Apple Transitioning Macs to Custom Chips With 'Incredible' Performance Starting Later This Year

[laptech]

The whole selling point about ARM processors is their ability to work at low power ratings. Anything that is heavy on the number crunching such as games, video rendering, audio rendering, some financial software is going to struggle on an ARM processor.

If people disagree then I would be interested to hear which movie studies use ARM pc's or music studios that use ARM pc's or financial companies involved in the stock market that use ARM pc's.

 

[cmaier]

The whole selling point about ARM processors is their ability to work at low power ratings. Anything that is heavy on the number crunching such as games, video rendering, audio rendering, some financial software is going to struggle on an ARM processor.

If people disagree then I would be interested to hear which movie studies use ARM pc's or music studios that use ARM pc's or financial companies involved in the stock market that use ARM pc's.

This is stupid.

No such companies use arm pcs because nobody has made arm pcs yet (other than a couple minor examples using crappy off-the-shelf arm chips).

The selling point of Arm processors is their licensing model. They were also the first to really focus on low power.

But you know what, Amazon used to just be a bookstore. That doesn’t mean they were incapable of selling hammers. Things change.

 

[laptech]

This is stupid.

No such companies use arm pcs because nobody has made arm pcs yet (other than a couple minor examples using crappy off-the-shelf arm chips).

The selling point of Arm processors is their licensing model. They were also the first to really focus on low power.

But you know what, Amazon used to just be a bookstore. That doesn’t mean they were incapable of selling hammers. Things change.

If nobody has made arm pc's yet then why is Apple promoting how good performace wise their ARM computers are going to be? Don't you think that is a very brash statement to make considering one has not been made yet (acording to your post). So how does Apple know how good it's performance is going to be?? basing it the performance of their ARM ipads??

 

[cmaier]

If nobody has made arm pc's yet then why is Apple promoting how good performace wise their ARM computers are going to be? Don't you think that is a very brash statement to make considering one has not been made yet (acording to your post). So how does Apple know how good it's performance is going to be?? basing it the performance of their ARM ipads??

What kind of spaghetti logic is this?

*apple* obviously has made Arm computers, and they have them in the labs, but they don’t sell them yet. Since apple has them, apple knows how fast they are going to be.

Since they don’t sell them yet, none of the types of customers you mention could possibly have used them yet.

I am also a cpu designer, and I assure you that Arm is just as fast as x86, so long as the chip designer intends for the chip to be just as fast.

 

[dgdosen]

It wouldn’t provide contest other than “it will run a lot faster than this,” but the fact that people seem to think it would provide context is exactly why they won’t allow it.

99% of the people on here will pay no attention, see the benchmark, and start shouting “see! It runs slower than intel’s [insert chip here]” and then that narrative starts to spread and it’s all based on ********.

To me, running perf numbers on the DTK would show me the performance floor of future chips. I think that's valuable. Running benchmarks would help developers tune the relative performance of their apps - giving confidence their apps are properly tuned for Apple Silicon.

I don't think benchmarks would be used as a hammer on Apple, unless they were really awful. In fact, forbidding the running of benchmarks seems to make it look as if Apple is hiding something, and I don't think they are. I'd expect Benchmarks for an A12z to be pretty close to benchmarks we already see on for the 2020 iPad Pro because it is the same chip. It might be slower because of more overhead, but it might be faster seeing is the device would have better thermals. There is nothing wrong with transparency. My trust in Apple has eroded a bit over the years, but I have no reason not to believe they'll deliver on the idea of "incredible performance" with these new macs.

If this were some 'beta' shipping hardware - which it's clearly not, I'd understand Apple's secrecy, but this is a "kit", a hack that Apple rightly wants out there to help devs with a smooth transition. They'll be out there. We know those benchmark numbers will get out one way or another anyway - so why (attempt to) hide it?

 

[joeblough]

So how does Apple know how good it's performance is going to be??

you know you can do performance simulations of proposed CPU architectures, right? there are a lot of ways for them to know approximately what the performance of some unrealized design is.

anyway, they probably already have silicon for the desktop machines, but it is still in bringup.

 

[cmaier]

To me, running perf numbers on the DTK would show me the performance floor of future chips. I think that's valuable. Running benchmarks would help developers tune the relative performance of their apps - giving confidence their apps are properly tuned for Apple Silicon.

