New References to AMD Processors Discovered in macOS 10.15.4 Beta

[koyoot]

Intel built really good ARM CPU back in the day in the name of Xscale. At that time it was the best ARM chip just like how Apple doing it right now.
They are really good indeed and even better than their later designed Atom CPUs under same power envelope.

Intel feared ARM advantage will kill its x86 monopoly so they stop the development and sold the design to Marvell.
Intel themself feel threatened by ARM and you still blindly believe into their 40 years legacy technology.

Intel FELT threatened by ARM in LOW-POWER Thermal envelopes. ARM is good at them but not in High thermal envelopes. All of its advantages from being designed for low-power are being mitigated in High-Thermal envelopes, and that is no speculation, that is a fact.

x86 for High-Performance will be the solution for the foreseeable future despite anything people will say about it. I was for ARM, heck I was ARM fanboy, back in the day, but since we're in 2020 and there is no ARM solution that can come close in terms of Clock speeds, core counts, for high-thermal envelopes. ARM and x86 can coegsist, but one will not replace the other.

P.S. If Intel felt threatened by ARM, why AMD does not feel that way, at all?

 

[dgdosen]

1) Cook is widely acknowledged as a supply chain genius, and you claiming otherwise does not make it so.

2) You don’t know what Apple is paying for Qualcomm modems, so you can’t claim Apple is overpaying.

3) Whatever chip Apple puts in MBA this year, next year or any year subsequent, there’s no reason to think Apple would overpay.

4) You not understanding how Apple uses multiple suppliers and other strategies to minimize costs and keep selling prices lower for customers is a you problem. It doesn’t change the facts. You not liking the facts isn’t relevant.

While I have no knowledge of Tim Cook's mental capacity - I think you're a bit misguided as to why Apple minimizes costs - They do it to maximize profit. They only care about not losing customers, and maximizing the revenue they can extract from any customer. Stone cold. If they wanted customers to enjoy lower prices they would:
- Sell hardware that can be upgraded
- Sell hardware that can be repairable
- Sell hardware that doesn't have batteries with expected lives of a year

By that logic - we still come to the same conclusion - that Apple will, if not migrate to AMD, use the threat of migration to AMD to negotiate a better deal with any Intel products.

We can assume Apple is doing the due diligence to kick the tires on a move to AMD. And I keep thinking - if their long term plan is to axe x86-64 and move to their own ARM chips anyway, why not enjoy that transition and temporarily move to AMD? As long as AMD was honest in their Ryzen 4000 keynote - I can't think of a reason NOT to. Especially if MS and Dell are already testing the waters.

 

[theluggage]

I think the idea of an Extinction Event is being over dramatic!

Each transition broke binary compatibility for older software, with emulation as an inferior alternative - the transition to Mac OS X meant switching to a totally new OS and broke source compatibility. Each time, anything no longer being actively maintained or that was uneconomic to update went extinct.

In this case the cost in building more powerful systems need for electricity and cooling.

That's one major factor, but not the only one. There's also the issue with Intel's problems moving to new processes, consequent supply shortages and the way that the specific chip models that Apple needs aren't always Intel's highest priorities. The separate problem is that, here and now, Intel has a weak spot in it's line up between i7/i9 and Xeon that is currently being aggressively filled by AMD.

ARM is low-power, yes, but with ARM, Apple gain the ability to design custom systems-on-a-chip with exactly the power/number of cores/GPU capability/other on-chip toys that they need.

Yes, ARM is a great design but it does have its limitations.

The current ARM CPUs/SOCs have their limitations because they were designed for mobile use. Apple can design their own processors - ARM sell licenses for everything from just the instruction set, though individual processor cores to complete SOCs. I doubt you'd see an A13 in anything bigger than a MacBook Air, if at all. If the Mac/iMac Pro ever went ARM, imagine lots more cores and/or afterburner technology on the same chip.

Besides low power, it's the pick'n'mix nature of ARM (if you're a Serious Caller like Apple) that makes it unique.

No matter, would a move to AMD be a true extinction event? It's still x86-64.

