8 core 14" - temperature sensors present for 'missing' cores

[fm2503]

TG Pro got update today with suport for M1 PRO, and sure enough, it really does look like the last 2 cores are there, but disabled. Amusingly the temperature readouts are there though.....

The (presumably) unused Performance Cores 7/8 (or whichever they are, see below) hotter than PC5

 

[baummer]

Yes - this is a practice called "chip binning".

 

[fm2503]

Yes - this is a practice called "chip binning".

I get that - I'm old enough to remember 486DX vs 486SX, I'm just slightly surprised they didn't disable the thermal sensors on the binned/disabled cores.

 

[PTLove]

Im more surprised by the fact it shows cores 7 and 8 as the disabled ones. I suppose it could just be how the software is ordering it, but if its not that implies that the cores on a single cluster were disabled, so we have one cluster with 2 cores and the other with 4. This is interesting because the way clock scaling works within a cluster is that when 3+ cores are active, it runs at 3.05ghz, when two are active it runs at 3.12ghz. Meaning having 2 cores on the same cluster disabled gives you *slightly* higher performance in a all out scenario.

 

[fm2503]

Im more surprised by the fact it shows cores 7 and 8 as the disabled ones. I suppose it could just be how the software is ordering it, but if its not that implies that the cores on a single cluster were disabled, so we have one cluster with 2 cores and the other with 4. This is interesting because the way clock scaling works within a cluster is that when 3+ cores are active, it runs at 3.05ghz, when two are active it runs at 3.12ghz. Meaning having 2 cores on the same cluster disabled gives you *slightly* higher performance in a all out scenario.

Hmm, it seems assumptions about which cores are disabled may be false - although even the numbering scheme below is open to interpretation - although it implies P cores 0 & 1 are the disabled ones....

root@Matts-MacBook-Pro Preferences # powermetrics | grep frequen
E-Cluster HW active frequency: 1583 MHz
CPU 0 frequency: 1633 MHz
CPU 1 frequency: 1631 MHz

P0-Cluster HW active frequency: 983 MHz
CPU 2 frequency: 1437 MHz
CPU 3 frequency: 1543 MHz
CPU 4 frequency: 1645 MHz

P1-Cluster HW active frequency: 616 MHz
CPU 5 frequency: 1118 MHz
CPU 6 frequency: 1128 MHz
CPU 7 frequency: 1124 MHz


GPU active frequency: 29 MHz
GPU requested frequency: (389 MHz: 7.5% 486 MHz: 0% 648 MHz: 0% 778 MHz: 0% 972 MHz: 0% 1296 MHz: 0%)

If I synthesise load, then each P cluster goes up to 3036MHz and each P core to 3228MHz
Based on observing temperature of each core under load, then I'd say it's P cores 4 & 8 (0 and 7 if you want to number if that way) which are disabled

 

[uecker87]

I am kind of surprised at just how warm this chip gets. My base 14" still gets up to around 90C in the first run of Cinebench. When I tried out a M1 MBP a ways back it rarely got above 70C even under sustained load.

 

[JamieLannister]

I am kind of surprised at just how warm this chip gets. My base 14" still gets up to around 90C in the first run of Cinebench. When I tried out a M1 MBP a ways back it rarely got above 70C even under sustained load.

my 14" 10c/16c 16GB/1TB model gets cool to warm near the right-center on keyboard and bottom center. Perhaps the fan really good at exchanging the heat venting everything out properly. I only watch YouTube 4k videos and it can be "felt". My m1 mini 16GB ram model is cool to touch (surface) all the time. I need to download a fan monitor to see if the fans are even on. I can't hear them
!

 

[uecker87]

my 14" 10c/16c 16GB/1TB model gets cool to warm near the right-center on keyboard and bottom center. Perhaps the fan really good at exchanging the heat venting everything out properly. I only watch YouTube 4k videos and it can be "felt". My m1 mini 16GB ram model is cool to touch (surface) all the time. I need to download a fan monitor to see if the fans are even on. I can't hear them
!

Fans don't kick on until the temps reach about 90C on my MBP14. Even then they are extremely low and impossible to hear. I'm sure if I set a custom fan curve I can make it cooler under load.

Using TG Pro for now until Macs Fan Control gets updated for the new chips.

 

[JamieLannister]

my 14" 10c/16c 16GB/1TB model gets cool to warm near the right-center on keyboard and bottom center. Perhaps the fan really good at exchanging the heat venting everything out properly. I only watch YouTube 4k videos and it can be "felt". My m1 mini 16GB ram model is cool to touch (surface) all the time. I need to download a fan monitor to see if the fans are even on. I can't hear them
!

So about over 42 minutes into watching 4k YouTube video and some split screen web browsing, the entire chassis is WARM. keyboard center area warm, entire bottom case warm. So yes, it's interesting how this m1 pro 10c is much warmer than the m1 8core

 

[JamieLannister]

Fans don't kick on until the temps reach about 90C on my MBP14. Even then they are extremely low and impossible to hear. I'm sure if I set a custom fan curve I can make it cooler under load.

Using TG Pro for now until Macs Fan Control gets updated for the new chips.

yes, TG pro was the app - thanks for the reminder. That's crazy it gets to 90C then the fans kick on!

 

[uecker87]

yes, TG pro was the app - thanks for the reminder. That's crazy it gets to 90C then the fans kick on!

Yeah I have been playing around with it with a custom fan curve. I prefer it to run cooler with the fans on, so I'm dialing it in. These fans are still really quiet when under 75% even.

 

[PTLove]

Hmm, it seems assumptions about which cores are disabled may be false - although even the numbering scheme below is open to interpretation - although it implies P cores 0 & 1 are the disabled ones....



If I synthesise load, then each P cluster goes up to 3036MHz and each P core to 3228MHz
Based on observing temperature of each core under load, then I'd say it's P cores 4 & 8 (0 and 7 if you want to number if that way) which are disabled

Nice. Certainly makes more sense if they do it that way (limits the binning, but explains 3 and 4 core max frequency being equal).