Leaked benchmarks from developer kit for Apple's home-baked silicon appear to give Microsoft a run for its money

Benchmarks from the Mac Mini-based Developer Transition Kit powered by Apple's homegrown silicon have started to appear on GeekBench, showing competitive performance despite the inevitable hit from Cupertino's Rosetta 2 compatibility software. At the time of writing, 22 Geekbench 5 benchmarks have been submitted. That's …

[Charlie Clark]

Interestingly, Geekbench notes that the chip running on the Transition Kits has four cores…

Not really, no point in shipping big.Little silicon for what is a non-mobile developer workstation and Apple will presumably be aiming to ship beefier kit in its own hardware later this year, otherwise it would probably selling it already.

[Anonymous Coward]

Maybe, but low power cores in a laptop would allow some background activity to continue when sleeping on battery.

[Charlie Clark]

Sure, which is why they're not in the developer box which doesn't have a battery.

[Pascal Monett]

If the laptop is sleeping then the only background activity that needs to take place is knowing when to wake up.

Let's not start confusing modes, shall we ? Sleep is sleep. If the laptop's CPU is working, then it's not sleeping.

[Len]

Apple has used an existing two year old mobile chip in the development kit. It's not custom made or the one they'll be selling later.

The most likely explanation is that Geekbench (developed for x86, not ARM and running under Rosetta 2) is incapable of detecting the little cores as it has never had to look for those types. The extra cores will be available to developers on the dev kit, it's just that Geekbench doesn't report them.

[Charlie Clark]

Apple's policy

I'm sure Apple knew that benchmarks would be released so I wonder why they decided to bother banning developers from providing them; apart from the usual paranoia that is. Presumably, because it gives them plausible deniability over benchmarks for code that is almost certainly not optimised, while at the same time knowing that the leaked scores will keep people talking about the speed of the new chips.

Will we see improvements over time as the compiler and emulator get better? Will we see a baseline that allows Apple to gather wows™ when new hardware is released?

[BebopWeBop]

Re: Apple's policy

One minor point, I don't think that the article suggested that they had been leaked, just mentioned that the developer need permission - so possibly a nod and a wink have been shown?

[Len]

Re: Apple's policy

Some people think that this is a deliberate ploy by Apple. The idea goes that Apple knows that benchmarking tools for x86 will not be able to accurately detect actual performance of ARM architecture chips but the reported performance is already quite good, putting people at ease about having to rely on Rosetta 2. When they release the actual chips (the CPU in the dev kit is an existing two year old mobile model, not the one they'll be selling later) they will be so much faster than the leaked speeds that it generates extra excitement upon launch.

I think the above theory is a bit far-fetched but it's not impossible.

[DS999]

Re: Apple's policy

Just about every pre-release of hardware or software that requires signing an NDA bans benchmarking. Some things ban benchmarking even for shipping code - read the contracts your business has to sign with Oracle sometime, publishing any benchmarks is a serious violation...

[DS999]

Re: Apple's policy

There is undoubtedly debugging code in both the OS and Rosetta, and both will still be tuned and improved so the current performance is a low bar that can only improve.

Also, the A12Z in this temporary dev machine will be replaced by an A14 derived SoC in the shipping Macs, so performance will improve by somewhere between 30% to 50% based on that alone.

[Anonymous Coward]

Ouch ...

“who paid $500 to loan the device from Apple” Borrow ?

[Anonymous Coward]

Re: Ouch ...

It stops the "I'll grab one of those for a play" hackers from making them unavailable for "real" developers.

[don't you hate it when you lose your account]

Re: Ouch ...

Somebody in Apples marketing department used to work for the cable companies

[Daniel von Asmuth]

Armut

The MacMini (4 ARM cores) achieves a score of 2871 points against 19,155 points for ye aulde Mack Pro (28 Xeon cores). I suppose Apple's next entry level Mac will have 64 cores put it through its paces.

[Waseem Alkurdi]

Of course it won't have the same chip.

"Another point worth noting: there's no guarantee the A12Z chip in the Developer Transition Kit will appear in the first consumer Arm-based Macs, which are expected to land later this year. "

The other well-known Developer Transition Kit, the one for PowerPC to x86, had a Pentium 4, while the final release had a Core Duo chip (which was much more powerful).

[tip pc]

GeekBench compiled for OSX on arm

will be interesting to see what a GeekBench compiled for osx on arm will produce.

looks like the 4 cores scaled to 28 would outgun the 28 core xeon.

How many cores can apple go to?

[cipnt]

Re: GeekBench compiled for OSX on arm

The Mac Pro (8+ cores) will probably be the last to get Arm chips. The main benefits of Arm – increased energy efficiency and the "little" cores – will be best used in the MacBook, maybe even MacBook Pro range.

But if Apple chips can outperform Intel, either at single core performance or by optimizing its software for a significantly larger core count, then they will be welcome in the Pro range as well. But Apple will either have to create a special layer at OS level to better split the load to multiple cores, or they will need to encurage devs to adjust their code for larger core count.

For example Adobe CC is not currently optimised to make full use of 12 or more cores even in the most demanding tasks, so a higher clock cpu performs better than one with more cores.