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

[nutjob]

The driver is talking to is a discrete GPU connected via PCIe. There is no single dependence here to CPU architecture. So the code is _IDENTICAL_. On the lowest level you are getting all the relevant system information (e.g. bus, device and functions) via PCIe device enumeration and i am sure MacOS has an interface for that.

It's really not as simple as that. Different hardware has different characteristics, bottlenecks and requirements. Even if the graphics side of the driver is the same for the same Nvidia hardware, getting data to the card fast, reliably and without affecting system performance can be tricky for something very sensitive to latency like high performance graphics.

But anyway, that's all academic. If Apple doesn't give you permission for your drivers, you're out of luck. That's what Apple did for Nvidia drivers for Mojave. I wonder why they did that. Because it's such a benefit for users?

Video: Nvidia support was abandoned in macOS Mojave, and here's why | AppleInsider

Apple's macOS Mojave is a great software update for most users. But, it isn't for those who are using an Nvidia graphics card in their Mac Pro or inside of an external GPU enclosure, so let's talk about why.

appleinsider.com

 

[Kastellen]

Yeah, Who would want to buy them?

I would. So I can keep using Macs for a little longer before I have to leave the platform.
[automerge]1592880568[/automerge]

You guys will be fine - Us who were in the PowerPC to Intel transition feared the same thing, but survived

Apple was making their computers MORE compatible then, not less.

 

[burgerrecords]

1. This switch to ARM is unquestionably beneficial with regard to laptops both for the user and for apple. (i find apple notebooks to be reasonably priced for what you get, and of course this will only make them more appealing - I would guess that margins are currently lower on their laptops than other products - but this transition will increase those margins both directly and likely to spread costs over greater volume of units.)

2. Similarly Apple phones and their internet pads are a fair value based on relative performance to android - apple retains the profits for the efficiencies they create in iOS and with their custom chips but a new android that runs as well as a new apple isn't any cheaper than the iphone/ipad equivalent.

3. The iMac and Mac Mini are currently poor values particularly if you are someone who keeps their monitors over 2 or 3 computers and don't have dramatic size constraints. Arm transition may fix this as perhaps a mac mini or imac on ARM will provide comparable performance per $ to non-Apple desktop machines?

 

[damphoose]

When Apple transitioned from PowerPC to Intel 10.5.8 was the last Mac OS that would run on PowerPC. The Power Mac (PowerPC) was discontinued in August 2006 and 10.6 (Intel only) was introduced in August 2009. I'd say you'll get 3 years support at best. Maybe less.

it depends on what you mean by support. With the release of Snow Leopard I could no longer get the latest OS but I received maintenance updates to OS installed for about 5-6 years after. I remember pulling out an old macbook because I needed a DVD drive and received an update in 2012. Good times. I don’t have the negative feeling about the transition I am seeing from some here.

 

[Scottgfx]

It took Apple less than four years to obsolete the last G5s after the switch to Intel. So another full-on architecture change. For this of us with 2019 MBPs or Mac Pros this’ll be fun.

And it also means goodby Hackintosh users or unsupported Macs. Been fun having you.

I bet someone could make a Hackintosh out of one of these...

Avantek's Arm Workstation: Ampere eMAG 8180 32-core Arm64 Review

www.anandtech.com

 

[mdriftmeyer]

The limited Demos on a custom Mac Pro with a lot of RAM and Afterburner were slow.

This is the only time I'll reference my old boss, Steven P. Jobs. He would have not jumped the shark until he had a fully baked system with a matching performance MacBook Air against Intel MacBook Air running Big Sur Apple Silicon vs. Big Sur x86 was ready.

They truly jumped the shark on this one. They talked about legendary power improvements with ARM as if there was any other ISA architecture competing with ARM in the mobile space when the iPhone, iPad or AppleTV were launched. The competitor was Android OS vs. iOS, but both on ARM designs.

