4 ms·

It looks like you lose about 30% performance to Rosetta. That's quite a lot, and makes it slower than current x86-64 processors, e.g. Ryzen 5000 series, or even

by zrm 6y ago

It looks like you lose about 30% performance to Rosetta. That's quite a lot, and makes it slower than current x86-64 processors, e.g. Ryzen 5000 series, or even the Ryzen 3000 series. But compared to x86-64 Macs it's not that much slower, because the x86-64 Macs are using old slower Intel CPUs.

jmgao 6y ago
And that's for static code. Chrome runs like garbage under rosetta for me, with frequent stutters, presumably because of the cost of dynamically recompiling freshly emitted code.
- nly 6y ago
  It's actually a marvel that it can even JIT translate the output of another JIT...
alwillis 6y ago
But compared to x86-64 Macs it's not that much slower, because the x86-64 Macs are using old slower Intel CPUs. Depends on the app but some of them run faster in Rosetta than they do on comparable PC hardware. Everyone keeps comparing a SoC designed for low-end (in the Apple lineup/ecosystem) to the Ryzen. As the AnandTech review noted, the M1 bests even the Ryzen on some benchmarks and Ryzen beats it in others. The point everyone conveniently seems to miss is a new SoC for mobile devices running at 10-20 watts shouldn't be in the same universe when it comes to performance with any Ryzen desktop processor—but here we are. It's disingenuous to state the only reason M1 seems fast is because all Macs are using old, slow processors. The 16-inch MacBook Pro ($2,799.00) comes with a 2.3GHz 8‑core 9th‑generation Intel Core i9 processor, Turbo Boost up to 4.8GHz and for an additional $200, you can get it with a 2.4GHz 8‑core 9th‑generation Intel Core i9 processor, Turbo Boost up to 5.0GHz. And yet the new 13-inch MacBook Pro with it's M1 processor blows them away when it comes to performance. Here's a direct quote from their extensive review of the M1-based Mac mini (https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...): The M1 undisputedly outperforms the core performance of everything Intel has to offer, and battles it with AMD’s new Zen3, winning some, losing some. And in the mobile space in particular, there doesn’t seem to be an equivalent in either ST or MT performance – at least within the same power budgets.
- zrm 6y ago
  > The point everyone conveniently seems to miss is a new SoC for mobile devices running at 10-20 watts shouldn't be in the same universe when it comes to performance with any Ryzen desktop processor—but here we are. 10-20 watts is how much desktop processors use for a single thread, so they're not claiming an unfair advantage there. It would be truly impressive if the M1 at that power level was competitive with desktop processors on multi-threaded workloads where they actually use their full TDP, but it isn't even close. > The 16-inch MacBook Pro ($2,799.00) comes with a 2.3GHz 8‑core 9th‑generation Intel Core i9 processor, Turbo Boost up to 4.8GHz and for an additional $200, you can get it with a 2.4GHz 8‑core 9th‑generation Intel Core i9 processor, Turbo Boost up to 5.0GHz. Which are old slower Intel CPUs. Intel's latest is 10th-generation and even those are slower than Ryzen, much less the older 9th-generation ones. And even those older Intel mobile CPUs defeat the M1 on multi-threaded workloads like Cinebench. Core i9-9880H scores 8311 compared to 7833 for the M1 (native): https://www.cpu-monkey.com/en/cpu-intel_core_i9_9880h-958 https://www.cpu-monkey.com/en/cpu-intel_core_i9_9880h-958 https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested/2 https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...
  - alwillis 6y ago
    Which are old slower Intel CPUs. Intel's latest is 10th-generation and even those are slower than Ryzen, much less the older 9th-generation ones. And even those older Intel mobile CPUs defeat the M1 on multi-threaded workloads like Cinebench. You seem to be ignoring the benchmarks where the M1 beats the Ryzen 9 5950X on several--but not all--benchmarks. [1] From the article: In SPECint2006, we’re now seeing the M1 close the gap to AMD’s Zen3, beating it in several workloads now, which increasing the gap to Intel’s new Tiger Lake design as well as their top-performing desktop CPU, which the M1 now beats in the majority of workloads. And there's this: The situation doesn’t change too much with the newer SPECint2017 suite. Apple’s Firestorm core here remains extremely impressive, at worst matching up Intel’s new Tiger Lake CPU in single-threaded performance, and at best, keeping up and sometimes beating AMD’s new Zen3 CPU in the new Ryzen 5000 chips. And regarding power consumption: While AMD’s Zen3 still holds the leads in several workloads, we need to remind ourselves that this comes at a great cost in power consumption in the +49W range while the Apple M1 here is using 7-8W total device active power. That's the point I previously made—Apple’s first go at making their own processors for their entry level laptops and the Mac mini is super competitive with AMD's newest and best processors and beats it in several benchmarks, at a fraction of the cost and power consumption. That's pretty impressive. It has to make one wonder what the results will be when Apple ships a processor made for higher-end desktops, where they won't have the thermal and power restrictions of a laptop processor? I'm sure when they were contemplating a transition away from Intel, they had to do their due diligence and consider if AMD could be an option. But clearly they knew they have a much higher ceiling with their architecture. You know there are prototypes in Apple's lab now that far surpass Ryzen that'll be in machines spring/summer 2021. Apple's been relentless with shipping faster A series chips every year for the iPhone; I suspect we'll see the same thing with the M series. [1]: https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested/4 https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...
alwillis 6y ago
It looks like you lose about 30% performance to Rosetta. Regarding Rosetta, the AnandTech conclusion is (https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested/7 https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...): The software side of things already look good on day 1 due to Apple’s Rosetta2. Whilst the software doesn’t offer the best the hardware can offer, with time, as developers migrate their applications to native Apple Silicon support, the ecosystem will flourish. And in the meantime, the M1 is fast enough that it can absorb the performance hit from Rosetta2 and still deliver solid performance for all but the most CPU-critical x86 applications.

