Solved! ARM Apple High-End CPU - Intel replacement

[Richie Rich]

There is a first rumor about Intel replacement in Apple products:

• ARM based high-end CPU
• 8 cores, no SMT
• IPC +30% over Cortex A77
• desktop performance (Core i7/Ryzen R7) with much lower power consumption
• introduction with new gen MacBook Air in mid 2020 (considering also MacBook PRO and iMac)
• massive AI accelerator

Source Coreteks:

 

[Eug]

It uses 45W CPU and 50W dGPU. 65W SoC will probably going to be easier to cool.

Oh I see what you mean.

I'm not convinced Apple will use only their own GPU for now. I could see Apple going back to AMD for the GPU in the high end models.

P.S. Note that the MacBook Pro is loud, and gets loud quickly, even with CPU-only tasks.

 

[JasonLD]

Oh I see what you mean.

I'm not convinced Apple will use only their own GPU for now. I could see Apple going back to AMD for the GPU in the high end models.

P.S. Note that the MacBook Pro is loud, and gets loud quickly, even with CPU-only tasks.

I think that is the problem every laptop can't avoid for now as long as they have active cooling.

I do think Apple might stick with AMD with their desktop lineup but I think they can scale their own GPU for laptops at least as good as their current offering. Apple doesn't need to make GPU as feature complete as AMD or Nvidia as they already have NPU for ML functions.

 

[ThatBuzzkiller]

Seriously?
Xerox? Not Altair? Not IBM? Not Apple (with the Apple II)? Not Commodore or Atari?
Not Zilog, or Intel, or Motorola, or MOS Technology?
Not Digital Research Inc?

I realize that it's an ill-posed question, and one can make an argument (a lousy argument, but an argument) that Xerox "created" the GUI; but claiming they invented the PC is a step way too far. Apple has a vastly better claim to that.

Hint: the personal computer began before 1984, and it even began before 1981.

Not arguing that the others like Intel didn't contribute but it was from Xerox where the concept of PCs had rightly originated and their integrated Alto system was the first to match the description of a multipurpose device fitted with a GUI that can be operated by an end user ...

 

[Doug S]

Not arguing that the others like Intel didn't contribute but it was from Xerox where the concept of PCs had rightly originated and their integrated Alto system was the first to match the description of a multipurpose device fitted with a GUI that can be operated by an end user ...

People who want to give Xerox all the credit need to watch the videos (you can find them on Youtube) showing what the Alto actually was. To say that Apple just copied that and there's the Mac ignores that it was rudimentary to the extreme. Yes, they did the hard part of rethinking the paradigm by creating the mouse and windowing, but by comparison the Alto makes Windows 1.0 look as polished as whatever you think is the best GUI in 2020. Apple had to add a lot to it to take it from a curiosity made by engineers for engineers into something that the average person could use.

 

[Gideon]

Apple had to add a lot to it to take it from a curiosity made by engineers for engineers into something that the average person could use.

It was sill total garbage compared to multitasking Amigaos released just a year later and remained begind for quite a while

 

[scannall]

They didn't create the "personal computer" that credit goes to Xerox's Alto system which went on to defining many of the features we see today in PCs ...

Apple was just another OEM for PCs ...

You really have no idea what you are talking about. At all. I was around back then, and obviously you weren't. While the Alto pre dates the Apple ][ series by a couple of years, they were not in the same market. Around 1,000 Alto's were built, for the price of $32,000 each. In 1979 dollars. About $113,000 each in 2020 dollars.

The Apple ][ series, that they made millions of was $1298 then. Roughly $5500 in 2020 dollars. Designed by primarily Steve Wozniak with help from a couple others, it is an all Apple product. There was nothing in the market resembling what we now think of as a PC at the time, that regular people could buy and use.

See the difference?

 

[ThatBuzzkiller]

People who want to give Xerox all the credit need to watch the videos (you can find them on Youtube) showing what the Alto actually was. To say that Apple just copied that and there's the Mac ignores that it was rudimentary to the extreme. Yes, they did the hard part of rethinking the paradigm by creating the mouse and windowing, but by comparison the Alto makes Windows 1.0 look as polished as whatever you think is the best GUI in 2020. Apple had to add a lot to it to take it from a curiosity made by engineers for engineers into something that the average person could use.

Sure the Alto may have had a somewhat cumbersome and primitive GUI but it had all the features in place that we take for granted today with current PCs such as office work, games (online functionality too!), online connectivity/communication (internet!), and you could even use it for programming too ...

What 'PCs' were introduced shortly afterwards were a massive downgrade including Apple's entry with their own systems ...

The Xerox Alto had it all in terms of system integration and advanced capabilities ...

