2009-2012 Mac Pro CPU core temps

[cfshelor]

Love this little app and its getting better all the time. My issue at this point is the CPU A and CPU B "core from PCECI" temps do not agree with the values displayed by Temperature Monitor. This is driving the BoostA and BoostB fans to high speeds.

Mac Pro - Early 2009 - 2 x 2.66 GHz Quad-Core - OS X 10.8.5

You can see that most sensors match, but not the CPU core temps:
Sensor MacsFanControl TempMonitor
Ambient 28 28
CPU A heatsink 30 29
CPU A core 57 32
CPU B heatsink 27 27
CPU B core 61 31
Northbridge chip 51 51
Northbrdige heatsink 41 40
PCIe Slot 2 59 59

[Steven]

Hi,

Please post technical information from the About dialog

the question is: who said that Temperature Monitor is correct? There's no official documentation for sensors description. But the values

<key>TCAC</key>
<string>CPU A core from PCECI</string>
<key>TCBC</key>
<string>CPU B core from PCECI</string>

Were found in the official document.

[cfshelor]

Yes, as to who is right… The behavior of the system combined with logic and common sense seems to favor the temperature monitor… But it wouldn't be the first time I've been wrong!

Here is technical info from the about dialog as requested.

[Steven]

Thank you. We need to have tech info for every mac in order to have MFC working correctly on any hardware.

[Steven]

We have just received another report from a customer via e-mail:

Running on a 2010 Mac Pro, you show CPU A core from PCECI as 68-70°C at idle. When at 100% CPU load, it shows 38-43°C. What you're reading is the delta-T reported by the DTS from #PROTCHOT activating; that is degrees below thermal throttling threshold. On most of these CPU's, it's 100°C.

At idle then, my CPU is 70°C from Tj. Max of 100°C, so it's 30°C. Under load, it's 40°C from Tj. Max of 100°C, so it's 60°C.

You currently have it inverted, so if you've got fan speeds linked to CPU core temp, the fans will in fact slow down as temps go up.

-Keith

[AGUSTIN]

Same for MacPro 3.1
Core Temp may not differ much from the heat sink or diode temperature, otherwise we would all have bad thermal paste between the CPU and the heatsink.

[Steven]

AGUSTIN

Can you test temperature with 100% CPU load? CPU A(b) Core temperatures raise or fall down?

Please run this test as well:
http://forums.crystalidea.com/viewtopic.php?f=11&t=1073

[AGUSTIN]

Hi Steven,

the core temps low down at 100% cpu load.

[AGUSTIN]

By the other hand, the Northbridge Chip temperature is not working for me. With high IO activity, the IOH Heatsink (Northbrige heatsink) temperature rises, but the Northbridge Chip Temperature not.

[Steven]

heatsink doesn't look high as well at 100% load.

Do you really think that actual temperatures for CPU A(B) Cores should be (85C - current value)?

85C is Tj max for your CPU.

[AGUSTIN]

Hi,

I don't understand your question. As CPU load rises, the heat sink temperature rises but displayed Core Temperature fall down.
Maybe:
RealCoreTemp=Tjmax-DisplayedCoreTemp

[Steven]

AGUSTIN
That's exactly what I mean. TjMax for your CPU is 85C, then

RealCoreTemp=85-DisplayedCoreTemp
does that look right? Can you check it with Temperature Monitor Lite or some other software?

[Steven]

Fixed in 1.2 version. Available as beta at the moment