Path: csiph.com!eternal-september.org!feeder.eternal-september.org!reader02.eternal-september.org!kreme.dont-email.me!.POSTED!not-for-mail From: Lewis Newsgroups: comp.sys.mac.system Subject: Re: Apple Silicon M1 Chip in MacBook Air Outperforms High-End 16-Inch MacBook Pro and All iOS Devices Date: Sat, 21 Nov 2020 17:54:43 -0000 (UTC) Organization: Miskatonic U Lines: 48 Message-ID: References: Reply-To: g.kreme@gmail.don-t-email-me.com Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit Injection-Date: Sat, 21 Nov 2020 17:54:43 -0000 (UTC) Injection-Info: kreme.dont-email.me; posting-host="b5a7d23a50aec55a6c7bbe1d37202e0d"; logging-data="10468"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX1+9RNBFxqp6QD6e9DZZEPjh" User-Agent: slrn/1.0.3 (Darwin) Cancel-Lock: sha1:XLdwXyBDkpQ9XoZpLVXK3cs4wuM= X-Face: )^b5"R:T7U>9~:PEn3YkzMfW*[b1qKeU.fP9C8~8HpU9}lA&6`bH1z Mail-Copies-To: nobody X-Clacks-Overhead: GNU Terry Pratchett Xref: csiph.com comp.sys.mac.system:135129 In message Alan Browne wrote: > On 2020-11-21 11:47, Lewis wrote: >> In message Alan Browne wrote: >>> Others (nospam, JR, Lewis) have posted many links to factual information >>> about M1 Macs beating many higher end intel Macs in performance (speed, >>> energy). That's really all you need to know. >> >> But he would need to 1) read 2) reread 3) reread 4) understand. >> >> And what are the odds of that? > Best? 3 out of 4. Obviously. >> >>> In effect when running Mac OS on Mx architectures it's not the same Mac >>> OS at the system level. And that really is the point: moving what used >>> to be complex OS ops out of the OS and into the chip allowing the CPU to >>> do higher level things much faster and by default the lower level things >>> blaze. >> >> The key thing that is most impressive about the M1 is that it can run >> translated Intel x64 code faster than Intel CPUs can. > That's a blackbox view so far. We don't know the cycle speed(s) so it's > hard to know what is actually what. We do know that Intel x64 apps are running faster on the M1 than on the remaining 13" Mac Book Pro. We also know that some of the APIs execute twice as fast in translation than on the Intel chips. I forget the exact call that Apple mentioned, but it was something like NSRelease which takes 30ns on intel, 8 on the M1, and 14ns on the M1 in Intel mode. Ah, yes, here it is: Fun fact: retaining and releasing an NSObject takes ~30 nanoseconds on current gen Intel, and ~6.5 nanoseconds on an M1 … and ~14 nanoseconds on an M1 emulating an Intel. This was David Smith, an Apple Engineer (not the be confused with the developer _DavidSmith). -- The real American folksong is a rag -- a mental jag A rhythmic tone for the chronic blues