MP 1,1-5,1 CPU ID

[fatespawn]

4,1 delidded 3.46 CPUs....

I just repasted my CPUs and I seemed to have done a good job... and a crappy job. This is under a Handbrake load. CPU 2 is 66c and CPU1 is 84c when everything stabilizes.

I realize that Handbrake won't utilize all 24 cores so CPU 2 may be getting the bigger hit, but the difference seems pretty large. Thoughts? Also, is there a way to identify which CPU cores are being used? If I repaste (again) I want to make sure that "CPU1" is CPU-A... the rear CPU.

 

[tsialex]

CPU socket A (or CPU 1) is the rear one. CPU socket B is the front one and is always cooler than CPU A, since the CPU A receive the already warmer air from the CPU B heatsink. It's normal CPU A to be warmer than CPU B, usually around 10 to 12ºC when in full load.

You can confirm if you botched something installing the current CPU A to the socket B and then testing again to see if anything changes.

 

[fatespawn]

CPU socket A (or CPU 1) is the rear one. CPU socket B is the front one and is always cooler than CPU A, since the CPU A receive the already warmer air from the CPU B heatsink. It's normal CPU A to be warmer than CPU B, usually around 10 to 12ºC when in full load.

You can confirm if you botched something installing the current CPU A to the socket B and then testing again to see if anything changes.

Thanks Alex. I thought 20-25c was a bit much of a spread. Plus, I've never seen the fans at 5200RPM before - apparently that is the max? CPU A stabilized around 82c during the test, but It spiked pretty quickly above 90-92 initially. Then the fans brought it down, but only to 82.

 

[h9826790]

It spiked pretty quickly above 90-92 initially. Then the fans brought it down, but only to 82.

That means CPU A thermal paste isn't working as expected.

Another sign of poor heat transfer is the delta temperature between CPU diode and CPU heatsink. As you can see, the delta of CPU B is just 14°C. But the delta of CPU A is 19°C. Which means the heat in CPU A is trapped inside the CPU, and unable to transfer to the heatsink effectively.

If the thermal paste is working well. When you stress the CPU, you should see the diode temperature going up slowly, then reach ~85°C and stay there (this is the target temperature of the Apple native fan profile. That's why your CPU stabilise at that temperature). But if your CPU diode temperature can increase to 92°C quickly, this means the thermal paste isn't working well.

By considering you just replace the CPU, the thermal paste is new, then most likely the heatsink isn't tight enough. Since your CPU are delidded. You may tighten the CPU A heatsink screw 1/8 or even 1/4 turn, and see if that let CPU A cool better.

Anyway, Prime95 can do better for CPU stress test. Handbrake not necessary able to use all CPU thread effectively.

 

[fatespawn]

Thanks - I tried giving the heatsink an addition 1/8 turn, but that didn't have any effect. So, I repasted.

I guess I'm a little happier. The CPU was slower to ramp up to the high 80's. That gave the natural fan curve time to compensate and keep it from spiking. CPU A did hit 89/90 fluctuating until Boost A got up to 2000+ RPM's.

That's good info about the delta between the heatsink and the packaage. In this test, the delta's are 17 and 19 respectively and the CPU's are within 11c of each other. That seems better.