AMD would be the easiest move to date since running existing Intel code is a very high priority for AMDs designers. The Hackintoshers already have MacOS running on AMD with a few caveats. Most intel-specific features can and should be fixed at OS level (e.g. ISTR power management doesn't work on AMD Hackintsoh and has to be disabled) and I'm sure that there will be some apps that will run disappointingly on AMD until they have been optimised - but much, much easier than a totally new ISA.

To re-iterate: the story that started this thread is probably bunkum - just some AMD graphics driver code - but the Intel vs AMD vs ARM debate is still a valid one.

 

[dgdosen]

Bitcode is arch dependent. So that's actually not a great idea. But just think about how Apple did it 15 year ago and how iOS running same kernel and almost same framework on ARM64 you can imagine a future transitioning will work out.

Are you sure that's true?

For example - here's an NSHipster post (on a somewhat different subject):

Language Server Protocol

Apple’s adoption of LSP is arguably the most important decision they’ve made for Swift since releasing the language as open source in 2014. It’s a big deal for app developers, and it’s an even bigger deal for Swift developers on other platforms.

nshipster.com

That chart sure makes it look easy to build ISA independent bitcode - and I hold that author in pretty high regard.

 

[MikeZTM]

Intel FELT threatened by ARM in LOW-POWER Thermal envelopes. ARM is good at them but not in High thermal envelopes. All of its advantages from being designed for low-power are being mitigated in High-Thermal envelopes, and that is no speculation, that is a fact.

x86 for High-Performance will be the solution for the foreseeable future despite anything people will say about it. I was for ARM, heck I was ARM fanboy, back in the day, but since we're in 2020 and there is no ARM solution that can come close in terms of Clock speeds, core counts, for high-thermal envelopes. ARM and x86 can coegsist, but one will not replace the other.

P.S. If Intel felt threatened by ARM, why AMD does not feel that way, at all?

OK. So you still didn't get it.

Atom 1st gen was built using 80486 legacy architecture because CISC to RISC (micro op) translator build in all Intel CPUs use more energy than a ARM CPU as a whole.

Just remove this CISC to RISC translator we will have a better Intel CPU at sightly lower power consumption. This is a fact.

And also simple math for you. Graviton 1 core to core is 50% of a Xeon. It use much less than 50% power and is cheaper than 50% of a Xeon per core.

Why this is worse for a heavily scalable workload?

BTW AMD had skybridge project but due to what every reason they killed it just like Intel killed Xscale. Coincidence?

Power efficiency is scalable and what's the best scenario for scalable computing? Servers.
And guess what's intel's most profitable market? Servers.

They fear ARM and you fear nothing. Sure they should let you build their roadmap right?

 

[smulji]

Nothing is 100% confirm but that's a bad deal if Apple put the icelake-y that doesn't offer any major improvements.

There are two major improvements:

1. Quad-core at the same TDP
2. Big improvement in the iGPU

 

[PickUrPoison]

While I have no knowledge of Tim Cook's mental capacity - I think you're a bit misguided as to why Apple minimizes costs - They do it to maximize profit. They only care about not losing customers, and maximizing the revenue they can extract from any customer. Stone cold. If they wanted customers to enjoy lower prices they would:
- Sell hardware that can be upgraded
- Sell hardware that can be repairable
- Sell hardware that doesn't have batteries obsolete in a year

By that logic - we still come to the same conclusion - that Apple will, if not migrate to AMD, use the threat of migration to AMD to negotiate a better deal with any Intel products.

We can assume Apple is doing the due diligence to kick the tires on a move to AMD. And I keep thinking - if their long term plan is to axe x86-64 and move to their own ARM chips anyway, why not enjoy that transition and temporarily move to AMD? As long as AMD was honest in their Ryzen 4000 keynote - I can't think of a reason NOT to. Especially if MS and Dell are already testing the waters.

Apple makes their margin, which has been essentially flat for many years. If their costs go up $65 dollars, they’ll tack on $100 to the selling price. If they can cut $65 in cost, they can sell $100 cheaper. Simple.

If Apple is moving to ARM, all the more reason not to move to AMD. They’ll simply extract better prices from Intel, and pass the cost savings to customers. That could take the form of a price reduction, or a higher spec’ed product at the same price.

 

[MikeZTM]

Are you sure that's true?

