How do we know ARM won’t be like PowerPC?

[SegNerd]

PowerPC processors were a RISC-based processor that everyone claimed was an advanced design, superior to Intel, that could outperform Intel even at lower clock speeds. But as we all know, within a few years on the market, x86 chips kinda... kicked their ***.

Now Apple is switching once again to a RISC architecture that they claim is superior, at least on paper.

What went wrong with PowerPC? How do we know it won’t happen with ARM?

 

[ruslan120]

PowerPC’s architecture became bloated and power hungry. Intel was logical choice at the time because performance per watt was significantly better, especially on mobile devices.

What changed now?

A vast library of applications readily available to the consumer that will instantly work. Mac App Store and iOS App Store. For most people that is enough, for professionals there will be growing pains (but things like Final Cut Pro will work as well).

 

[casperes1996]

Nothing really went wrong with PPC. It’s doing pretty well actually. Power8 servers are actually releasing at the end of October. Pretty massive throughput. There’s also a reason that both the PlayStation 3 and Xbox 360 used PowerPC processors. They are and were fantastic. The issue was not so much PPC but the direction that PPC was going versus Intel - IBM didn’t really have the chip-focused resources to do everything and they wanted to focus PPC on mainframes and servers, so PPC chips were not suited for the low power markets Apple wanted like laptops and consumer devices. But put a Power chip in a mainframe and it has very high throughput.

As for Apple Silicon; Honestly it’s not that much about RISC vs. CISC. Or even the instruction set, really. I’m sure you also wouldn’t consider a Pentium equivalent to a Ryzen even though they are both x86. Apple Silicon is totally in the hands of Apple and so far they’ve shown to have fantastic chipmakers. That doesn’t mean they’ll be on top forever - Intel was king for a while and AMD took over and one day Intel will take back the crown - but Apple is competitive; Very competitive. And they can do things they couldn’t while relying on others, be that IBM, Motorola or Intel - like custom accelerators

 

[velocityg4]

While we don't know for certain how desktop processors will scale up. We do know and have known for quite a while. That they are much faster per watt than x86.

PowerPC failed early on in the consumer market place. Basically just Apple used them. There were others but their sales volume was way too small to matter. Which means it didn't have the R&D budget to compete.

It did fairly well with servers and still does. So, while Motorola gave up and sold of their PPC business. IBM focused on the server market where power usage didn't matter.

That means IBM was trying to scale down from a power inefficient CPU. They didn't put much effort into it either.

What it boils down to. There wasn't enough funding to compete.

ARM however has a massive market. Apple has a vested interest in tons of R&D. Given these devices are the core of their staggeringly profitable business. If they have problems. They can invest more than AMD, Intel and Qualcomm combined.

Even if the Macs aren't sold in anywhere near the volume of iPhone. It doesn't matter as much. Most of the legwork is already done for the iPhone. They just need to scale up for laptops and desktops.

Who knows how PowerPC would have done. If Windows for PowerPC did well and Dell and HP were on board from the 601 days.

 

[casperes1996]

While we don't know for certain how desktop processors will scale up. We do know and have known for quite a while. That they are much faster per watt than x86.

PowerPC failed early on in the consumer market place. Basically just Apple used them. There were others but their sales volume was way too small to matter. Which means it didn't have the R&D budget to compete.

It did fairly well with servers and still does. So, while Motorola gave up and sold of their PPC business. IBM focused on the server market where power usage didn't matter.

That means IBM was trying to scale down from a power inefficient CPU. They didn't put much effort into it either.

What it boils down to. There wasn't enough funding to compete.

ARM however has a massive market. Apple has a vested interest in tons of R&D. Given these devices are the core of their staggeringly profitable business. If they have problems. They can invest more than AMD, Intel and Qualcomm combined.

Even if the Macs aren't sold in anywhere near the volume of iPhone. It doesn't matter as much. Most of the legwork is already done for the iPhone. They just need to scale up for laptops and desktops.

Who knows how PowerPC would have done. If Windows for PowerPC did well and Dell and HP were on board from the 601 days.

Yes, exactly. Though while the ISA matters, it really isn’t the end all-be all. I mean of course it has some importance; While you can make a Turing Complete system with just a single instruction, it’d basically be impossible to make that fast. Likewise having too many instructions and too complicated instructions might encourage a penalty on the front end instruction decoding.

