Because you don't pre-announce hardware, but yet you still want to give developers a chance to recompile and test in advance.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Rosetta 2 Benchmarks Surface From Mac Mini With A12Z Chip
- Thread starter MacRumors
- Start date
- Sort by reaction score
Because you don't pre-announce hardware, but yet you still want to give developers a chance to recompile and test in advance.
Won't put in the effort for what? These are dev kits that they have to go into a lottery just for the chance to rent, anyone jumping through that many hoops is going to put in the effort. And those that don't now can catch up later when the final hardware launches.
What exactly did I get wrong? Rosetta produces native ARM code. I did not claim it was optimized. Interestingly enough, the code may actually inherit some high level optimization delivered by the C++ compiler on x86.
Spot on analysis. I'm actually impressed with the performance based on the limited cores, rosetta, etc.
Why in the world would you want “benchmarks” from an underclocked version of the A12 meant for developers to test their apps knowing that Rosetta is adding in overhead and, more importantly, that the chip itself will never see the light of day in a product that will ship to consumers?
Someone posted his own benchmarking results for iPad a few posts earlier. His observation was that the small core performance is 10% of the big core. If this is correct, adding 4 small cores would add 10% performance at best. The impact of Rosetta is hard to estimate.
I am actually blown away as this is a chip that's two generations old binned chip running MacOS and being able to translate the code well. I suspect the new iMac will blow socks away in terms of performance with its new form design.
The problem they want to solve, is that they can't make their laptops so THIN that you can't see it when viewed in profile with anything Intel is currently manufacturing.
All they seem to care about is making everything, even desktops like the iMac, obnoxiously thin, regardless of how it affects basic function. (see butterfly keyboard, proper cooling most iMacs for the past 6 years, and now losing the option to run Windows on the same machine)
It wouldn't shock me if the first ARM based Mac laptop came with a zero-key-travel keyboard, just to make it as obnoxiously thin as possible. Don't think for a second there isn't a worse version of the POS butterfly KB in the works.
Form over function. And Apple has shown time and time again, thin is the only form they care about.
I'm just going to sit back and enjoy the show. I'm pretty sure I'm long since blocked, but maybe some of you who still care to ... could ...
apple.com/feedback
?
its the ROSETTA emulation being benchmarked for those curious about the emulation performance.
Mac is going to be downgraded in performance and software availability lol. Emulation is just a bandaid, and in the case of Macs... it’s going to be the way of life. Emulate everything that matters but hey! you get native crappy iOS apps on a $4000 machine as a trade off! Woohoo!
Last edited:
Because this is not a real product so what difference does it make? Or perhaps since the chip will throttle down far less often than in an iphone, and it will run a high percentage of the time at this higher clock rate, there is some sort of limitation on whatever power supply parts they are using, or some other issue.
[automerge]1593453990[/automerge]
I disagree entirely. We now know that at least one app takes something like a 30% hit. We don’t know anything about any other apps. And it’s only a percentage - the real chips may run 5x faster than this. We have no idea. So 30% of 5x faster is as much a possibility as anything.
I'm still not excited for or in favor of this transition, but I will say I don't know why some people are so negative on these results (perhaps they're just being brought back down to reality... LOL).
Keep in mind GeekBench 5's baseline is a Core i3 8100, a 3.6Ghz 4C/4T CPU that scores roughly 1000 in ST (duh, its the baseline) and 3000~3500 in MT, so a score of 830/2800(ish) for non-native code on the A12Z running at 2.4Ghz on only 4 cores is actually pretty damn impressive IMHO.
Does it mean this is an acceptable level of performance for a Mac in late 2020/early 2021? No, but assuming that the Mac chips will be A14 derivatives, that the A14 delivers a measurable jump over the A13, and that they will likely feature more cores and higher clocks/power envelopes vs the A12Z, it's not a bad initial showing. While I'm going to need to see more benchmarks (Cinibench please!) to really come to a conclusion, these initial Geekbench scores are encouraging and Apple at least gets a passing grade for Rosseta 2 in my book (certainly better than I was expecting)
Keep in mind GeekBench 5's baseline is a Core i3 8100, a 3.6Ghz 4C/4T CPU that scores roughly 1000 in ST (duh, its the baseline) and 3000~3500 in MT, so a score of 830/2800(ish) for non-native code on the A12Z running at 2.4Ghz on only 4 cores is actually pretty damn impressive IMHO.
Does it mean this is an acceptable level of performance for a Mac in late 2020/early 2021? No, but assuming that the Mac chips will be A14 derivatives, that the A14 delivers a measurable jump over the A13, and that they will likely feature more cores and higher clocks/power envelopes vs the A12Z, it's not a bad initial showing. While I'm going to need to see more benchmarks (Cinibench please!) to really come to a conclusion, these initial Geekbench scores are encouraging and Apple at least gets a passing grade for Rosseta 2 in my book (certainly better than I was expecting)
I find it ironic that two of those Macs are Macs I use daily, right now. The 2012 iMac is my workshop Mac, and I'm typing this on a 2012 non-retina 15" MBP?
Like deja vu all over again ;-)
Means little to Intel. ARM needs more memory for instruction because they are larger. x86 has smaller memory footprint for instructions.
It will be much more interesting if some developer would post the results of his professional application natively compiled for ARM vs the same software on a comparable Intel.
I know Apple and Tim Cook. They are masters of supply chain. They want to minimize costs and risk while maximizing throughput. IMO, the reason to go to ARM is to vertically integrate their supply chain.
An idea for fun - Apple will create 1 chip for mobile, 1 chip for laptops, and 1 chip for desktop - all chips will have the maximum cores available but Apple will disable some of them for the cheaper machines. However, imagine being able to pay for performance when you need it. Like, I purchase a MBP with 4 core running at 2Ghz for $1k. But if I need a little more juice, and if apple is using a 1chip design, then I can pay to unlock more cores boosting performance.
Just an idea, but this move will open up many cool options for Apple to explore. Tim Cook, if you're reading this - DM me. I can make your company fly
Last edited:
Then the developer is responsible for having let an unauthorized third party have access to the dev kit machine. Also against the agreement they signed.
There is no intent here to make a powerful machine. What Apple is producing here is something that works, that's all. They are not selling these machines.
Arm working sets are not much bigger than Intel working sets. Given current compiler technology and assuming we are talking about x86-64 and not using 32-bit instructions, the main arm overhead is just additional bits for the predicate field, and occasionally more opcode bits.