For example - here's an NSHipster post (on a somewhat different subject):

Language Server Protocol

Apple’s adoption of LSP is arguably the most important decision they’ve made for Swift since releasing the language as open source in 2014. It’s a big deal for app developers, and it’s an even bigger deal for Swift developers on other platforms.

nshipster.com

That chart sure makes it look easy to build ISA independent bitcode - and I hold that author in pretty high regard.

LLVM IL is arch dependent but not strictly arch dependent.
armv7 IL is mostly compatible with arm64. Apple Watch is already using this compatibility now since S4 is arm64 only and do not have ability to run armv7 code.

Compatibility between arm64 and x86 is currently still not finished yet and this doesn't looks like useful--especially by this slow speed of progress.
Right now Mac AppStore still didn't take off and still doesn't mandatory Bitcode. Apple Watch was requiring Bitcode support for all apps from start.

I think they may just push everyone to Catalyst platform and just run iOS framework(or maybe run iPadOS app natively) and let developer to port their AppKit apps themself.

 

[smulji]

Apple makes their margin, which has been essentially flat for many years. If their costs go up $65 dollars, they’ll tack on $100 to the selling price. If they can cut $65 in cost, they can sell $100 cheaper. Simple.

If Apple is moving to ARM, all the more reason not to move to AMD. They’ll simply extract better prices from Intel, and pass the cost savings to customers. That could take the form of a price reduction, or a higher spec’ed product at the same price.

Apple doesn't have to extract lower prices. That's already beginning to happen without Apple

Intel officially craters Xeon pricing

SemiAccurate has been warning that Intel is slashing prices to OEMs for months now, something that is now official.

semiaccurate.com

 

[MikeZTM]

There are two major improvements:

1. Quad-core at the same TDP
2. Big improvement in the iGPU

Intel Icelake is not more power efficient than Whiskey lake except you running AVX512.
So same TDP but less performance in general.

iGPU still much worse than AMD's offering.

 

[cmaier]

As a Ultra SPARC T1 user I hate Intel not innovating back in 2000s.
And thanks for the details. I never thought about chopping off the x86 compatibility stuff could just make a Intel CPU an ARM one.

And also a DEC Alpha machine was my childhood dream that I never got.

I worked on UltraSparc V. I was responsible for the register file renaming unit, as I recall, and the global floor planning methodology. I quit after one month and went to AMD. Friggin' Sun.
[automerge]1581356515[/automerge]

Atom 1st gen was built using 80486 legacy architecture because CISC to RISC (micro op) translator build in all Intel CPUs use more energy than a ARM CPU as a whole.

Eh. I have no particular knowledge of how Intel does things, but my rough estimate based on my work on x86 cpus (which may be out of date now) was that the instruction decoding complications (microcode rom, sequencer, load/store stuff, etc.) are maybe 20% of the transistors for a given core. Probably less nowadays since cores are becoming more complicated with multithreading, etc. So while it may use more power than an ARM core, it would be a pretty slow ARM core. More like 15% or something is my rough guess.

 

[feldpos]

I don't know if Apple is listening but if you make Ryzen laptops/Mac Minis (and fix the damn keyboard) I will order the instant you announce it.

 

[koyoot]

OK. So you still didn't get it.

Atom 1st gen was built using 80486 legacy architecture because CISC to RISC (micro op) translator build in all Intel CPUs use more energy than a ARM CPU as a whole.

Just remove this CISC to RISC translator we will have a better Intel CPU at sightly lower power consumption. This is a fact.

And also simple math for you. Graviton 1 core to core is 50% of a Xeon. It use much less than 50% power and is cheaper than 50% of a Xeon per core.

Why this is worse for a heavily scalable workload?

BTW AMD had skybridge project but due to what every reason they killed it just like Intel killed Xscale. Coincidence?

Power efficiency is scalable and what's the best scenario for scalable computing? Servers.
And guess what's intel's most profitable market? Servers.

They fear ARM and you fear nothing. Sure they should let you build their roadmap right?

AMD killed project Skybridge because it was financially nonfeasible. The power/performance ratio was simply not there.

I will put the bar even higher. IF ARM would be that Be-all, End-All product as you picture it, AMD would invest it in years ago! And don't tell me about "AMD does not have skilled enough Engineers...". They have - which Zen architectures has been proving.