You have Big Sur, Snow Leopard for Mojave/Catalina with Services on steroids and Widgets 2.0, to everything is a rounded rectangle and mixed minimalist/photorealist icon UI that should have happened for Lion moving forward as an extended everything old is new again and then some for the UI. You finally have Photos Metal accelerated as if that should have happened in Mojave as somehow innovative.

You have talk of performance with demos in the lab were slow and not actual demos. Screen redraws were slow with the ARM motherboard inside the Mac Pro connected with an Afterburner. They show three basic Video streams in 4k at 30fps but don't do any actual work on them, beyond a quick scrub through.

Nothing in these demos were impressive. If they were impressive the team would have spec'd out the Mac Pro with how much Memory they used [They have a Dev Kit Mac Mini with 16GB RAM and 512SSD [as if they couldn't spare a 1TB] and that A12z in-house CPU to get people to port code for ARM executables.

What failure in the server market? Oh you mean the Intel failure. For example the new servers that AWS is rolling out use ARM SOCs, not Intel.

The new server instances are a few ARM SoCs of their own and more AMD EPYC instances. Look it up.

 

[entropys]

So...it starts with the Mac Mini.

Well not really. The A12z is an iPad chip. We haven’t seen the Mac versions yet.

 

[joeblough]

I bet someone could make a Hackintosh out of one of these...

Avantek's Arm Workstation: Ampere eMAG 8180 32-core Arm64 Review

www.anandtech.com

well i hope so but just like there are no iphone clones apple can certainly make it impossible to clone an ARM mac. especially since their SoC is, well, an SoC and not just a CPU.

 

[mdriftmeyer]

The AArch64 series instruction set has better support for numerical calculation than X64.

No it doesn't and besides the best support for Numerical Calculations in precision is Fortran, and that's platform agnostic.

 

[ChrisBos]

I had a Late 2013 rMBP which I updated last summer to the last 15" with an 8-core i9. This thing gets loud and hot when under a full load in something like Premiere.

The 16" is still a very good machine, especially coupled with the new 5600M. The benchmarks look quite outstanding but the i9 itself is the problem that I don't see changing anytime soon, although the i9 thermal throttling was definitely improved over the 15 I have (and a much better keyboard)

I don't disagree. But way more machine than I need even for future proofing, and way too large of a laptop for my liking.

At that price point, I'd personally grab whatever the equivalent of the $1299 iMac ends up being–which is also way more than I need but can at least double as a 4K media screen in my office/spare bed and enhance what little gaming I do, at a great price–and a standard config i3 Air for my portability needs.

But either way I don't think it's the right time for me to buy a long-term machine, and really am torn between riding out my 2015 MBP, or grabbing an Air as a bridge machine until ARM kinks get worked out.

 

[Alan Wynn]

I would love to work on the cpu for the next Mac Pro. Fun design problem.

Here you go! Granted if you take the job, we will not learn much from you any more.

 

[SteveW928]

I am old enough to remember the PPC > Intel transition. They backed that up because PPC was garbage in terms of performance compared to Intel. ...

It was more about getting processing power per watt. The performance didn't come until later when Intel started adding more cores. The G5 was beating anything Intel made by a good margin.

That said, a lot of people didn't have G5s (and Intel was ahead of the rest of Apple's line) and certainly for laptops.

... If you are a pro, you don't care about the silicon, only what it can do for you. Do you really care about Intel, AMD, PowerPC, or A12Z if it delivers your product to your clients faster than the last one with fewer bugs? ...

The problem is the software. Vendors of more specialized software were more likely to port to Mac on Intel, but even if they didn't, you could run it in virtualization or Bootcamp. It doesn't matter how awesome the hardware is if you can't get the software you need.

Apple's own GPU's. The whole point is not to have third party CPU/GPU's. All Apple Silicon. Tim even showed the 1000x GPU gains. Clearly no more Nvidia or AMD.

Hmm, I kind of doubt that. Obviously more more Intel graphics, but you're not going to get GPU performance like AMD/Nvidia unless Apple builds something similar. The GPU in your iPad Pro isn't even close.

