recommendation and post specifics.

 

[Studio K]

Hi again. I used Macs Fan Control. The reading was for the CPU A Heatsink. There is another "CPU A Core from PCECI". I searched quickly but wasn't sure what it was. Currently the machine is idle and this reads 77C, BUT, and big BUT, I was concerned about this then I ran the Handbrake encode and this dropped way way down and the CPU A Heatsink was the one that read the highest at 64C.

I can run the encode again with a specific monitoring tool as per your recommendation and post specifics.

"CPU A core from PCECI" is how many degrees you are from the temperature you never want to reach (TJMax they call it). It is a countdown, so the higher the number, the better. Many apps assume TJMax to be 100C. And this is to do with cores, not diode.

Macs Fan Control is an app that allows you to control fan speeds based on temperature readings, isn't it? Do your fan speeds increase at a certain preset temperature?

"Temperature Monitor" is a free app that displays lots of readings including cpu core temperatures. Just Google it by name and install if you want to. There's a driver it needs to install to get the core temps, so be sure to let it install that so you can see those.

One way to stress test your CPU to get an idea about how hot your CPU will get and how soon fans come to the rescue is to run the 'yes' command in Terminal. At the Terminal command line, type:

yes > /dev/null &

Enter this 8-10 times to fully max your cpu cores. then watch your core temps in Temperature Monitor.
Can you monitor your fans too? Watch the BOOSTA fan to see if and when it speeds up.

To stop the stress test, type: killall yes

 

[08380728]

"CPU A core from PCECI" is how many degrees you are from the temperature you never want to reach (TJMax they call it). It is a countdown, so the higher the number, the better. Many apps assume TJMax to be 100C. And this is to do with cores, not diode.

Macs Fan Control is an app that allows you to control fan speeds based on temperature readings, isn't it? Do your fan speeds increase at a certain preset temperature?

"Temperature Monitor" is a free app that displays lots of readings including cpu core temperatures. Just Google it by name and install if you want to. There's a driver it needs to install to get the core temps, so be sure to let it install that so you can see those.

One way to stress test your CPU to get an idea about how hot your CPU will get and how soon fans come to the rescue is to run the 'yes' command in Terminal. At the Terminal command line, type:

yes > /dev/null &

Enter this 8-10 times to fully max your CPU cores. then watch your core temps in Temperature Monitor.
Can you monitor your fans too? Watch the BOOSTA fan to see if and when it speeds up.

To stop the stress test, type: killall yes

I entered the command 12 times, since there are 6 cores and Hyper Threads. I let this run for 10 mins then took screen shots. I've linked an image to the CPU Sensor temperature overview. CPU A Temp Diode was 31C before the yes commands were issued. I'd be interested to know the layout of each core under the heat spreader on the CPU die so I could visualise where the highest temperature cores are are placed and figure out if my thermal paste application is affecting them. I used a variation of Artic Silvers' application method, since I couldn't rotate the heatsink I decided to place 3 single lines of paste on the spreader rather than one. I'm wondering if I could lower temps of the higher cores by applying the thermal compound the old skool way, that is spreading a layer over then entire spreader.

Interestingly, the Macs Fan control image shows very little deviation for any of the Fan speeds including BOOSTA. Could be an issue?

 

[Studio K]

It looks like your cpu temperatures peaked in the upper-80's without any intervention by the fans. Fan speeds were at their minimums still.

I wonder if the cpu temps would continue rising if you let the stress test run longer (20-30 minutes).

10 minutes is enough time for my cpu to heat up quite a bit more than yours. My Tdiode reaches 84 and the cores will peak at 97C. This seems to be the trigger for the fans to start speeding up. INTAKE, EXHAUST, and BOOST will intervene at this point to keep the core temp from rising any further. The cores will drop to 95C where they will remain for the rest of the testing. Fan speeds will be running a little faster to maintain this temp.

Your fans must not increase speed because your temperatures never rise enough to trigger them. Either the W3690 cpu stays cooler than the W35xx cpu's due to its design, or the Arctic Silver paste is doing a better job than my Noctua NT-H1 paste.
I'm not certain how much influence thermal grease has on temperatures when stressing a CPU, though. It seems to me that the fans would play a bigger role in keeping temps under control in those circumstances.