Intel feared ARM not because of performance, but power efficiency, and costs. This is the point.

In the context of this debate. There is no such thing as a Free Lunch. And you should know this perfectly well.

 

[PickUrPoison]

Apple doesn't have to extract lower prices. That's already beginning to happen without Apple

It’s less that it’s beginning to happen without Apple and more that it’s been happening for years for Apple (and others).

Thanks to competitive pressure from AMD, Intel’s been discounting for awhile now. It’s the point I was making to that brainwashed dude in post #347. Apple isn’t going to pay crazy prices when they can get competitive product for much less.

Intel enjoyed monopoly pricing power for many years. Now that they have serious competition, they’ve got to be price competitive.

 

[MikeZTM]

I worked on UltraSparc V. I was responsible for the register file renaming unit, as I recall, and the global floor planning methodology. I quit after one month and went to AMD. Friggin' Sun.
[automerge]1581356515[/automerge]

Eh. I have no particular knowledge of how Intel does things, but my rough estimate based on my work on x86 cpus (which may be out of date now) was that the instruction decoding complications (microcode rom, sequencer, load/store stuff, etc.) are maybe 20% of the transistors for a given core. Probably less nowadays since cores are becoming more complicated with multithreading, etc. So while it may use more power than an ARM core, it would be a pretty slow ARM core. More like 15% or something is my rough guess.

80486 part was from a media interview artical from a Intel engineer from that time.

And I heard they use 496 compiler flag to optimize code for atom.

 

[cmaier]

80486 part was from a media interview artical from a Intel engineer from that time.

And I heard they use 496 compiler flag to optimize code for atom.

Ok. Now that cores are much more complicated (out-of-order issue and retire, 64-bit, threading support, complicated branch predict, etc.) it's probably a smaller percentage.

The hardware to decode x86 (especially 32-bit) is in many ways similar to a tiny little cpu on its own.

 

[MikeZTM]

AMD killed project Skybridge because it was financially nonfeasible. The power/performance ratio was simply not there.

I will put the bar even higher. IF ARM would be that Be-all, End-All product as you picture it, AMD would invest it in years ago! And don't tell me about "AMD does not have skilled enough Engineers...". They have - which Zen architectures has been proving.

Intel feared ARM not because of performance, but power efficiency, and costs. This is the point.

In the context of this debate. There is no such thing as a Free Lunch. And you should know this perfectly well.

Power efficiency and cost means performance.
There's no such thing has poor performance when power efficiency and cost are all good.

Told you. Simple math. 50% performance , less than 50% cost and less than 50% power.
Just buy 2 you got 100% performance and less than 100% cost and power.

Sure not all workloads are scalable but hey this is server workload. It's highly scalable.

I'm a full stack software developer. I understand what you thinking because I'm earning money from that kind of thought -- Enterprise customer usually lack any knowledge of technology but have strong preference of technology they didn't even understand. Sure most of their preference are "time proven" but most of them doesn't make any sense and nowhere near cost effective.

That's why Java is still THE first class enterprise platform even with all those ugly business action Oracle did.

You should know more than those who are only target of us ripping money from.

 

[Zdigital2015]

While I have no knowledge of Tim Cook's mental capacity - I think you're a bit misguided as to why Apple minimizes costs - They do it to maximize profit. They only care about not losing customers, and maximizing the revenue they can extract from any customer. Stone cold. If they wanted customers to enjoy lower prices they would:
- Sell hardware that can be upgraded
- Sell hardware that can be repairable
- Sell hardware that doesn't have batteries with expected lives of a year

By that logic - we still come to the same conclusion - that Apple will, if not migrate to AMD, use the threat of migration to AMD to negotiate a better deal with any Intel products.

We can assume Apple is doing the due diligence to kick the tires on a move to AMD. And I keep thinking - if their long term plan is to axe x86-64 and move to their own ARM chips anyway, why not enjoy that transition and temporarily move to AMD? As long as AMD was honest in their Ryzen 4000 keynote - I can't think of a reason NOT to. Especially if MS and Dell are already testing the waters.

Tim Cook’s job is to keep customers and shareholders happy. It’s not an easy task for any CEO who gives a damn. I don’t envy him that task. He has a fiduciary duty to maximize profit and sales. It’s a business, not a non-profit.