by zrm 6y ago

jmgao 6y ago

And that's for static code. Chrome runs like garbage under rosetta for me, with frequent stutters, presumably because of the cost of dynamically recompiling freshly emitted code.

nly 6y ago
It's actually a marvel that it can even JIT translate the output of another JIT...

alwillis 6y ago

But compared to x86-64 Macs it's not that much slower, because the x86-64 Macs are using old slower Intel CPUs. Depends on the app but some of them run faster in Rosetta than they do on comparable PC hardware. Everyone keeps comparing a SoC designed for low-end (in the Apple lineup/ecosystem) to the Ryzen. As the AnandTech review noted, the M1 bests even the Ryzen on some benchmarks and Ryzen beats it in others. The point everyone conveniently seems to miss is a new SoC for mobile devices running at 10-20 watts shouldn't be in the same universe when it comes to performance with any Ryzen desktop processor—but here we are. It's disingenuous to state the only reason M1 seems fast is because all Macs are using old, slow processors. The 16-inch MacBook Pro ($2,799.00) comes with a 2.3GHz 8‑core 9th‑generation Intel Core i9 processor, Turbo Boost up to 4.8GHz and for an additional $200, you can get it with a 2.4GHz 8‑core 9th‑generation Intel Core i9 processor, Turbo Boost up to 5.0GHz. And yet the new 13-inch MacBook Pro with it's M1 processor blows them away when it comes to performance. Here's a direct quote from their extensive review of the M1-based Mac mini (https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...): The M1 undisputedly outperforms the core performance of everything Intel has to offer, and battles it with AMD’s new Zen3, winning some, losing some. And in the mobile space in particular, there doesn’t seem to be an equivalent in either ST or MT performance – at least within the same power budgets.

zrm 6y ago
> The point everyone conveniently seems to miss is a new SoC for mobile devices running at 10-20 watts shouldn't be in the same universe when it comes to performance with any Ryzen desktop processor—but here we are. 10-20 watts is how much desktop processors use for a single thread, so they're not claiming an unfair advantage there. It would be truly impressive if the M1 at that power level was competitive with desktop processors on multi-threaded workloads where they actually use their full TDP, but it isn't even close. > The 16-inch MacBook Pro ($2,799.00) comes with a 2.3GHz 8‑core 9th‑generation Intel Core i9 processor, Turbo Boost up to 4.8GHz and for an additional $200, you can get it with a 2.4GHz 8‑core 9th‑generation Intel Core i9 processor, Turbo Boost up to 5.0GHz. Which are old slower Intel CPUs. Intel's latest is 10th-generation and even those are slower than Ryzen, much less the older 9th-generation ones. And even those older Intel mobile CPUs defeat the M1 on multi-threaded workloads like Cinebench. Core i9-9880H scores 8311 compared to 7833 for the M1 (native): https://www.cpu-monkey.com/en/cpu-intel_core_i9_9880h-958 https://www.cpu-monkey.com/en/cpu-intel_core_i9_9880h-958 https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested/2 https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...
- alwillis 6y ago
  Which are old slower Intel CPUs. Intel's latest is 10th-generation and even those are slower than Ryzen, much less the older 9th-generation ones. And even those older Intel mobile CPUs defeat the M1 on multi-threaded workloads like Cinebench. You seem to be ignoring the benchmarks where the M1 beats the Ryzen 9 5950X on several--but not all--benchmarks. [1] From the article: In SPECint2006, we’re now seeing the M1 close the gap to AMD’s Zen3, beating it in several workloads now, which increasing the gap to Intel’s new Tiger Lake design as well as their top-performing desktop CPU, which the M1 now beats in the majority of workloads. And there's this: The situation doesn’t change too much with the newer SPECint2017 suite. Apple’s Firestorm core here remains extremely impressive, at worst matching up Intel’s new Tiger Lake CPU in single-threaded performance, and at best, keeping up and sometimes beating AMD’s new Zen3 CPU in the new Ryzen 5000 chips. And regarding power consumption: While AMD’s Zen3 still holds the leads in several workloads, we need to remind ourselves that this comes at a great cost in power consumption in the +49W range while the Apple M1 here is using 7-8W total device active power. That's the point I previously made—Apple’s first go at making their own processors for their entry level laptops and the Mac mini is super competitive with AMD's newest and best processors and beats it in several benchmarks, at a fraction of the cost and power consumption. That's pretty impressive. It has to make one wonder what the results will be when Apple ships a processor made for higher-end desktops, where they won't have the thermal and power restrictions of a laptop processor? I'm sure when they were contemplating a transition away from Intel, they had to do their due diligence and consider if AMD could be an option. But clearly they knew they have a much higher ceiling with their architecture. You know there are prototypes in Apple's lab now that far surpass Ryzen that'll be in machines spring/summer 2021. Apple's been relentless with shipping faster A series chips every year for the iPhone; I suspect we'll see the same thing with the M series. [1]: https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested/4 https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...

alwillis 6y ago

It looks like you lose about 30% performance to Rosetta. Regarding Rosetta, the AnandTech conclusion is (https://www.anandtech.com/show/16252/mac-mini-apple-m1-tested/7 https://www.anandtech.com/show/16252/mac-mini-apple-m1-teste...): The software side of things already look good on day 1 due to Apple’s Rosetta2. Whilst the software doesn’t offer the best the hardware can offer, with time, as developers migrate their applications to native Apple Silicon support, the ecosystem will flourish. And in the meantime, the M1 is fast enough that it can absorb the performance hit from Rosetta2 and still deliver solid performance for all but the most CPU-critical x86 applications.