iMac 2019 and ARM transition

[theluggage]

This was seemingly confirmed by remaining high-end RISC CPUs (IBM POWER, Oracle SPARC). They aren't generally any faster nor consume less power than Intel's top-end Xeons.

Not surprising, since current Intel processors are basically RISC-like cores with x86-to-RISC instruction decoders. The last "true CISC" processors were the Pentium 4/Netburst chips which Intel eventually abandoned as a dead end because of heat problems. Intel backtracked to the older Pentium Pro design to produce the Core range (which probably helped convince Apple to switch to Intel). I'm not sure that POWER or SPARC have ever been designed to be low-power...

(Its ancient history and of dubious relevance now, but the original ARM chips were designed as desktop processors and outperformed their Intel contemporaries while still using less power. Famous anecdote is that after successfully testing the first ARM chip, the developers realised that they hadn't connected the main power line to the chip, and it was just running off what it could pull from the address/data busses...)

This seems to imply that something fundamental has changed in A-series CPU design.

I think its more a case of proving the potential of building a bespoke CPU (or, rather, system-on-a-chip) for a device rather than building the device around whatever Intel deigns to produce. I suspect its also as much to do with the GPU and other bells-and-whistles built onto the A12x as the ARM CPU itself. The Macs that would benefit most from ARM will be the MB, the MBA and the 13" MBP that make sense as system-on-a-chip based designs - but then those are also Apple's most commercially valuable Macs.

For the "pro" Macs, currently sporting i9s and discrete GPUs, to successfully switch to ARM, Apple might need to bring something new to the table - like lots of cores and more specialist accelerator devices on-chip that tie into MacOS frameworks (Metal, Accelerate framework etc.) Thing is, like it or not, Apple could probably afford to take risks with the "pro" market (not updating the Mac Pro since 2013 is a near-terminal risk and will already have driven off anybody remotely ambivalent about MacOS).

It will be interesting to see what happens in respect of the Mythical Modular Mac Pro at WWDC - if nothing is announced then its dead, if they've sunk a lot of R&D into a Xeon system then they're probably planning to stick with Intel for a while, or there's the slim off-chance that they'll come up with some exotic ARM-based supercomputer with a zillion cores and specialist processing units etc. starting with a few highly-optimised Pro apps from Apple and strategic partners.

 

[theluggage]

ARM chips as usually designed are low-power chips.

...and Apple's #1 requirement is low-power systems-on-a-chip for their bread-and-butter MBA and 13" MBP models. The A12X in the top iPad Pro models is really proof-of-concept for that.

A similar Intel box cost about $4,000 but CPU and GUI performance were five times the SPARC box at one-fifth the cost. But the SPARC box ran Oracle RDBMS far better than the Intel box did. So it comes down to workload.

It also comes down to operating system type and tuning. Even if you're talking about an Intel box running the same flavour of Unix the kernel has a plethora of optimisation options to tune it for various workloads. If you're comparing Windows with Unix then all bets are off (and would be even on the same hardware). I once worked with VAX/VMS systems in a university where the systems had to be 'jack of all trades' - running Ingres, a DBMS written for Unix, brought a VMS system configured for general-purpose use to its knees - ISTR it was because Ingres forked subprocesses in a way which (regardless of hardware) was horribly slow in VMS. (NB: Windows NT was mainly written by the VMS guys...)

(Sure, by the same reasoning you have to be a bit careful comparing iPad Pro benchmarks to MacOS but those are closely related OSs both configured for interactive, high-graphics, GUI use).

As for GUI performance - there are precisely 3 *nix-type OSs with a decent GUI - MacOS, iOS and Android. Everything else is based on/derived from the chronically over-engineered X11 system and designed by people who think the point of a GUI is to run 8 instances of VIM in translucent windows over a Hubble space telescope image. p). Basically, want a server - use Linux, want a GUI, use MacOS - and that's nothing to do with CPU.

If it were the better mousetrap, though, Apple could just reintroduce the X-Serve with this processor and win the server market which would be worth a massive amount of money.

Take Apple's A12X chip and run Oracle, SAP, Genomic workloads, Apache, a few cloud workloads, etc. on it and let me know how it compares

Now you are being ridiculous. No sensible person would suggest using an A12X for server workloads - any more than they's suggest using the I5-8210Y from a MacBook Air - which is what an A12X might be a credible replacement for.

I think you want one of these instead:

https://www.marvell.com/server-processors/thunderx-arm-processors/

...that is designed for server use (not sure about Oracle or SAP but most of the open source web/cloud/DBMS stuff is already on ARM64).