Apple laptops can be upgraded at purchase time, but customers haven’t been able to upgrade both DRAM and storage since 2012 when the Retina MacBook Pros were released. Why is this still being groused about?

Mac desktops are upgradeable, some just DRAM, some both.

All Macs are repairable, but not necessarily by the end user.

My 2012 15” MacBook Pro is nearly 8 years old, has 316 cycles and still hold 78% of its capacity. It last slightly more than a year, I think.

Apple might be doing it’s due diligence, but if they move to AMD, then there will be no move to ARM. Apple is not going to make a short transition to AMD AMD then go to ARM a few years later. It’s just not in their DNA.

 

[Unregistered 4U]

(b) someone at Intel who can't keep a secret saw an an internal memo saying

Having working in huge companies before, it’s quite easy for folks internally to know how large a given contract is as it’s not excessively privileged information. However, if you have the number, you can infer quite a bit. I think someone outside supplier management saw the value of the Apple contract, saw the number was FAR less than any of the prior few years, and inferred that Apple’s either getting an INCRIDBLE deal OR that Apple just wasn’t going to be buying many Intel chips. Which would mean that Apple intends to sell FAR fewer Macs OR they’re using someone else’s CPU.

For these folks, the modern smartphone will fit all of their needs.

And that’s the point. If one focuses on ARM can’t be as powerful as Intel’s highest end processors, you miss the point that the vast majority of folks don’t need Intel’s highest end processors. For those that prefer the desktop/laptop form factor, if the modern smartphone can fill all of their needs, a laptop/desktop with the SAME processor would do the same.

The top selling vehicle type is an SUV.

Is it SUV as a market or is the top selling mainly in the crossover area? Because a crossover is just a car that’s higher with more room for stuff which would seem to be what folks are looking for (especially as their knees are going bad).

it's because they forgot they had hungry competitors. Itanium failed because of Intel's arrogance in assuming the world would just fall in line because they declared IA-64 the future

This was my feeling, too. The initial Itanium was a bear, BUT if companies felt forced to buy and use it, Intel would likely have been able to iterate and improve. Having hungry competitors meant that there was an opening for a company to release something “as good” and undercut them.

Nobody has mentioned Apple's move to LLVM bitcode/IL which - if I'm not mistaken, would allow for most applications built in objective c or swift to be be compiled to bitcode - and from there to any one of several backed ISAs (PowerPC G5...)?

There was an article I read that explained this very well (can’t find it now) and that’s when I understood that they’d been prepping for the next transition, whatever it may be.

The question is, how much "legacy" software and abandonware is still hanging around that has x86-specific code

But the folks depending on legacy software is the smaller market. Apple sells almost half of their Macs every year to folks that have never owned a Mac before, so that concern is growing less and less important every year. It may never be of ZERO importance, but it’s probably already so low that Apple’s not even considering the pain it may cause.

They only care about not losing customers,

I’d change this to say that they don’t mind losing customers if it’s because those customers want what they aren’t going to make. The “extinction events” that someone mentioned earlier in this thread cost Apple LOTS of customers each time. And each time, they pivoted to a new group of willing customers.

AMD would invest it in years ago!

AMD’s focus is contesting Intel, so ARM wouldn’t benefit from doing anything with ARM... it would have pulled valuable resources from their x64 effort, which is now yielding dividends.

 

[theluggage]

Mac desktops are upgradeable, some just DRAM, some both.

Only the iMac 27" and the Mac Pro have officially user/third party-upgradeable RAM.

The Mini and iMac Pro RAM upgrades are - officially - Apple Authorized Service Provider only - and require completely dismantling the computer (including the tricky, special-tools-needed job of removing and replacing the glass on the iMac Pro) - if you break it, bye bye warranty and Applecare, if you don't break it Apple can claim you did the next time you try and use Applecare (maybe that's technically unenforceable in your jurisdiction, but good luck with that).

AFAIK none of the current SSD blades are available directly to consumers and, once the iMac has adopted the T2 chip (probably at the next revision) they won't be replaceable without some magic woo-woo from an Apple Service Provider anyway. Looks like the $6k Mac Pro gets by because you can add a PCIe SSD.