 

[Gideon]

Sure the Alto may have had a somewhat cumbersome and primitive GUI but it had all the features in place that we take for granted today with current PCs such as office work, games (online functionality too!), online connectivity/communication (internet!), and you could even use it for programming too ...

What 'PCs' were introduced shortly afterwards were a massive downgrade including Apple's entry with their own systems ...

The Xerox Alto had it all in terms of system integration and advanced capabilities ...

While technically correct, for $32,000 each this is an entirely different market, it's like comparing commerial HEDT workstations to chromebooks.

Regardless, I don't get the 80-ies macs hype. Commodore Amiga lineup embarassed Macs in both performance and price and OS capability every step in the way. In graphics it was a total no-contest but despite having the same Motorola 68000 CPUs Amiga had a better hardware-architecture (and OS). For instance The late A2500s and the Mac IIci both had the MC68030, but an A2500 running an emulator was faster at being a IIci than a IIci itself was.

And It's also not like it was Wozniak in a garage vs the world. Commodore's business practices were abysmal and R&D budgets nonexsistant after Jack Tramiel left (at times having 1 person work on future hardware). Irvin Gloud absoultely tanked a company that could have stopped Wintel from gaining anywhere the prelevance they did during the 90ies.

I really recommend watching this Ahoy's video about the machine and what it could do. Will perhaps stop some people form idealizing 80ies macs as somehow being innovative or good value.

 

[soresu]

While technically correct, for $32,000 each this is an entirely different market, it's like comparing commerial HEDT workstations to chromebooks.

Regardless, I don't get the 80-ies macs hype. Commodore Amiga lineup embarassed Macs in both performance and price and OS capability every step in the way. In graphics it was a total no-contest but despite having the same Motorola 68000 CPUs Amiga had a better hardware-architecture (and OS). For instance The late A2500s and the Mac IIci both had the MC68030, but an A2500 running an emulator was faster at being a IIci than a IIci itself was.

And It's also not like it was Wozniak in a garage vs the world. Commodore's business practices were abysmal and R&D budgets nonexsistant after Jack Tramiel left (at times having 1 person work on future hardware). Irvin Gloud absoultely tanked a company that could have stopped Wintel from gaining anywhere the prelevance they did during the 90ies.

I really recommend watching this Ahoy's video about the machine and what it could do. Will perhaps stop some people form idealizing 80ies macs as somehow being innovative or good value.

Personally I loved my grandads ancient 20 MB HDD Amstrad from that era with it's 5.25 inch floppy drive, I think it might have been this.

 

[scannall]

While technically correct, for $32,000 each this is an entirely different market, it's like comparing commerial HEDT workstations to chromebooks.

Regardless, I don't get the 80-ies macs hype. Commodore Amiga lineup embarassed Macs in both performance and price and OS capability every step in the way. In graphics it was a total no-contest but despite having the same Motorola 68000 CPUs Amiga had a better hardware-architecture (and OS). For instance The late A2500s and the Mac IIci both had the MC68030, but an A2500 running an emulator was faster at being a IIci than a IIci itself was.

I like the Macs from that era, but you're right. The Amiga was a great machine. Had a few of them, including a couple with a Video Toaster setup. The problem was, Commodore owned it. Had they been able to stay independent, or be taken over by just about anyone other than Commodore it might be a different computer world out there.

 

[Eug]

I like the Macs from that era, but you're right. The Amiga was a great machine. Had a few of them, including a couple with a Video Toaster setup. The problem was, Commodore owned it. Had they been able to stay independent, or be taken over by just about anyone other than Commodore it might be a different computer world out there.

Maybe it was our personal bias, but at that time it seemed to me and my friends the only point of the Amiga was Video Toaster. Otherwise the Amigas and the Atari ST line were dead ends.

The real mainstream options to us were the Apple ][e (or //c), the IBM PC, or the Macintosh. I personally skipped the entire Macintosh line though until the 21st century, when Mac OS X* came out.

With regards to the Atari ST, maybe it was the reputation of the Atari 1200XL that doomed it for us. Perhaps I was esp. biased because I had originally salivated over the 1200XL hardware, until I realized it was going to be a sales and support disaster.

*PS. Its only been two decades but now we finally will get macOS 11 this year.

 

[soresu]

I personally skipped the entire Macintosh line though until the 21st century, when Mac OS X* came out.

Apparently you and everyone else - given the base for OSX was an entirely different kettle of fish it's probably not even worth mentioning those 2 in the same sentence.

It's probably only the stubborn pride/egomania streak of Jobs that led to it being called iMac in the first place after the lackluster market performance of the Macintosh and his departure from Apple back then.