Apart from Geekbench, the A12X has been benchmarked on things like compressing JPEGs and transcoding movies - plus user impressions of responsiveness - that are relevant to MacBooks.

 

[pshufd]

...and Apple's #1 requirement is low-power systems-on-a-chip for their bread-and-butter MBA and 13" MBP models. The A12X in the top iPad Pro models is really proof-of-concept for that.

It also comes down to operating system type and tuning. Even if you're talking about an Intel box running the same flavour of Unix the kernel has a plethora of optimisation options to tune it for various workloads. If you're comparing Windows with Unix then all bets are off (and would be even on the same hardware). I once worked with VAX/VMS systems in a university where the systems had to be 'jack of all trades' - running Ingres, a DBMS written for Unix, brought a VMS system configured for general-purpose use to its knees - ISTR it was because Ingres forked subprocesses in a way which (regardless of hardware) was horribly slow in VMS. (NB: Windows NT was mainly written by the VMS guys...)

(Sure, by the same reasoning you have to be a bit careful comparing iPad Pro benchmarks to MacOS but those are closely related OSs both configured for interactive, high-graphics, GUI use).

As for GUI performance - there are precisely 3 *nix-type OSs with a decent GUI - MacOS, iOS and Android. Everything else is based on/derived from the chronically over-engineered X11 system and designed by people who think the point of a GUI is to run 8 instances of VIM in translucent windows over a Hubble space telescope image. p). Basically, want a server - use Linux, want a GUI, use MacOS - and that's nothing to do with CPU.

Now you are being ridiculous. No sensible person would suggest using an A12X for server workloads - any more than they's suggest using the I5-8210Y from a MacBook Air - which is what an A12X might be a credible replacement for.

I think you want one of these instead:

https://www.marvell.com/server-processors/thunderx-arm-processors/

...that is designed for server use (not sure about Oracle or SAP but most of the open source web/cloud/DBMS stuff is already on ARM64).

Apart from Geekbench, the A12X has been benchmarked on things like compressing JPEGs and transcoding movies - plus user impressions of responsiveness - that are relevant to MacBooks.

I don't think of the 13 inch MBP models as low power but I only have a 2015 13 at the top CPU level. If it were easy, then they'd do it. I don't see it as being easy.

I was running Oracle on Windows vs Oracle on Solaris and the performance was like night and day. There are lots of companies that run Oracle on Windows but Windows was a second-class citizen back then and likely remains so today.

I used to do Ingres engineering support on Ultrix-32 when I was at DEC.

Windows NT had VMS source code. Dave Cutler took the team (either Seattle or Portland) to Microsoft. VMS + 111 = WNT. I received an email from Cutler when they had WNT up and running.

I'm responding to the idea that the A12X can do anything and everything and that it will take Intel out. I'd guess that someone would have to pay Oracle or SAP for a port.

I'm familiar with writing hand-coded assembler to do JPEG work myself. I haven't worked with video but I think that there are a lot of similarities. How does their performance compare to Intel's hand-coded SIMD libraries for multimedia processing?

 

[theluggage]

I'm responding to the idea that the A12X can do anything and everything and that it will take Intel out.

...but who is claiming that? The A12X was designed by Apple to drive the iPad Pro and might be suitable for something like a 12” MacBook. Other chipmakers are producing ARM-based chips for server applications that might end up running Oracle, probably under Linux. PostgreSQL is already on ARM64 Linux.

 

[pshufd]

...but who is claiming that? The A12X was designed by Apple to drive the iPad Pro and might be suitable for something like a 12” MacBook. Other chipmakers are producing ARM-based chips for server applications that might end up running Oracle, probably under Linux. PostgreSQL is already on ARM64 Linux.

After years of slowly closing the gap between itself and Intel, Apple can fairly claim to have slashed the difference between its ARM CPUs and Intel’s x86 chips to ribbons — at least in one test. Geekbench 4 reports an iPad Pro with a single-core score of 5030 and a multi-core score of 17995. This compares well against the Apple MacBook Pro — but not necessarily as well as you think.

https://www.extremetech.com/mobile/...-nearly-matches-top-end-x86-cpus-in-geekbench

The problem is for people that don't read the last sentence or the rest of the article.

