Very high CPU temperature readings Mac Pro 8-core 2007

Your temperatures look a little high, but not outrageous. The temperature measured on the heat-generating silicon will always be higher than the temperature measured on the heatsink. That is what the heatsink is supposed to be doing, is carrying away the heat toward the fins, which would measure even cooler.

Thanks for your response. From what I understand, 100º C for any CPU die is generally critical, so 91º C could be OK. But what exactly is measured here?

According to the specs, the maximum operating TCASE is 63º C. Since the sensor for the heat-sink sits right above the copper plate that touches the CPU, it may come close to measuring the TCASE temperature. I guess that would make sense, however, assuming that the heat-sink is always going to be somewhat cooler than the CPU, 58º C would already be too high.

The CPU/Memory intake fans pretty much stay at 500 - 600 rpm when this happens. Even when running a stress test for 30 minutes, the temperature of the heat-sink and CPU core maintains the same 30º C - 35º C difference. My best guess is that the CPU does not displace the heat form the silicon die to the IHS properly. Unfortunately Intel does not make CPU testing tools for OS X to see if the CPU is reducing its speed when reaching that temperature.

FWIW, I did some more testing:

cat > cycle.c <<-'EOF'
#include <stdio.h>
#include <stdint.h>
#include <unistd.h>

uint64_t rdtsc(void)
{
    uint32_t ret0[2];
    __asm__ __volatile__("rdtsc" : "=a"(ret0[0]), "=d"(ret0[1]));
    return ((uint64_t)ret0[1] << 32) | ret0[0];
}

int main(int argc, const char * argv[])
{
    uint64_t startCount = rdtsc();
    sleep(1);
    uint64_t endCount = rdtsc();

    printf("Clocks per second: %llu\n", endCount - startCount);

    return 0;
}
EOF

Compiled and executed above code:

$ cc cycle.c

$ ./a.out
Clocks per second: 2995579161

Then I stressed the CPU:

$ for i in `seq 50`; do { yes > /dev/null & }; done

Here's a picture of the temp readings:

I suppose if there was a temperature problem, I would not read the following when the CPU reaches 90º C:

$ ./a.out
Clocks per second: 2993508810

So according to this, everything seems fine.

Btw, the forum editor is terrible - I guess it's Jive.

Thanks for your response. I was wondering what temperatures other people are reading who have the same type of Mac.

My ambient temperature is nearly the same at 22 C. So I used smfancontrol and set it to 1214 rpm like you have.

I read the following, when the computer is not running much:

Heatsinks: 27/29 C. CPUs 38/40 C.

Using the same settings and running a CPU stress test for several minutes:

Heatsinks: 38/41 C. CPUs 65/70 C.

Anyway, I find running the CPU fan at 1214 rpm a bit too noisy. During the stress test, it only increased up to 1220 though. My fans speeds also do not really increase much regardless. Since the firmware upgrade from Mac Pro 1,1 to Mac Pro 2,1 did not upgrade the SMC, I wonder about it.

Could you provide some temp readings while the CPU is busy? You can download Prime95 (freeware) from http://www.mersenne.org/download/ and run a "Torture" test, which might be easier than my command line example.

I also thought that the 5365 should run cooler. I got both CPU's together for $70 from ebay.

Regarding 10.9, I played with a couple of options, but the following worked:

http://oemden.com/sixty-four-on-thirty-two-sfott/privatetest/

Yes I keep my computers for a while, but what I'm doing here with the Mac Pro is just for fun.

Thanks. I'm getting very similar readings except my CPU, which runs 10 C hotter.