I don't think benchmarks would be used as a hammer on Apple, unless they were really awful. In fact, forbidding the running of benchmarks seems to make it look as if Apple is hiding something, and I don't think they are. I'd expect Benchmarks for an A12z to be pretty close to benchmarks we already see on for the 2020 iPad Pro because it is the same chip. It might be slower because of more overhead, but it might be faster seeing is the device would have better thermals. There is nothing wrong with transparency. My trust in Apple has eroded a bit over the years, but I have no reason not to believe they'll deliver on the idea of "incredible performance" with these new macs.

If this were some 'beta' shipping hardware - which it's clearly not, I'd understand Apple's secrecy, but this is a "kit", a hack that Apple rightly wants out there to help devs with a smooth transition. They'll be out there. We know those benchmark numbers will get out one way or another anyway - so why (attempt to) hide it?

We already have benchmarks of the a12z. Not much new to be learned. And the distance between this “floor” and the actual performance of the shipping product will be much larger than people assume, so Apple rightly doesn’t want people making bad assumptions.

 

[09872738]

The whole selling point about ARM processors is their ability to work at low power ratings. Anything that is heavy on the number crunching such as games, video rendering, audio rendering, some financial software is going to struggle on an ARM processor.

If people disagree then I would be interested to hear which movie studies use ARM pc's or music studios that use ARM pc's or financial companies involved in the stock market that use ARM pc's.

You are skating to where the puck is going to be, not where it has been. There once was a time where peope were certain flight „heavier than air“ is impossible. Just sayin‘

 

[gnasher729]

The whole selling point about ARM processors is their ability to work at low power ratings. Anything that is heavy on the number crunching such as games, video rendering, audio rendering, some financial software is going to struggle on an ARM processor.

If people disagree then I would be interested to hear which movie studies use ARM pc's or music studios that use ARM pc's or financial companies involved in the stock market that use ARM pc's.

The worlds fastest super computer is ARM.

Being able to run fast at low power means these chips are going to fly if that restriction is gone. Six core chips will beat all existing Macs up to eight processors. And it doesn't end there.

Today, no movie studios use Macs with ARM processor, and no financial companies do. That's going to change.
[automerge]1593097105[/automerge]

I am also a cpu designer, and I assure you that Arm is just as fast as x86, so long as the chip designer intends for the chip to be just as fast.

I'm a software developer, and I've written code that ran faster on an iPhone XR than on a quad core iMac. With all cores running. Even after running for fifteen minute and slowing down a bit due to heat, the 2 cores + 4 tiny cores on the iPhone XR were faster than four 2.8 GHz Intel cores. And the iPad chips are faster.

For the future: Right now the number of iPhones sold per year is about 220 million. PCs sold are 260 millions. Add a few million iPads, consider a few million PCs (named Macs) moving from Intel or AMD to ARM, and soon the number of Apple ARM devices will be higher than the number of x86 PCs sold, definitely higher than the number of Intel PCs sold. So Apple can afford to invest a lot of money into these chips, and over the last ten years they have been definitely more successful with this than Intel.

 

[gnasher729]

x86 won (for that time) because it was faster and would scale better than PPC (despite PPC being a RISC architecture).

I would say PowerPC lost because Motorola was only interested in use for low end routers (much too slow for a Mac), and IBM was only interested in server chips (where power and cost was of no concern, much too expensive and power hungry for a Mac). And at the time Apple was much too small and poor to develop its own PowerPC chips.

 

[cmaier]

I would say PowerPC lost because Motorola was only interested in use for low end routers (much too slow for a Mac), and IBM was only interested in server chips (where power and cost was of no concern, much too expensive and power hungry for a Mac). And at the time Apple was much too small and poor to develop its own PowerPC chips.

Steve shouldn’t have killed Exponential Technology, then. We had a high performance chip, and would certainly have been happy to develop a low power chip if that’s what apple asked us to do.

 

[Comed1an]

The whole selling point about ARM processors is their ability to work at low power ratings. Anything that is heavy on the number crunching such as games, video rendering, audio rendering, some financial software is going to struggle on an ARM processor.

If people disagree then I would be interested to hear which movie studies use ARM pc's or music studios that use ARM pc's or financial companies involved in the stock market that use ARM pc's.

I disagree with you and people like you that live in yesterday and are not really up to date on what is happening overall in this space.