 

[soresu]

While technically correct, for $32,000 each this is an entirely different market, it's like comparing commerial HEDT workstations to chromebooks.

Wut!!!

What HEDT workstations have you been buying and from where?!

$32,000 back then was well over $150,000 today.

That's more like comparing a small business supercomputer/render farm today to a single HEDT workstation.

 

[Gideon]

Wut!!!

What HEDT workstations have you been buying and from where?!

$32,000 back then was well over $150,000 today.

That's more like comparing a small business supercomputer/render farm today to a single HEDT workstation.

True that. I meant the relative price difference though, rather than actual price (2000-3000$ vs 200-300$) but tyour example is more accurate.

 

[Doug S]

Sure the Alto may have had a somewhat cumbersome and primitive GUI but it had all the features in place that we take for granted today with current PCs such as office work, games (online functionality too!), online connectivity/communication (internet!), and you could even use it for programming too ...

What 'PCs' were introduced shortly afterwards were a massive downgrade including Apple's entry with their own systems ...

The Xerox Alto had it all in terms of system integration and advanced capabilities ...

At the price of a small house (not in Silicon Valley, but in many parts of the US) in the late 70s I should hope it was an upgrade over PCs costing less than a tenth as much. Building something in a "cost is no object" fashion is one thing, building something affordably enough that it has a potential market in the millions takes a totally different set of skills - you know you will have to cut out some features or refactor parts of the design to reduce cost.

The skill to do that is equally difficult to master as designing the cost is no object product in the first place. If you look at "cheap clone" versions that always follow on the heels of new expensive product classes you will see a lot of ways how NOT to do those things right.

If I give you a plot of land and some money to build a house, the house you can build if I give you $100,000 is very different from the house you can build if I give you $1 million. You can't complain that the $100K house is a "massive downgrade" from the $1 million one, that's kind of impossible to avoid with such a big difference in resources. If you insisted on making the $100K house 4000 sq ft in size to match the $1 million house you'd end up with a pole barn with one bathroom and no kitchen, if you insisted on making the $100K house to the same standard of fit and finish as the $1 million house you'd end up with a 150 sq ft "tiny home" that wouldn't work if you have or ever plan to have kids.

 

[soresu]

At the price of a small house (not in Silicon Valley, but in many parts of the US) in the late 70s I should hope it was an upgrade over PCs costing less than a tenth as much. Building something in a "cost is no object" fashion is one thing, building something affordably enough that it has a potential market in the millions takes a totally different set of skills - you know you will have to cut out some features or refactor parts of the design to reduce cost.

There's also time based cost reduction in play there - this was still during the heyday of Moore's Law, so the reduction in cost (and size/volume) for the same performance in 10 years would be pretty significant to say the least.

 

[Doug S]

There's also time based cost reduction in play there - this was still during the heyday of Moore's Law, so the reduction in cost (and size/volume) for the same performance in 10 years would be pretty significant to say the least.

For the hardware that's affected by Moore's Law, yes. Software doesn't obey Moore's Law though.

One thing I've always wondered about the original Apple/Xerox deal is exactly how many stock options Xerox was given and what they sold them for. I wonder how many tens or hundreds of millions - or perhaps billions - they'd be worth today if Xerox had held onto them all this time?

 

[scannall]

For the hardware that's affected by Moore's Law, yes. Software doesn't obey Moore's Law though.

One thing I've always wondered about the original Apple/Xerox deal is exactly how many stock options Xerox was given and what they sold them for. I wonder how many tens or hundreds of millions - or perhaps billions - they'd be worth today if Xerox had held onto them all this time?

Rough back of the napkin math puts it at about 7.5 billion.

 

[Doug S]

Rough back of the napkin math puts it at about 7.5 billion.

Considering Xerox's current market cap is less than half that, I guess we'll file it under "missed opportunity"

 

[Eug]

Considering Xerox's current market cap is less than half that, I guess we'll file it under "missed opportunity"

Tell me about it... I bought at a little over $50 and then sold at around $80. I thought I did well with that 50-60% return.

That stock has since had a 7:1 stock split and now is $383.

In other words, had I kept it, it'd be worth about 50X what I paid for it. Instead, I cashed out when it was worth 1.5X what I paid for it.

 

[JasonLD]

Keeping the thread on topic, here is Dirt Rally and Cinema4D running through Rosetta 2

 

[name99]

Keeping the thread on topic, here is Dirt Rally and Cinema4D running through Rosetta 2

So, for someone who does not use either of these apps, how do these compare to native on say a MacBook or MacBook Air (ie a low-end mac, matching what will be below low-end shipping Apple silicon).
Are these apps limited by iGPU or by CPU? And are the scenes being demoed considered demanding (ie this shows great performance) or is the point more jut that the apps work? If you took scenes that made a MacBook struggle, are they still fluid on the Rosetta2 dev Apple Silicon?