I wonder if when we are in the ARM world if Desktop and Portable Macs will basically be the same performance.

No, because a desktop will always be able to apply more power/cooling, so higher performance chips will be made for them. You might raise the bar in how much performance you can stuff into a laptop, but the desktop can have many times more.

Soooo, you run 15+ year old software? You’re not operating the backend of a bank or a power plant or something are you?

AutoCAD is over 15 years old, and still widely used. There is actually a lot of over 15 year-old software used by professionals in all sorts of fields.

Imagine if Apple’s chips become so powerful that virtualized Windows performance is close to native Intel performance.

Unfortunately, it isn't a matter of imagine, but of necessity. If that isn't the case, a lot of us won't be able to pick Apple Silicon. We'll have to wait and see, but are these things going to be THAT much faster?

Because Pros want a complicated interface instead?

No, but an adequate one. The interface on an iPad doesn't cut it. Not complexity, capability.

 

[4jasontv]

Maybe they will launch a first gen low cost, sub $500 Mac Mini ARM that can run Final Cut Pro, DEVONThink server, and stream video to the rest of the house all at the same time.

 

[mdriftmeyer]

People who bought Mac Pros did so because they need the power, now. They could not wait for a couple more years for a possible new platform. They will surely pay for those Macs before an Arm-based replacement is ready. I would not be surprised if Apple couldn’t offer an Apple Silicon daughter card to pop into a Mac Pro to give it an upgrade.

And Apple spent 5 years developing them, then sold them as the future of Professional Workstations. Then this stunt happens. Instead of moving to AMD Zen and becoming best of breed now, they hope to be something in 3-5 years hoping the Joe Public is stupid enough not to realize that even Intel systems today won't remotely resemble those in 3-5 years and all the work Jim Keller's team have put in. The biggest payoff from Keller is AMD whose road map is solid for at least ten more years, living on cutting edge the entire time.

Apple did this to exploit more profit, not give a better product.

 

[4jasontv]

I am old enough to remember the PPC > Intel transition. They backed that up because PPC was garbage in terms of performance compared to Intel. They're going to have to give us all some serious benchmarks to prove the performance is there. Especially graphics.

So we can safely assume you are at least 15. /s

 

[mdriftmeyer]

It was more about getting processing power per watt. The performance didn't come until later when Intel started adding more cores. The G5 was beating anything Intel made by a good margin.

That said, a lot of people didn't have G5s (and Intel was ahead of the rest of Apple's line) and certainly for laptops.



The problem is the software. Vendors of more specialized software were more likely to port to Mac on Intel, but even if they didn't, you could run it in virtualization or Bootcamp. It doesn't matter how awesome the hardware is if you can't get the software you need.



Hmm, I kind of doubt that. Obviously more more Intel graphics, but you're not going to get GPU performance like AMD/Nvidia unless Apple builds something similar. The GPU in your iPad Pro isn't even close.



No, because a desktop will always be able to apply more power/cooling, so higher performance chips will be made for them. You might raise the bar in how much performance you can stuff into a laptop, but the desktop can have many times more.



AutoCAD is over 15 years old, and still widely used. There is actually a lot of over 15 year-old software used by professionals in all sorts of fields.



Unfortunately, it isn't a matter of imagine, but of necessity. If that isn't the case, a lot of us won't be able to pick Apple Silicon. We'll have to wait and see, but are these things going to be THAT much faster?



No, but an adequate one. The interface on an iPad doesn't cut it. Not complexity, capability.

FWIW: AutoCad is over 30 years old.

 

[4jasontv]

And Apple spent 5 years developing them, then sold them as the future of Professional Workstations. Then this stunt happens. Instead of moving to AMD Zen and becoming best of breed now, they hope to be something in 3-5 years hoping the Joe Public is stupid enough not to realize that even Intel systems today won't remotely resemble those in 3-5 years and all the work Jim Keller's team have put in. The biggest payoff from Keller is AMD whose road map is solid for at least ten more years, living on cutting edge the entire time.