The number one spot in super computers is as of now held by ARM based system (TOP 500 June 2020)

Also Amazon AWS has invested heavily into ARM based Graviton.

You are correct that it is difficult to say what companies are using ARM PC's, but the reason for this is that ARM PC's are not yet really readily available.

Apple is the first major company to start disrupting the legacy Intel x86 PC space.

 

[fishkorp]

The whole selling point about ARM processors is their ability to work at low power ratings. Anything that is heavy on the number crunching such as games, video rendering, audio rendering, some financial software is going to struggle on an ARM processor.

If people disagree then I would be interested to hear which movie studies use ARM pc's or music studios that use ARM pc's or financial companies involved in the stock market that use ARM pc's.

Umm, you do realize the world’s fastest super computer uses ARM chips, right? Please educate before making nonsensical statements.

 

[laptech]

What kind of spaghetti logic is this?

*apple* obviously has made Arm computers, and they have them in the labs, but they don’t sell them yet. Since apple has them, apple knows how fast they are going to be.

Since they don’t sell them yet, none of the types of customers you mention could possibly have used them yet.

I am also a cpu designer, and I assure you that Arm is just as fast as x86, so long as the chip designer intends for the chip to be just as fast.

If spaghetti logic is what you call it then it is of your own doing. All i did was repond to your post that said arms pcs have not been made yet, and i quote

No such companies use arm pcs because nobody has made arm pcs yet (other than a couple minor examples using crappy off-the-shelf arm chips).

Therefore for Apple to make the statements it has about just how good ARM processors are going to be and the performance improvements said processors are going to make is brash in my opinion and my opinion is still valid based on the point YOU made that no arm pc's have been made yet. Lab tests and real world test have always proven to be drastically different.

 

[cmaier]

If spaghetti logic is what you call it then it is of your own doing. All i did was repond to your post that said arms pcs have not been made yet, and i quote


Therefore for Apple to make the statements it has about just how good ARM processors are going to be and the performance improvements said processors are going to make is brash in my opinion and my opinion is still valid based on the point YOU made that no arm pc's have been made yet. Lab tests and real world test have always proven to be drastically different.

You are still using the same broken logic.

1) apple has made no quantitative statements about performance. They have only provided a demo of an a12z and let you judge for yourself.

2) you asked “if arm is so fast why don’t banks use it” - the answer is that nobody has yet sold such a machine

3) then you argue “if nobody sells such a machine, apple can’t know how fast such a machine is.” That’s the dumbest part of the argument. Apple *has* such machines. Just because they aren’t on sale yet doesn’t mean apple doesn’t know how fast they are.

When I designed CPUs I knew how fast they’d be while I was still designing them. Think about it - do you think chip designers just do their thing and then hope for the best? We have specific performance targets that we are meant to hit, and the job of designing a chip is making sure you hit those targets.

Now, if the chip designer knows how fast the chip is going to be before it even exists, don’t you think that Apple, which has the actual chips in actual boxes and can actually test it, knows how fast they are?

And, again, I reiterate, apple has NOT made any claims about performance other than “it’s fast!”

 

[gnasher729]

AMD and Intel have built and shipped multi-threaded CPUs - Apple hasn't. AMD and Nvidia have shipped discreet GPUs - Apple hasn't. Intel will beat them to the punch also.

Hyperthreading doesn't do much for performance (especially if you have optimised code in the first place). And in the recent years, it has been an absolute nightmare with security concerns. If I was in Apple's place, I'd keep my fingers far, far away from hyper threading. And hyper threading is much more significant for a single core than for a machine with say 8 cores.

 

[firewood]

I can one-up that. I designed a bipolar logic cpu using current mode logic.

In addition to building vacuum tube radio kits, the first logic circuits I wired up (including a JK flipflop) used DPDT relays recycled from junked pinball machines.

Also fairly steep uphill... both ways

 

[cmaier]

In addition to building vacuum tube radio kits, the first logic circuits I wired up (including a JK flipflop) used DPDT relays recycled from junked pinball machines.

Also fairly steep uphill... both ways

I once had to design and build a wire wrapped interface board to connect a paper tape reader to a PDP-11.

And I liked it.

 

[Alan Wynn]

I would say PowerPC lost because Motorola was only interested in use for low end routers (much too slow for a Mac), and IBM was only interested in server chips (where power and cost was of no concern, much too expensive and power hungry for a Mac). And at the time Apple was much too small and poor to develop its own PowerPC chips.

