4 ms·

The M1 peaks at about 27W at max compute load. Apple TDPs are like Intel TDPs - they’re a benchmarking number pulled out of the air for marketing reasons that h

by pja 6y ago

The M1 peaks at about 27W at max compute load. Apple TDPs are like Intel TDPs - they’re a benchmarking number pulled out of the air for marketing reasons that have at best a tenuous relationship with reality. (This is not to undermine the fact that the M1 is a /great/ chip, but the 10W figure is a marketing number.)

ogrisel 6y ago
A MacBook Air with M1 does not have a fan and I find it faster than my Dell XPS 13 from 2015 when running the scikit-learn test suite which is quite CPU intensive. Running those tests on the XPS 13 would typically cause the fan to turn on after a minute or two if I remember correctly.
- pja 6y ago
  One would hope that a CPU from 2020 would be faster than one shipped in 2015! As I said, the M1 is a /great/ CPU. But the numbers are marketing numbers & should be treated as such. The same goes for the CPU in your XPS. It was sold with an advertised TDP, was probably configured by the manufacturer to a /different/ TDP target subject to the thermal constraints of the laptop & as an end-user you can shift the set-points of thermals vs power consumption yourself if you want to. None of this is new & Apple is no exception. If Intel hadn’t completely failed to deliver their last 2 fab shrinks, you’d be sat here raving about how amazing your passively cooled Intel Mac was I imagine.
GeekyBear 6y ago
There were a few caveats attached to those numbers. >It’s especially important to keep in mind that the figure of what we usually recall as TDP in processors is actually only a subset of the figures presented here, as beyond just the SoC we’re also measuring DRAM and voltage regulation overhead, something which is not included in TDP figures nor your typical package power readout on a laptop. https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...
- kllrnohj 6y ago
  The 27W is the number Anandtech estimates after accounting for those. Total power draw was 31W.
traveler01 6y ago
Yeah but in real world results these SocS are shining.

by pja 6y ago

ogrisel 6y ago

A MacBook Air with M1 does not have a fan and I find it faster than my Dell XPS 13 from 2015 when running the scikit-learn test suite which is quite CPU intensive. Running those tests on the XPS 13 would typically cause the fan to turn on after a minute or two if I remember correctly.

pja 6y ago
One would hope that a CPU from 2020 would be faster than one shipped in 2015! As I said, the M1 is a /great/ CPU. But the numbers are marketing numbers & should be treated as such. The same goes for the CPU in your XPS. It was sold with an advertised TDP, was probably configured by the manufacturer to a /different/ TDP target subject to the thermal constraints of the laptop & as an end-user you can shift the set-points of thermals vs power consumption yourself if you want to. None of this is new & Apple is no exception. If Intel hadn’t completely failed to deliver their last 2 fab shrinks, you’d be sat here raving about how amazing your passively cooled Intel Mac was I imagine.

GeekyBear 6y ago

There were a few caveats attached to those numbers. >It’s especially important to keep in mind that the figure of what we usually recall as TDP in processors is actually only a subset of the figures presented here, as beyond just the SoC we’re also measuring DRAM and voltage regulation overhead, something which is not included in TDP figures nor your typical package power readout on a laptop. https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...

kllrnohj 6y ago
The 27W is the number Anandtech estimates after accounting for those. Total power draw was 31W.

traveler01 6y ago

Yeah but in real world results these SocS are shining.