New iPad Pro Benchmarked: This Blows Away Windows PCs
When Apple introduced the new iPad Pro, the company boasted that its slim slate is more powerful than 92 percent of PCs out there. Now that we've benchmarked the 12.9-inch iPad Pro for our review over at Laptop Mag, it looks like that claim could very well be legit.

https://www.tomsguide.com/us/new-ipad-pro-benchmarks,news-28453.html

Prior to Apple’s official unveiling of the third-generation iPad Pro this week, it was clear that a new A12X Bionic processor could finally match the speeds of current-generation MacBook Pro laptops if Apple was ready to pull the trigger. Now early Geekbench benchmark results are confirming as much: the iPad Pro rivals the latest Intel Core i7 CPU-equipped Apple laptops, seemingly despite significant price gaps.

https://venturebeat.com/2018/11/01/...s-rival-macbook-pros-with-intel-core-i7-cpus/

ARM processors like A12X are nearing performance parity with desktop processors

https://news.ycombinator.com/item?id=19325548

ARMed Attack: Intel And AMD Do Not See The Torpedo Headed Their Way

Now ARM is coming for them where it will really hurt — PCs and servers.

https://seekingalpha.com/article/4227086-armed-attack-intel-amd-see-torpedo-headed-way

2018 will be the year Arm finally gains a foothold in the datacenter. Qualcomm QCOM +0% announced Centriq in late 2017 (see my review here) to much fanfare. Cloud providers Azure and Alibaba BABA +0% contributed to this launch by announcing plans to deploy Centriq in cloud environments. Additionally, a robust ecosystem came out in support of Qualcomm QCOM +0%—from the OS ( Red Hat RHT +0%) to database (MariaDB), to cloud (Azure, Alibaba BABA +0%, Packet). Centriq’s performance-per-watt advantage will cause cloud providers to find use cases that support deployment.

My prediction: look for Qualcomm QCOM +0% to exit 2018 with 3% of the server market. While this may seem insignificant, it isn’t. Those that witnessed the uprising of x86-based servers in the datacenter can appreciate the difficulties of new architectures establishing relevance. Once the performance and power characteristics of Centriq are understood, cloud providers will deploy more broadly.

https://www.forbes.com/sites/moorinsights/2018/01/30/the-silicon-wars-heat-up-in-2018/#29af0aaa7ab1

One of the buzziest chip startups in recent memory is ready to start shipping its Arm-based data center processor, in hopes of carving out a foothold in a market dominated by Intel.

Ampere plans to announce later on Tuesday that it is shipping two varieties of its eMAG server processor, which could finally give backers of the Arm chip architecture a few wins in the data center market. Lenovo plans to release a server based on the processor, and cloud computing vendors will be able to obtain the chips in volume production to help determine if there’s room for another company inside the data centers of the world.

https://www.geekwire.com/2018/amperes-arm-based-server-chip-ready-challenge-intel-data-center/

The problem is that people that read articles or papers don't really understand the area so you get claims that are ridiculous. I've hung around the Linux community and remember when people said that it was going to take over the desktop. I recall Netscape saying that they were going to replace Microsoft on the desktop before they got killed by Internet Explorer. I remember a lot of tech claims made by executives of tech companies. And those claims by fans of particular software or platforms. And the articles about the A12X being faster than Intel chips gets fans think that benchmark results in a few areas means a lot more than it actually means.

 

[theluggage]

https://www.extremetech.com/mobile/...-nearly-matches-top-end-x86-cpus-in-geekbench

The problem is for people that don't read the last sentence or the rest of the article.

Well, the rest of the extremetech article just cherry-picks the specific sub-tests from the close-run A12X vs i7-8559U MBP Geekbench 4 where the i7 had a clear advantage and throws up a lot of FUD (maybe just accidental 'bias by balance' - I'm not saying there is anything untoward about the article ) about benchmark weighting that makes it sound like some statistical fudge. Its a general, synthetic benchmark that shouldn't be taken too seriously one way or the other.

The Tom's Hardware 'blows away windows PCs' headline is stupid, too, but arguably defensible in that their tests did show it soundly beating the Surface and XPS on all of their tests (including battery life).

There's no reasonable expectation that the tablet-orientated A12X itself needs to match a premium i7 laptop blow-for-blow, and we won't know for sure whether it does unless and until there is a A12X MacOS machine to test. If such a machine does appear its more likely to be a 12" MB replacement, which is a much softer target.

 

[giopiar]

As I said, following WWDC I ordered an i5 27-inch iMac with 256Gb SSD.

I’d like to share with you some considerations, based on your answers and several online sources. IMHO, in light of WWDC announcements, an ARM Mac is not going to happen soon. Nowadays Macs represent only a minimal percentage of Apple business, and I doubt that Apple is going to bear the hassle of managing tho different macOS architectures... Even less other devs... I think that if a transition ever happen, it is going to spread quickly over all Mac lineup, and recent Mac Pros, with Xeon processors and extremely high price tag, are going to stay here for a long time.

