Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Possible link to benchmarks here (I wouldn't read too much into it, but there isn't any other news):


It's funny for me that this "thing" emulating a CPU gives little better results than given by my old and good boy rMBP 15" 2012.

Now we can imagine how could be the "real Apple silicon" features and performance. We can expect it will bring PCIe, USB 4 (with TB3 compatibility), and other desktop features added to the final silicon design.
 
  • Angry
Reactions: djjeff
Can someone with DTK run Shadow of Tomb Raider to confirm performance and that it wasn't running on x64 hardware since the device was hidden from view?
 
  • Disagree
Reactions: hxlover904
I thought these things didn’t even have Thunderbolt 3. How did they even connect a Pro Display XDR?

do you really think Apple can design and engineer a world class chip but can’t figure out how to put thunderbolt port on it? Especially when we know they’re shipping out working minis with these chips? I doubt the minis have lightening ports on them.
 
Last edited:
Rosetta just statically translates (usually), so it shouldn’t affect how many cores are used. Could be that the os, itself, doesn’t yet schedule threads on all cores, or, more likely, it doesnt use the ”little” cores, at least for a12 (not a lot of reason for apple to spend time tuning the thread scheduler for a12, after all)
I think what you see is how many cores Rosetta reports to Intel code. Intel code assumes 8 cores = 8 fast cores. Scheduling 8 threads assuming they are all fast might be quite suboptimal. Not telling Intel code about four slow cores might make less difference. I would assume the scheduler is running on ARM and knows the truth.
 
I think what you see is how many cores Rosetta reports to Intel code. Intel code assumes 8 cores = 8 fast cores. Scheduling 8 threads assuming they are all fast might be quite suboptimal. Not telling Intel code about four slow cores might make less difference. I would assume the scheduler is running on ARM and knows the truth.
yes, surely the scheduler knows. The question is was the benchmark running on 8 cores or 4, and if only 4, why? I highly doubt Apple’s plan is to allow translated apps to run only on The high-performance cores. I wouldn’t be surprised if the low performance cores are just disabled at this point, because they didn’t spend the time to optimized the scheduler for this fake machine.
 
  • Like
Reactions: CarlJ
T2, M1 etc. You know this.
To be blunt: I don't think YOU know what you're talking about. The T2 chip only exists because Apple doesn't make the main chipset in the Mac. Once the Mac is on Apple Silicon the functionality of the T2 will either not be needed, or it will be rolled into the main SoC die.
You notice there's no iOS device that ships with a T2 because the functionality is build into the system chip.
 
Did Apple mention anything what'll happen to the $500 developers paid to get these DTKs once they are returned?
 
Did Apple mention anything what'll happen to the $500 developers paid to get these DTKs once they are returned?
Not that I’ve seen. The last time, on the transition to Intel, if I recall correctly, the participating devs got a seriously good deal on a well-equipped Intel iMac. No guarantees they’ll do the same here, but it wouldn’t surprise me.
 
Not that I’ve seen. The last time, on the transition to Intel, if I recall correctly, the participating devs got a seriously good deal on a well-equipped Intel iMac. No guarantees they’ll do the same here, but it wouldn’t surprise me.
I do reading about it, I read through their eligibility and it doesn't seem like anything is mentioned as well. Right now it seems like a $500 fee to rent an exclusive ARM Mac for a year. But you are right, since the $500 is non refundable, they might announce devs are able to upgrade to a Apple Silicon Mac with a hefty discount with this.
 
It's related, and I don't know where else to ask this. Should we/would we expect with new Macs to come out with new chips to be in a similar situation where back in the days of Power PC G5s, where they moved from Mac chips to Intel chips, that it caused the situation where we had to abandon our PowerPCs with MAc chips becuase Adobe and others didn't support the old Macs and only supported Intel? Similarly, new OS didn't support old Macs beyond Jaguar?
 
I wonder if developers have to sign NDAs regarding performance testing on the A12Z chip. I would like to see what it really can do versus an i7 or i9 without the Apple propaganda.
As promised, here are the numbers, I was actually wrong. Actually 2018 iPad CPUs are pretty much up to par on all benchmark related scores when compared to MBP 2020 10th gen highest end CPUs. Even multicore scores are faster on iPads.

Geekbench interesting Scores, I have highlighted the ones I was referring to, but also left a few others to see the ‘relevant range’ of scores.

GPU - Metal:

Image.png


Single core:

IMG_0131.jpeg


IMG_0132.jpeg


Multi-core:

IMG_0134.jpeg


IMG_0133.jpeg



———————
Excuse the weird aspect ratio, but again, to MacRumors people if you ever read this: on iOS devices either writing with the screen keyboard, physical keyboard or manipulating images is a frustration heavy endeavor... sometimes the text editor starts with an upper case letter after a space (not dot or similar), images can’t be uniformly scaled respecting it’s original aspect ratio, the ‘scale handles’ at the image borders do nothing and the only option to scale, which is via the box type-in values, doesn’t say what was the original pixel sizes per axis... if interesting in enhancing or full-on fixing, I can capture videos of each of those mentioned issues.
 
As promised, here are the numbers, I was actually wrong. Actually 2018 iPad CPUs are pretty much up to par on all benchmark related scores when compared to MBP 2020 10th gen highest end CPUs. Even multicore scores are faster on iPads.

Geekbench interesting Scores, I have highlighted the ones I was referring to, but also left a few others to see the ‘relevant range’ of scores.

GPU - Metal:

View attachment 929224

Single core:

View attachment 929223

View attachment 929227

Multi-core:

View attachment 929225

View attachment 929226


———————
Excuse the weird aspect ratio, but again, to MacRumors people if you ever read this: on iOS devices either writing with the screen keyboard, physical keyboard or manipulating images is a frustration heavy endeavor... sometimes the text editor starts with an upper case letter after a space (not dot or similar), images can’t be uniformly scaled respecting it’s original aspect ratio, the ‘scale handles’ at the image borders do nothing and the only option to scale, which is via the box type-in values, doesn’t say what was the original pixel sizes per axis... if interesting in enhancing or full-on fixing, I can capture videos of each of those mentioned issues.
Thanks for posting this information. This is very interesting and along the lines with what I was hearing about Apple's mobile chips. The ARM chips will probably be very impressive in a MBP.
 
  • Like
Reactions: amartinez1660
It’s only using half the cores.
it’s clock speed is lower than even an ipad pro
it’s running a Rosetta-translated version of the benchmark
it’s not the chip that will be in real macs

So it’s all pretty meaningless.

Yes, it's pretty meaningless. But it does tell us that an A12 emulating Intel with Rosetta 2 actually scores about the same as the current 3600Mhz i3 Mac Mini . That bodes very well for the future of Arm Macs.
 
Last edited:
Yeah but this isn't going to stop people.
AppleInsider did the benchmark and took the clip from youtube within the hour es. The results were not staggering, but I’m not sure the DTK is build for performance and stand the test of benchmarks. Those machines are prototypes.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.