Apple M1 CPU

[jigger]

Microsoft already do this for windows ARM devices, they must have either found a way around it or paid Intel for a special license. I'm sure it's not beyond Apple to do the same.

Apple paying Intel is good business for Apple. Also licensing tech from the competition* is erm.. topical.

 

[tommoT.Striker]

Interested to see where this goes. I use a mac book for software development, but never been sold on the apple ecosystem. But apple must be confident they can scale this class of M1 CPUs performance up to compete with the very best x86 CPUs. Let's see if software dev follows.

 

[Jimmy Weirdarms]

They don't though, do they? Windows arm devices can't run x86 apps. I thought that was why they run the silly cut down version of windows.

Yep, Microsoft ARM devices (like Surface Pro X) running Win 10 for ARM (not the older RT) can already run 32-bit x86 binaries, with 64-bit emulation starting to roll out on their "Insider" program this month, meaning it'll hit the mainstream a few months down the line.

 

[Vince]

Yep, Microsoft ARM devices (like Surface Pro X) running Win 10 for ARM (not the older RT) can already run 32-bit x86 binaries, with 64-bit emulation starting to roll out on their "Insider" program this month, meaning it'll hit the mainstream a few months down the line.

Nice. Who knew. Ive been so busy with life i missed this on Shall have a lil look.

 

[smilingcrow]

Yep, Microsoft ARM devices (like Surface Pro X) running Win 10 for ARM (not the older RT) can already run 32-bit x86 binaries, with 64-bit emulation starting to roll out on their "Insider" program this month, meaning it'll hit the mainstream a few months down the line.

The difference is that Apple are ruthless and have a small market share so they can just drop x86 support completely over a relatively short period.
Whereas MS struggle just to stop releasing 32 bit versions of Windows.
So for them to move to ARM would take about 200 years.

 

[Jimmy Weirdarms]

The difference is that Apple are ruthless and have a small market share so they can just drop x86 support completely over a relatively short period..

Seems unlikely they will any time soon - those expensive, expensive Mac Pros aren't going to be using ARM any time soon.
But yes, point taken, they can move faster than MS. How long did MS keep 16-bit support? It was ages...

That said, in the long past I worked for IBM and their mainframe series had a back compatibility date sometime in 1964!

 

[smilingcrow]

Seems unlikely they will any time soon - those expensive, expensive Mac Pros aren't going to be using ARM any time soon.

Apple stated about 2 years to complete the transition.
The base line is ancient Xeons on 14nm using DDR4, PCIe 3.0 and up to 28 cores.
In two years time Apple may be using 3/4nm, DDR5, PCIe 5.0 and at least 64 cores.
They don't have to annihilate AMD, but just trounce what they had previously and hopefully what Intel have when they release it.
Given Intel's troubles that doesn't seem hard.

 

[Jimmy Weirdarms]

Apple stated about 2 years to complete the transition.

Sure, for hardware, but I think they'll have to continue supporting their previous x86 stuff with software updates and compatibility for some time after that.

 

[smilingcrow]

Sure, for hardware, but I think they'll have to continue supporting their previous x86 stuff with software updates and compatibility for some time after that.

They will, but they won't sell any more Intel based systems, so they will have replaced the Mac Pro which is what you disputed.

 

[chef-borjan]

There are currently a handful of scores for the new M1 chips up on Geekbench, have averaged them out here.

Macbook Air - M1 3.2Ghz - 8GB - Single Core = 1689. Multi-Core: 7083 (4 tests. Note that the single core score seems to be sitting just below 1700, but there is a bit more variance on the multi core. Two scores are at 7400-ish, one around 7000 and one at 6500.)

Macbook Pro - M1 3.2Ghz - 16GB - Single Core = 1714. Multi-Core: 6802 (Just one test so far here)

iMac (27-inch Retina Mid 2020) - Intel Core i9-10910 @ 3.6 GHz (10 cores) - Single Core = 1251 Multi-core = 9012
MacBook Pro (13-inch Mid 2020) - Intel Core i5-1038NG7 @ 2.0 GHz (4 cores) - Single Core = 1148 Multi-core = 4240
MacBook Pro (16-inch Late 2019) - Intel Core i9-9980HK @ 2.4 GHz (8 cores) - Single Core = 1095 Multi-core = 6869
MacBook Air (Early 2020) - Intel Core i7-1060NG7 @ 1.2 GHz (4 cores) - Single Core = 1140 Multi-core = 3080


AMD Ryzen 9 5950X -3.4 GHz (16 cores) - Single Core = 1628 Multi-core = 15744
Intel Core i9-10900K - 3.7 GHz (10 cores) - Single Core = 1410 Multi-core = 11069

Made some quick comparisons. You probably know better about which CPU's to compare against... which ones are around the M1's TDP range?

 

[DVD Cinema]

But can the M1 work with graphic cards, because it isn't compatible with eGPU?

 

[Destination]

I wonder are the reported speeds swift burst, or sustained.
That single core does look good, but can it actually run any software?
Is the instruction set compatible? or will someone at ARM encourage the purchase of VIA to give them licensing?

 

[Armageus]

That single core does look good, but can it actually run any software?
Is the instruction set compatible? or will someone at ARM encourage the purchase of VIA to give them licensing?

It doesn't run x86 software natively if that's what you mean?

Under MacOS it will run x86 MacOS applications using an emualation layer (Rosetta2), that converts x86 OS calls to the appropriate Arm one's, and performs emulation on anything that can't be converted.

Geekbench above has apparently been recompiled to run on Arm.

Future Mac Apps will ship as "fat" universal binaries containing both x86 and Arm versions in the same executable, so will run on either processor

 

[DVD Cinema]

The MacBook Pro 13 GPU up to 5x faster test was M1 v i7 1.7Ghz (Pro 13) on a 10 second Final Cut Pro project.

 

[Destination]

Ah so what we actually need is a comparison of geekbench running in emulation mode, to see what the processor manages when tested in that fashion, rather than a recompile, which makes it a different test altogether.

 

[DVD Cinema]

Geekbench 5 :-

MacBook Air M1 3.2Ghz 10W passive cooling laptop posted an impressive 1687 and 7433 in the single- and multi-core tests, respectively v Core i9-9880H 2.3Ghz 48W 16-inch MacBook Pro. Down 35% and 15% respectively against the M1.

 

[gbgt2]

Geekbench is worthless as a benchmark. Cinebench R23 has been released with support for M1 though

 

[HACO]

Geekbench is worthless as a benchmark. Cinebench R23 has been released with support for M1 though

Geekbench, SPEC and Cinebench agree more or less completely (R^2 is over 0.95 between any two of them), the only differentiating factor being thermal throttling which Geekbench doesn't take into account because it's quick (hence the passively cooled 13-inch MBA and actively cooled MBP/MM get the same scores when they shouldn't).

Cinebench R23 runs for 10 minutes, that will be a pretty good measurement of sustained performance.

 

[Deleted member 209350]

Geekbench is worthless as a benchmark. Cinebench R23 has been released with support for M1 though

Finally someone who understands lol.

They just massively favour arm and apple chips in general, hence why its unreliable.

 

[Jimmy Weirdarms]

They will, but they won't sell any more Intel based systems, so they will have replaced the Mac Pro which is what you disputed.

No, I 'disputed' that they're going to drop x86 support within a couple of years.

Do try to keep up!