By the way, I’m now happy with my purchase because I feel like we’re going to see many macOS releases before my new Mac turns into an expensive paperweight!

 

[Freida]

As I said, following WWDC I ordered an i5 27-inch iMac with 256Gb SSD.

I’d like to share with you some considerations, based on your answers and several online sources. IMHO, in light of WWDC announcements, an ARM Mac is not going to happen soon. Nowadays Macs represent only a minimal percentage of Apple business, and I doubt that Apple is going to bear the hassle of managing tho different macOS architectures... Even less other devs... I think that if a transition ever happen, it is going to spread quickly over all Mac lineup, and recent Mac Pros, with Xeon processors and extremely high price tag, are going to stay here for a long time.

By the way, I’m now happy with my purchase because I feel like we’re going to see many macOS releases before my new Mac turns into an expensive paperweight!

have you not watched the presentation? Apple is doing all the steps to go ARM.
Catalyst is precisely the first step. Now that devs can literally code for mac with just a simple box tick its looking more likely than before that ARM is just around the corner.
I wouldn't be surprised if we see ARM computer next year and by 2022 all will be ARM.

 

[BigBoy2018]

have you not watched the presentation? Apple is doing all the steps to go ARM.
Catalyst is precisely the first step. Now that devs can literally code for mac with just a simple box tick its looking more likely than before that ARM is just around the corner.
I wouldn't be surprised if we see ARM computer next year and by 2022 all will be ARM.

x86 is dead! Long live x86!

Marco Armant speculated during the ATP WWDC live podcast that perhaps Apple would in fact, stay with x86, if for no other reason that it's a big, costly, customer alienating change ... and the fact that the Mac isn't nearly as important to Apple as it was in 2005, when the PPC to Intel transition happened. That earlier transition was critical to the health and well-being of Apple at that time ... there's no such need nowadays.

Additionally, one could argue that now that it's easy for developers to code for both the iPad and the Mac at the same time ... why bother with a big pain in the ass transition ... the big advantage of ARM (same apps across platforms) is largely taken care of by Catalyst.

 

[Freida]

x86 is dead! Long live x86!

Marco Armant speculated during the ATP WWDC live podcast that perhaps Apple would in fact, stay with x86, if for no other reason that it's a big, costly, customer alienating change ... and the fact that the Mac isn't nearly as important to Apple as it was in 2005, when the PPC to Intel transition happened. That earlier transition was critical to the health and well-being of Apple at that time ... there's no such need nowadays.

Additionally, one could argue that now that it's easy for developers to code for both the iPad and the Mac at the same time ... why bother with a big pain in the ass transition ... the big advantage of ARM (same apps across platforms) is largely taken care of by Catalyst.

Its simple, Apple's chips are becoming superior to anything else and Intel is struggling as it is which causes apple to be their slave. By going ARM not only that Apple can set itself free but can make a massive performance gap vs competition. That alone would get bigger market share and more customers overall. Imagine good computer from DELL, ASUS, LENOVO etc. and then similar machine from Apple that is 50% faster, way more secure etc.
What would the customer choose? Apple most likely, right? Apple could essentially kill the market with their advances just like they pretty much dominate the smartphone/tablet market.

 

[giopiar]

have you not watched the presentation?

Yes, I watched the presentation, and that’s the reason I think x86 in Mac may still have long life. I think project Catalyst is not about architecture, indeed it is a way to port simple apps to Mac sharing code with iPad counterpart. It doesn’t necessarily means that “native/real” Mac apps are going to die.

One day we’ll get to the truth, by now it’s all about speculations

 

[BigBoy2018]

Yes, I watched the presentation, and that’s the reason I think x86 in Mac may still have long life. I think project Catalyst is not about architecture, indeed it is a way to port simple apps to Mac sharing code with iPad counterpart. It doesn’t necessarily means that “native/real” Mac apps are going to die.

One day we’ll get to the truth, by now it’s all about speculations

Furthermore, even if Apple started to roll out ARM Macs even as soon as next year, it would be a good 2-3 years before they’re established enough to buy one.

And how long after that before all the mac programs you rely on have been ported to ARM ... if ever.

 

[Freida]

Yes, I watched the presentation, and that’s the reason I think x86 in Mac may still have long life. I think project Catalyst is not about architecture, indeed it is a way to port simple apps to Mac sharing code with iPad counterpart. It doesn’t necessarily means that “native/real” Mac apps are going to die.

One day we’ll get to the truth, by now it’s all about speculations

Yep, true. It will take time but I do think its heading there. Step by step Apple is aiming for that result.
Will be interesting to see what they will come up with.