If you feel like it, go ahead and run the 'yes' command (or Handbrake) for a longer time. 30 minutes should be enough to get an idea how high the cpu temps will go. Perhaps the six-core takes a longer time to peak.
You could open Activity Monitor while you are testing or handbraking to make sure that you are fully utilizing your cpus.

In case I did not mention it before, my system is a 4,1 flashed to 5,1 with W3570 cpu. Very similar to yours, but with quad 3.2 instead.

 

[08380728]

It looks like your cpu temperatures peaked in the upper-80's without any intervention by the fans. Fan speeds were at their minimums still.

I wonder if the cpu temps would continue rising if you let the stress test run longer (20-30 minutes).

Fired up the machine since it had been off for a few hours. Fan speed didn't increase.
Initiated yes ran for 1 hour: results below:

 

[Studio K]

Fired up the machine since it had been off for a few hours. Fan speed didn't increase.
Initiated yes ran for 1 hour: results below:

Image

I want to thank you for running these tests and posting these shots.

Your temperatures look good compared to what I am experiencing. This lastest test you ran confirms that your cpu temperatures stabilize on their own (80C for diode & 90C average for cores) without need for the fans to get involved.
My temps continue rising until the fans speed up. Your processor simply does not get as hot as mine.

It could simply be the CPU itself that accounts for the difference. We are using different models (mine a 35xx series; yours a 36xx). The specs are slightly different (45 nm vs. 32 nm).

Your ambient air is quite cool, I noticed. Mine is always 32C. Do you live in a cold climate? Or is the room just kept very cold?

 

[08380728]

Your ambient air is quite cool, I noticed. Mine is always 32C. Do you live in a cold climate? Or is the room just kept very cold?

I was going to mention this. The temperature during the day was probably about 17C. Later this evening when I ran this hour long test, it was probably closer to 13C, it's still winter where I am. We'll have to see when it gets into summer, 30C+ haha

 

[Studio K]

I was going to mention this. The temperature during the day was probably about 17C. Later this evening when I ran this hour long test, it was probably closer to 13C, it's still winter where I am. We'll have to see when it gets into summer, 30C+ haha

Later in the evening? It's mid-day here! And hot!

I'm not sure it's worth re-applying your thermal grease. Your temperatures are better than mine.
It would be interesting though to remove the heat sink once more to see how well your application method worked.

 

[08380728]

Later in the evening? It's mid-day here! And hot!

I'm not sure it's worth re-applying your thermal grease. Your temperatures are better than mine.
It would be interesting though to remove the heat sink once more to see how well your application method worked.

yeah, open the image in a new window and look at the time in the file name shown in the address bar of your browser.

Yes i'm interested as well about the thermal compound application and its coverage. Maybe I'll consider that when my 2 x 16GB sticks of RAM arrive from macsales.

 

[adr1974]

It looks like your cpu temperatures peaked in the upper-80's without any intervention by the fans. Fan speeds were at their minimums still.

I wonder if the cpu temps would continue rising if you let the stress test run longer (20-30 minutes).

10 minutes is enough time for my cpu to heat up quite a bit more than yours. My Tdiode reaches 84 and the cores will peak at 97C. This seems to be the trigger for the fans to start speeding up. INTAKE, EXHAUST, and BOOST will intervene at this point to keep the core temp from rising any further. The cores will drop to 95C where they will remain for the rest of the testing. Fan speeds will be running a little faster to maintain this temp.

Your fans must not increase speed because your temperatures never rise enough to trigger them. Either the W3690 cpu stays cooler than the W35xx cpu's due to its design, or the Arctic Silver paste is doing a better job than my Noctua NT-H1 paste.
I'm not certain how much influence thermal grease has on temperatures when stressing a CPU, though. It seems to me that the fans would play a bigger role in keeping temps under control in those circumstances.

If you feel like it, go ahead and run the 'yes' command (or Handbrake) for a longer time. 30 minutes should be enough to get an idea how high the cpu temps will go. Perhaps the six-core takes a longer time to peak.
You could open Activity Monitor while you are testing or handbraking to make sure that you are fully utilizing your cpus.

In case I did not mention it before, my system is a 4,1 flashed to 5,1 with W3570 cpu. Very similar to yours, but with quad 3.2 instead.

Hi StudioK - question. I just upgrade my 2010 with two W5590s. Temps seem decent under load (76C/67C), but fans spin up to around 3000 rpm. What kind of fan speeds are you seeing under load?