Is Silicon M1 a Game Change?

[xraydoc]

I don’t think so. There is no reason to disallow RAM upgrades. The current Mac Pro has tremendous memory bandwidth. 140 GB/s according to Apple. That is more than twice as fast as the M1.

The SSDs on the current Mac Pro are faster than the SSDs on the M1 Macs. So again there is no reason to change that design.

As for slots, Apple is rumored to be creating their own GPU. There is no reason why they couldn’t put that on an MPX card and allow users to upgrade to newer versions over time. I also am not sure that Apple is going to block AMD graphics cards on the Mac Pro. Having access to very high end GPUs is something that Apple might not be able to do in house.

Those are Intel designs, however.

Hey, I hope I’m wrong. I’m afraid I’m not, unfortunately.

 

[jdb8167]

Those are Intel designs, however.

Not sure why you think that matters. Apple can design any memory controller that they need. It doesn’t have to be like the M1. The SSD controller on the Mac Pro isn’t Intel. It is a T2 designed by Apple. PCIe is standard. Again not Intel.

Apple can design a chip for the Mac Pro that matches or exceeds the current Xeon design in the current Mac Pro.

 

[alien3dx]

i still writing in old intel compare m1 . m1 is good but apple platform preety lack of software /games

 

[xraydoc]

Not sure why you think that matters. Apple can design any memory controller that they need. It doesn’t have to be like the M1. The SSD controller on the Mac Pro isn’t Intel. It is a T2 designed by Apple. PCIe is standard. Again not Intel.

Apple can design a chip for the Mac Pro that matches or exceeds the current Xeon design in the current Mac Pro.

I'm totally with you. Apple has the ability to design some pretty impressive stuff.

But my take on the situation is that Apple will continue it's iPadification of the Mac, making things more and more closed. They don't have to do it -- I believe they want to do it.

Like I said, I hope I'm wrong. Back in the day, I bought a first generation Mac Pro (MP1,1) to replace my G5 tower because of it's expandability. Upgraded it several times. Drove it into the ground. Definitely got my money's worth. I'd love an expandable Mac, though the current generation of Mac Pro is too expensive at it's base config and doesn't offer enough extra performance at it's base config compared with other machines for me personally to justify the price. Still pining for the legendary xMac. Sorry, I digress....

 

[nquinn]

It's a leap forward, but let's not pretend it is world changing. Intel/AMD/others would eventually hit 5nm and beyond for cool/quiet chips. The new Tiger Lake and AMD 6000 series stuff is already making a lot of progress.

ARM will continue to have less overhead than x86 and be more integrated so it is a little more advanced, but it's not like Apple/TSMC here created alien like tech. Apple is just something like 2-3 years ahead.

 

[v0n]

Apple can design a chip for the Mac Pro that matches or exceeds the current Xeon design in the current Mac Pro.

I don't think they can. And I think current lineup is a good proof of that. Not even in PowerPC or Motorola times did Apple throw so few eggs into all the baskets and have so many different products use one and the same chip.
We've all gone overboard expecting them to have M1X or M2 update 9 months from M1. It's not coming this year. This (taps with finger on his M1 mac mini the desk) is all they have. This is the "wow". And as many of us know already - the "wow" is nice but it has inherent vices and issues. Which I don't think they expected to crop up and it's not something they can fix any time soon.
So I recon it will be a long while before they have anything else. I know we all expected Apple to actually have "IT", the tech, and go all the way to Mac Pro with "IT" in a snap of their fingers. But they don't and they won't. It's going to be a long, painful, three, maybe four year long process of changing device after device to just a little better arm, just a tad more memory, with just a single port more, until the entire lineup but Mac Pro run but four or five different versions of the same processor, same gen, two of which we know already, and the only variation will be between basic version and severely cut down version. That's my optimistic prediction.

 

[neinjohn]

Apple basically today launched a very complete Zoom competitor on three very functionally different devices with functions that need deep system integrations and some peculiar hardware support (the AI functions) and it probably took less work than developing Microsoft Teams to Windows and Android.