Exactly. See this history. To be clear, he never responded to my post, because it makes it clear that his argument is wrong, and that would force him to stop raising the same points over and over. Oh well. :-D

Amplifying this point, while the current system ranked number 1 on the top 500 list is ARM SoC based, the next two are PowerPC based. Intel is not in the top 5 (did not look to see how low they fall).

 

[hobowankenobi]

...that’s, that’s just not how any *consumer* makes electronic purchasing decisions today.

...notice the word: savvy. Sure, not the majority, but savvy users do tend to set the buying trends for lots of consumer goods.

 

[hobowankenobi]

The whole selling point about ARM processors is their ability to work at low power ratings. Anything that is heavy on the number crunching such as games, video rendering, audio rendering, some financial software is going to struggle on an ARM processor.

If people disagree then I would be interested to hear which movie studies use ARM pc's or music studios that use ARM pc's or financial companies involved in the stock market that use ARM pc's.

Data centers are all about efficiency and power/heat ratio, similar to 3D render farms and video production. I'll just leave this and this here. And Apple arguably has the best ARM CPUs.
[automerge]1593105000[/automerge]

Exactly. See this history. To be clear, he never responded to my post, because it makes it clear that his argument is wrong, and that would force him to stop raising the same points over and over. Oh well. :-D

Amplifying this point, while the current system ranked number 1 on the top 500 list is ARM SoC based, the next two are PowerPC based. Intel is not in the top 5 (did not look to see how low they fall).

...Just to add some flavor, PPC still gets used in higher-end laser printers too...best bang for the buck: power/heat/cost. ARM will probably take over this market as well, though it might take longer before it is cost-effective to port or re-write printer OS/firmware.

 

[djjeff]

I would say PowerPC lost because Motorola was only interested in use for low end routers (much too slow for a Mac), and IBM was only interested in server chips (where power and cost was of no concern, much too expensive and power hungry for a Mac). And at the time Apple was much too small and poor to develop its own PowerPC chips.

Can you support what you say? Or is it merely conjecture?

 

[cmaier]

Can you support what you say? Or is it merely conjecture?

As a PowerPC designer I will agree, for sure, that IBM really primarily cared about its workstations. Nothing in any of IBMs chips was ever designed for mobile - they didn’t use clock gaters, they didn’t use power supply regions, they didn’t modulate slew rates to minimize leakage, etc. They were always focused on fastest possible speed, which, frankly, we all were at the time, because that’s where the money was.

Motorola, I have no idea what their issue was. We never thought of ourselves as competing against Motorola, only against IBM (and indirectly intel). Motorola didn’t have a reputation in that time frame as being all that competitive in the area of CPU design anymore.

 

[djjeff]

Exactly. See this history. To be clear, he never responded to my post, because it makes it clear that his argument is wrong, and that would force him to stop raising the same points over and over. Oh well. :-D

Completely and 100% incorrect.
[automerge]1593112613[/automerge]

As a PowerPC designer I will agree, for sure, that IBM really primarily cared about its workstations. Nothing in any of IBMs chips was ever designed for mobile - they didn’t use clock gaters, they didn’t use power supply regions, they didn’t modulate slew rates to minimize leakage, etc. They were always focused on fastest possible speed, which, frankly, we all were at the time, because that’s where the money was.

This qualification is meaningless to the question asked. If you don't have knowledge then please avoid commenting. I want facts, not supposition or conjecture.

 

[cmaier]

Completely and 100% incorrect.
[automerge]1593112613[/automerge]

This qualification is meaningless to the question asked. If you don't have knowledge then please avoid commenting. I want facts, not supposition or conjecture.

I gave you the basis for my knowledge so you could determine its veracity and so the limits of my knowledge would be understood. I then told you which part of the statement I had knowledge about and which I did not.

I don't know what sort of evidence you would accept, but it's highly unlikely that any IBM employees with contemporaneous knowledge are willing to log in here and tell you what they were thinking at the time.

So, as someone who worked at Exponential Technology, a company funded by Apple and which made PowerPC chips in competition with IBM and Motorola, I was telling you what the understanding in silicon valley was at that time.

You can weigh that evidence how you'd like, but it's exceptionally rude to think you get to determine who has to "avoid commenting" and who does not.