But even if we strip out all the instructions that’ve been added to x86 over the years, if we compare an old Intel 80386 to a modern i7 10700, running the 10700 in 32-bit mode and clock normalizing everything to even it out, the 10700 is still a bajillion times faster, even though they are both i386 ISAs (aka what we normally refer to as x86) - Of course the 10700 has many more instructions as well that’ve come over the years, especially for floating point and SIMD which could accelerate it even further ahead as well.

Likewise an Apple Silicon chip is not just ARM. And ARM isn’t even a single instruction set really, but that’s a whole other story - The binaries are mostly compatible so an ARMv7 binary will run on Aarch64. Anyways, ranting now. Point was that an A14 is far from the old Acorn Risk Machine chips that started out what we now know as ARM and it’s not really doing justice to the chip design to reduce it all to just the ISA

 

[cmaier]

PowerPC processors were a RISC-based processor that everyone claimed was an advanced design, superior to Intel, that could outperform Intel even at lower clock speeds. But as we all know, within a few years on the market, x86 chips kinda... kicked their ***.

Now Apple is switching once again to a RISC architecture that they claim is superior, at least on paper.

What went wrong with PowerPC? How do we know it won’t happen with ARM?

IBM had no interest in targeting processors to Macs, Motorola didn’t know how to design processors.

Apple knows how, and they are highly motivated.

 

[cmaier]

PowerPC’s architecture became bloated and power hungry.

No, the architecture was not bloated and power hungry.

The designs were not bloated, but were power hungry. Because they were targeted at IBM’s own machines, and not at Macs. IBM knew how to do designs but didn’t care about apple. Motorola had no idea what they were doing. And Steve Jobs put Exponential out of business.

 

[Spindel]

PowerPC processors were a RISC-based processor that everyone claimed was an advanced design, superior to Intel, that could outperform Intel even at lower clock speeds. But as we all know, within a few years on the market, x86 chips kinda... kicked their ***.

Now Apple is switching once again to a RISC architecture that they claim is superior, at least on paper.

What went wrong with PowerPC? How do we know it won’t happen with ARM?

Well at least for a fair number of years the PowerPC CPU in mac actually outperformed x86 pretty well, but as time wen on x86 cought up and surpassed PowerPC. PowerPC was around in Macs for about 12 years.

Now after about 14 years everything points at Apples Apple Silicon implementation has cought up and surpassed x86 in performance.

It's almost as if the development is moving forward and the world isn't a static place.

 

[Project Alice]

PowerPC processors were a RISC-based processor that everyone claimed was an advanced design, superior to Intel, that could outperform Intel even at lower clock speeds. But as we all know, within a few years on the market, x86 chips kinda... kicked their ***.

Now Apple is switching once again to a RISC architecture that they claim is superior, at least on paper.

What went wrong with PowerPC? How do we know it won’t happen with ARM?

PowerPC was way better than x86 up until around 2004. Intel had nothing on PPC til that point. It was mainly the fault of IBM and Motorola, not the architecture itself.

 

[Project Alice]

No, the architecture was not bloated and power hungry.

The designs were not bloated, but were power hungry. Because they were targeted at IBM’s own machines, and not at Macs. IBM knew how to do designs but didn’t care about apple. Motorola had no idea what they were doing. And Steve Jobs put Exponential out of business.

G5s were power hungry only be because it was a server grade chip.

Pentium 4s of the time were equally if not more power hungry.

 

[mr_roboto]

And Steve Jobs put Exponential out of business.

Can you elaborate on that one? I remember Exponential as looking a little dubious from a technical standpoint. It was BiCMOS, to push clocks high, but BiCMOS also meant power was going to be an issue. And the caches seemed small given how fast they wanted to clock the core.

But those are all vague impressions from very afar, I don't know how good they actually were in practice.

 

[quarkysg]

According to https://en.wikipedia.org/wiki/Exponential_Technology, Exponential, or it's successor, was bought by Apple. So it looks like their work lives on in Apple Silicon.

 

[Mikael H]

How do we know it won’t happen with ARM?