...and "upgradeable at Apple Service Provider" means paying Apple's usurious several-hundred-percent upgrade prices.

 

[dgdosen]

AFAIK none of the current SSD blades are available directly to consumers and, once the iMac has adopted the T2 chip (probably at the next revision) they won't be replaceable without some magic woo-woo from an Apple Service Provider anyway.

While not true today, that new Apple tax sounds disappointingly foreseeable.

 

[MikeZTM]

Only the iMac 27" and the Mac Pro have officially user/third party-upgradeable RAM.

The Mini and iMac Pro RAM upgrades are - officially - Apple Authorized Service Provider only - and require completely dismantling the computer (including the tricky, special-tools-needed job of removing and replacing the glass on the iMac Pro) - if you break it, bye bye warranty and Applecare, if you don't break it Apple can claim you did the next time you try and use Applecare (maybe that's technically unenforceable in your jurisdiction, but good luck with that).

AFAIK none of the current SSD blades are available directly to consumers and, once the iMac has adopted the T2 chip (probably at the next revision) they won't be replaceable without some magic woo-woo from an Apple Service Provider anyway. Looks like the $6k Mac Pro gets by because you can add a PCIe SSD.

...and "upgradeable at Apple Service Provider" means paying Apple's usurious several-hundred-percent upgrade prices.

T2 is awesome.
This is an answer to Intel's slow progression.

T2 support h265 encode and is good enough for compressing personal stuffs and runs super fast.
T2 shows a tiny bit of what Apple could do if they are not limited by Intel. It is basically a hardware patch for Intel CPUs to get some important feature from iOS.

More compact lead to less upgradability but if they can pack in more performance in lighter chassis then I'm all in.

Soldered LPDDR ram give them power efficiency upgrade. I bet in near future they will utilize new soldered storage chip to do the same.

It's insane to think about soldered CPU just 10 year ago and now you have to try really hard to find a laptop without soldered CPU on it.

 

[theluggage]

Apple sells almost half of their Macs every year to folks that have never owned a Mac before, so that concern is growing less and less important every year.

That may be true for the lower-end Macs - but it would still be "courageous" to launch a new Mac that couldn't run (say) MS Office or Photoshop on day one - and it may not be top on MS or Adobe's agenda to fix that. That's an improvement on 10 years ago when it would have been "suicidal". Presumably there would be some sort of x86 emulation (ISTR that there were some IP issues with the x86 instruction set, but MS seems to have got away with it) but x86 to ARM wouldn't carry the sort of raw power boost seen with previous processor transitions, so performance might be a bit rubbish.

I don't think its true of the higher-end Macs, though: I'm afraid I don't see many people buying those unless they're already committed to MacOS... and think of all the third party plugins that people rely on with things like Logic Pro.

Still - I've already said several times that switching to ARM is not as unthinkable as some people make out - but its still non-trivial!

 

[sublunar]

T2 is awesome.
This is an answer to Intel's slow progression.

T2 support h265 encode and is good enough for compressing personal stuffs and runs super fast.
T2 shows a tiny bit of what Apple could do if they are not limited by Intel. It is basically a hardware patch for Intel CPUs to get some important feature from iOS.

More compact lead to less upgradability but if they can pack in more performance in lighter chassis then I'm all in.

Soldered LPDDR ram give them power efficiency upgrade. I bet in near future they will utilize new soldered storage chip to do the same.

It's insane to think about soldered CPU just 10 year ago and now you have to try really hard to find a laptop without soldered CPU on it.

It certainly appears that the T2 is Apple's answer to several questions:

1. Quicksync - the encoding improvement means that Apple aren't so reliant on the (very good) Intel hardware decoding/encoding of simple jobs for features such as AirPlay2.

2. SSD Controller - as it runs Apple's Secure Boot and encrypted storage it's effectively Apple's DRM solution for killing of Hackintosh within a few generations.

3. System management controller

4. Image signal processing from the FaceTime HD camera (face detection, exposure control, tone mapping)



Apple really need too keep upgrading it to keep it up to date, given that they sometimes let things lie for years there is a possible danger of it becoming a bottleneck in the future.

 

[high heaven]

We are talking about different TDP from AMD APU that can be used.

Stop ignoring the base TDP. That's totally nonsense.