Apple did this to exploit more profit, not give a better product.

I don't know if I agree with everything you said but I assume a base MacBook will get 8 years of functional life. I don't know if a pro machine should be more or less than that.

 

[Metrosey]

Because:

a) it's cheaper to make a single line of machines
b) Apple make better margins on their own chips than Intel's
c) it doesn't look good to continue selling a competitor's CPU when you claim yours is the future and the superior product

1. Both Intel and ARM machines are occurring at the same time anyway. Adding the choice onto each MacBook improves adoption rate (especially when ARM will be cheaper) while allowing more choice for the consumer.

2. Again, both continuing. As Apple gets better margins on ARM, it would be better to spread it across a large platform rather than one machine.

3. Again, Apple are releasing both ARM and Intel MacBooks.

 

[cmaier]

Here you go! Granted if you take the job, we will not learn much from you any more.

I’m more of a physical design engineer, logic designer, microarchitect, EDA or methodology engineer

 

[4jasontv]

Here you go! Granted if you take the job, we will not learn much from you any more.

Ugh! So close but I have a Ph.B. in Electrical Science. I wonder if they would still consider me.

 

[Alan Wynn]

So I just spent $14,000 for a NEW MAC PRO with an Intel processor. Hope there will be a motherboard for MAC PRO 2019 owners when Apple Processor is out.

What configuration? You do not expect to have amortized the cost by the end of next year? Certainly by the time its replacement ships, I would hope you recovered your costs. It is possible that there will be a CPU upgrade, but I would not count on it.

 

[mdriftmeyer]

I bet they mean apps like MATLAB, LabView, FEMLab, etc...

ANSYS, COMSOL, etc.

COMSOL: https://www.comsol.com/comsol-multiphysics

ANSYS: https://www.ansys.com

CATIA: https://www.3ds.com/products-services/catia/

PTC Creo (formerly Pro/Engineer): https://www.ptc.com/en/products/creo

DDS SolidWorks: https://www.solidworks.com

 

[Metrosey]

Yes.

But it also a way to not kill all your hardware sales today promising gear that is not ready to ship for quite some time.

Obviously folks will bench the heck out of the first new hardware; Apple can't hide from it or sweep it under the rug. They know that, and we all know that. If there are no substantial improvements to the cost/performance/heat/battery-life matrix...this transition will be a flop. There is nothing to bench yet (that we can buy or even see), so we have to wait.

Honestly...if they gave us benchmarks today, many would not believe them anyway, without production hardware that anybody could test and verify. Pointless.

Oh I agree, adding onto my previous point, from a presentation standpoint, it needs to be quick with a good flow. Showing benchmarks and numbers isn’t that interesting when using this presentation style. Furthermore, I doubt it’s been fully optimised yet, so they don’t want to provide numbers. Moreover, I agree with you, many would not believe them, but sometimes you do have to criticise first party benchmarks because they do like to handpick the best ones (fair enough from a business standpoint).

 

[SkyBell]

Hoping this means Intel Mac prices in the used market will crash - wouldn't mind picking up a Late 2015 iMac for cheap.

 

[NickName99]

And Apple spent 5 years developing them, then sold them as the future of Professional Workstations. Then this stunt happens. Instead of moving to AMD Zen and becoming best of breed now, they hope to be something in 3-5 years hoping the Joe Public is stupid enough not to realize that even Intel systems today won't remotely resemble those in 3-5 years and all the work Jim Keller's team have put in. The biggest payoff from Keller is AMD whose road map is solid for at least ten more years, living on cutting edge the entire time.

Apple did this to exploit more profit, not give a better product.

The Mac Pro is the best computer it can be right now. When Apple can make it better using their own processors, they will do so, most likely keeping the same enclosure design.

Apple knows Intel’s roadmap, they’re saying they’ll be transitioned to their own chips in 2 years, so it stands to reason that Apple knows they’ll overtake Intel within that time frame.