Google and Microsoft don't even have three mature hardware platform to do that at the same time nor, at the moment, tailored hardware for their software goals.

I would say that today was another nail on game changer.

 

[cocoua]

My concern is about what will happen once ARM reach its smallest manufacturer size, as it seems until today has Been one of the biggest reason of its rocket career

 

[gustavopi]

Yeah, I know about RISC and CISC. I frequently see explanations like the one you give here but in the end, if you look at all this in more detail, I don’t think there is much left of this argument. The main difference between Aarch64 and x86 is that the first is a load-store architecture with fixed instruction length and the second is a register memory architecture with variable instruction length. This means that you can occasionally encode a sequence of operations in a slightly more compact way in x86, and that’s about it… and even then it mostly applies to toy examples. When you look at real world code, especially optimized one, there is no significant difference between x86 and Aarch64 binary sizes.

As to complex vs. simple instructions… frankly, I don’t even know where do these misconceptions came from. ARM has instructions that store/load multiple registers at once, it has matrix multiplication, auto-increment addressing modes, pointer encryption, half-float operations, interleaved simd loads and stores, fully featured horizontal vector operations… in fact, I can’t think of any instruction that x86 has that ARM lacks (except maybe the weird old-school CISC instructions which are dead slow and never used in modern code).

In the end, Aarch64 is a modern instruction set that has been meticulously designed to simplify the lives of both the CPU designer and the compiler writer as well as enable high-performance out of order execution. And x86 is a very old design that has been continuously adding more modern features on top of its legacy core.

I wrote that answer and did look a little further... the differences CISC - RISC are polemic when it comes to these points. I can't have my own opinion because I only worked with Z-80 and 8086 in this matter. They were very different to me but this was a long time ago. Thought this discussion is for some academic thesis.

I agree. However, I'm not sure if Apple sees it that way.
In the current Mac Pro, what's the expansion for? GPU cards, drives and RAM.
I'm expecting an absence of GPUs, since it'll be integrated into the CPU as it is on the M1. Same with RAM. So we're left with maybe a couple SSD slots for additional storage (main SSD will be integrated like on the M1 SOC).

I like my pro 2012 because it is powerful and very customizable (therefore I am now using Big Sur). What a professional user expect, usually, is an "open" tool. I have looked for a newer used Macbook Pro but it seems they are not like this anymore, with memory and/or ssd soldered. If Apple doesn't change this policy, I will go to a hackintosh. A professional must have a suitable tool, not just a good brand.

If I remember it correctly, those rectangular Intel CPU modules were actually an attempt to have a proprietary slot standard to prevent the use of AMD Athlon CPUs - which were compatible with the "square" sockets Intel had been using (this was around the first time AMD had produced CPUs that could compete with what Intel was cranking out.) Fortunately, what Intel try to do failed, and they had to go back to the more conventional sockets.

As for the PPC CPUs, I believe one of the main reasons Apple stopped using them was the same reasons they decided to replace Intel CPUs: as the CPUs got more powerful, the thermals became more of a problem (for instance, the fasted G5 PowerPC ran so hot that the CPU had to be liquid-cooled.)

Ironically, at the same time Apple is abandoning Intel CPUs, AMD has mostly surpassed them in X86 performance again with their Ryzen CPUs …?

Is not the first time AMD overcome Intel, but now it is more dangerous competition, AMD is now bigger enough to defy Intel's leadership.
I recall Steve Jobs justifying this change to fans by the lack of interest of IBM in keep processors in the top, and also provide the needed amount of parts so Apple could expand the market. I think powermac G5 was an awesome machine, but the bigger differences was due to the board project (as the serial bus). That was difficult times, Intel only could supply Job's wishes with Core2 arch. I remember Dual Core being slower than G5 and things like this (If somebody got better memory please share). But IBM gave up of the competition, leaving to AMD and Intel. Both are good brands to me, but I miss the wish of to change the game that Apple still have. So Apple was right at that time either.