Pretty much any technology you can point at today will be obsolete in a number of years. The fact that a technology company recognizes this and moves with the times without losing a significant following during these upheavals is pretty impressive. I'm pretty sure in another 15-20 years people at Apple will be seriously considering a migration to the platform that shows most promise to them at that time.

(At that point I'm sure at least some corporations will have managed to move at least a few of their old Win32 apps to a more modern framework...)

 

[iKrivetko]

We don't until we try, what kind of a question is that?

 

[thenewperson]

Well, we don't. I can't imagine Apple isn't using this opportunity to also make any future transition they make easier for themselves + their developers.

 

[casperes1996]

Pretty much any technology you can point at today will be obsolete in a number of years

*Points towards light bulb*

 

[thenewperson]

*Points towards light bulb*

Well, the specific tech used for those does tend to go obsolete.

 

[casperes1996]

Well, the specific tech used for those does tend to go obsolete.

*Scratches head, thinking.*

*Points to wall power socket* (I'm joking btw. not disagreeing with you)

 

[calstanford]

It could, but then it's Apple's own fault.

PowerPC was a dependency on two companies who weren't major chip players anymore at it's day (Moto + IBM)

 

[cmaier]

Can you elaborate on that one? I remember Exponential as looking a little dubious from a technical standpoint. It was BiCMOS, to push clocks high, but BiCMOS also meant power was going to be an issue. And the caches seemed small given how fast they wanted to clock the core.

But those are all vague impressions from very afar, I don't know how good they actually were in practice.

Our chips ran fine. First tape out came out at around 420MHz with similar power budget to IBM. The speed issue was a cross coupling bug caused by automated power reduction software that was used on the netlist, and we had already taped out a fix that would get us over 500MHz.

we had orders from the Mac clone makers, but to use our processor the bios had to be modified slightly. Jobs came back to Apple and wanted the clone makers gone, and wouldn’t allow the bios mods.

we were not traditional bicmos. All of the logic circuits were pure bipolar, mostly ECL but some CML. It was bicmos only in the sense that all memory circuits were CMOS.

As far as power, we used no more dynamic power than anyone else. It’s just that our static power was equal to our dynamic power, so power didn’t fluctuate. Not suitable for mobile, but fine for anything plugged in to a wall. Plus it’s then immune from all sorts of side-channel attacks - you can’t guess passwords by measuring power fluctuations
[automerge]1602682728[/automerge]

It could, but then it's Apple's own fault.

PowerPC was a dependency on two companies who weren't major chip players anymore at it's day (Moto + IBM)

IBM was certainly a major chip player at that point. In fact, so was Motorola.

 

[cmaier]

According to https://en.wikipedia.org/wiki/Exponential_Technology, Exponential, or it's successor, was bought by Apple. So it looks like their work lives on in Apple Silicon.

sort Of.

our Austin office was working on another project. When we went out of business, most of those folks stuck around and became EVSX, which was mostly a consulting thing to keep themselves going. Then they became intrinsity and then they got bought. The guy in the cubicle next to me, who was a key architect of our work, moved to Austin and became a critical part of all that.

 

[D.T.]

*Points towards light bulb*

Light bulb you say? Hmm, researching this as our candle and lantern oil costs have really escalated this year ...

 

[MevetS]

*Points towards light bulb*

Some do last a while ...

Centennial Light - Wikipedia

en.wikipedia.org

 

[casperes1996]

Some do last a while ...

Centennial Light - Wikipedia

en.wikipedia.org

Better reference:

 

[ArPe]

PowerPC processors were a RISC-based processor that everyone claimed was an advanced design, superior to Intel, that could outperform Intel even at lower clock speeds. But as we all know, within a few years on the market, x86 chips kinda... kicked their ***.

Now Apple is switching once again to a RISC architecture that they claim is superior, at least on paper.

What went wrong with PowerPC? How do we know it won’t happen with ARM?

Aside from what others said, these two aren’t comparable. ARM has been around for decades and making an in-house CPU+GPU means you can optimize the software to it.

Look how fast and smooth iPadOS is. This is what macOS should feel like. Long term Mac users have always been annoyed the macOS support for third party graphics chips hasn’t been as good as Windows/Linux. With Apple using their own GPU and decoder/encoded chips they no longer have to lean on third party documentation and or have high power consumption causing throttling.