Apple MacBook Air (M1, 2020) much faster than XPS 9730 (Intel Core i9-13900H, 2022)

[Martin Schroeder]

Dear stata forum,
I have recently used a code along these lines both on an Apple MacBook Air (M1, 2020) and an XPS 9730 (Intel Core i9-13900H, 2022)


use "ESS-Data-Wizard-subset-2023-08-15.dta"

timer clear
foreach num of numlist 1 2 4 6 8 {
set processor `num'
timer on `num'
quiet: regress x y
timer off `num'
}
timer list

And I was pretty surprised by the results. Regardless of how many cores I have used, the M1, which is supposed to be a far more mobile and slower machine than the i9, was never far off in processing speed, often being pretty much on par (and never getting hot). It also made no difference whether I used e.g. meologit, mixed, ologit etc.

So my question to you would be: have you had similar experiences? Are Apples chips just so much faster with Stata? I would love to hear how e.g. a MacBook Pro is doing, which is actually designed for speed, while the MacBook Air is a much more portable machine that should not even come close to an i9.

I do a lot of heavy calculations, so if others had similar experiences, I will switch from Intel/Windows to a Mac. I am also sure others would be keen to hear what the most potent machine for Stata is.

Let me know what you think!

[Leonardo Guizzetti]

I think that your result is counter-intertuitive. Is it possible that your Intel machine is using power conservation to effectively run the CPU at a lower frequency, therefore making it appear slower than its full capability?

[Mike Lacy]

Seemingly small differences in the Stata setup on the two machines can matter a lot: I once found that a program ran *much* slower on a machine on which the maximum matrix size was set to a large value. (My posting about that should be somewhere in the archives here.) I'm not saying that's the issue here, but it illustrates what can happen.

[Martin Schroeder]

Yes, the result is definitely counter-intuitive. My i9's power limits were set to PL1 75 watts and PL2 115 watts and it reached 5 Ghz. But I totally agree that the result is very surprising, which is why I would love to hear from others how their Macs perform relative to their PCs. Maybe Stata just runs much more efficient on Macs compared to PCs?