What I would consider interesting (approaching things from a different direction) is the performance of the Mathematica benchmarks. The last time I did this, comparing my iMac Pro to my iPad Pro (running Wolfram Player) a few things were clear:

- Wolfram Player was clearly not doing a lot of performance stuff (like supporting the Mathematica Parallel constructs). This (like lots of horrible choices in Wolfram Player) seems to be a decision to try to prevent Player (free) from competing with Mathematica ($$$)

- Wolfram Player was clearly not using NEON for many purposes where it would make sense. Unclear whether this reflects: compiler sucks, they didn't get round to it yet, or again they don't want to compete with Mathematica

- Wolfram Player has terrible bignum performance. This one does not seem to reflect a deliberate choice not to compete with Mathematica. Mathematica uses GNU bignum, and that has been ported to ARM, but I don't know if ARM is using an old version, or the current ARM version sucks. Maybe Apple has (or will) get round to improving this as part of their open source on ARM effort.

Anyway, with those caveats in mind, what I found was that for everything that didn't fall into those categories the iPad Pro was pretty much equivalent to the iMac Pro; sometimes a little worse, usually a little better.

So I'd expect Rosetta Mathematica to run at maybe 80% of native speed and so ~70 to 80% of iMac Pro (optimistic? but it's running a lot of loops...); but hopefully there will be a Native Mathematica by end of the year which should crush the iMac Pro -- I'd expect 1.5 to 2x faster pretty much across the board.

 

[Hitman928]

So, for someone who does not use either of these apps, how do these compare to native on say a MacBook or MacBook Air (ie a low-end mac, matching what will be below low-end shipping Apple silicon).
Are these apps limited by iGPU or by CPU? And are the scenes being demoed considered demanding (ie this shows great performance) or is the point more jut that the apps work? If you took scenes that made a MacBook struggle, are they still fluid on the Rosetta2 dev Apple Silicon?


What I would consider interesting (approaching things from a different direction) is the performance of the Mathematica benchmarks. The last time I did this, comparing my iMac Pro to my iPad Pro (running Wolfram Player) a few things were clear:

- Wolfram Player was clearly not doing a lot of performance stuff (like supporting the Mathematica Parallel constructs). This (like lots of horrible choices in Wolfram Player) seems to be a decision to try to prevent Player (free) from competing with Mathematica ($$$)

- Wolfram Player was clearly not using NEON for many purposes where it would make sense. Unclear whether this reflects: compiler sucks, they didn't get round to it yet, or again they don't want to compete with Mathematica

- Wolfram Player has terrible bignum performance. This one does not seem to reflect a deliberate choice not to compete with Mathematica. Mathematica uses GNU bignum, and that has been ported to ARM, but I don't know if ARM is using an old version, or the current ARM version sucks. Maybe Apple has (or will) get round to improving this as part of their open source on ARM effort.

Anyway, with those caveats in mind, what I found was that for everything that didn't fall into those categories the iPad Pro was pretty much equivalent to the iMac Pro; sometimes a little worse, usually a little better.

So I'd expect Rosetta Mathematica to run at maybe 80% of native speed and so ~70 to 80% of iMac Pro (optimistic? but it's running a lot of loops...); but hopefully there will be a Native Mathematica by end of the year which should crush the iMac Pro -- I'd expect 1.5 to 2x faster pretty much across the board.

I can only comment on Dirt Rally because I don't use Cinema4d or do rendering work. Dirt Rally is very light on CPUs, like any CPU with at least 2 cores and 4 threads from the last 8 years could probably run the game very well. It's not too harsh on GPUs either but you'll actually see a hierarchy of performance on GPUs as you crank the settings and (obviously) if you bring it to 4K resolution.

 

[ThatBuzzkiller]

What's the refresh rate like on all of the Rosetta 2 demos ?

On second thought, Apple's 'tiler' GPU design will end up being a bigger issue for game performance than the translation layer ...

 

[tamz_msc]

I can only comment on Dirt Rally because I don't use Cinema4d or do rendering work. Dirt Rally is very light on CPUs, like any CPU with at least 2 cores and 4 threads from the last 8 years could probably run the game very well. It's not too harsh on GPUs either but you'll actually see a hierarchy of performance on GPUs as you crank the settings and (obviously) if you bring it to 4K resolution.

Dirt Rally is GPU-bound in the benchmark and rally stages, but is more CPU-bound in Rallycross.