Apple M1 CPU

[Armageus]

I'm looking at this and I can't help but look for the flaw. There's no such thing as a free lunch, and I'm looking for the trade off. Anyone got any ideas where it might be?

I'm struggling to find one, but there must be one, otherwise Intel and AMD would have done it already.

The only thing I can think of is where do they go from here? Can they scale up to 16, 32 or more cores or is it limited by architecture.

How much of the performance is due to the on package memory? That clearly isn't cost effective to scale up, hence only 8gb and 16gb parts being released so far.

 

[q974739]

Free lunch? Since why?
I don't know why people love to troll so much.

Do you understand the reference?

 

[4K8KW10]

Intel and AMD have no technological competence to build ARM RISC architectures. If it was so easy, there would be dozens of companies doing it already.

 

[q974739]

I'm struggling to find one, but there must be one, otherwise Intel and AMD would have done it already.

The only thing I can think of is where do they go from here? Can they scale up to 16, 32 or more cores or is it limited by architecture.

How much of the performance is due to the on package memory? That clearly isn't cost effective to scale up, hence only 8gb and 16gb parts being released so far.

Inability to scale would be an interesting one. Memory would also be interesting... both might well be good calls.

 

[ChrisD.]

How much of the performance is due to the on package memory? That clearly isn't cost effective to scale up, hence only 8gb and 16gb parts being released so far.

I wonder if it's the size of the SoC and the cooling capacity? Then again the Mini has plenty of room inside it.

 

[Armageus]

Intel and AMD have no technological competence to build ARM RISC architectures. If it was so easy, there would be dozens of companies doing it already.

I'm sure Intel and AMD could apply the technology competence they have to any CPU architecture

 

[Vince]

Intel and AMD have no technological competence to build ARM RISC architectures. If it was so easy, there would be dozens of companies doing it already.

It's an open source supported platform, if you believe that any other chip provider be that intel, amd or anybody else do not have experience and capability in fpga designs then you are mad. Didn't amd just aquire xilinx a market leading fpga maker?

 

[4K8KW10]

Do you understand the reference?

Do you understand that RISC architectures are much more power efficient and it's in their very basics?

Where did this bizarre "idea" of free lunch ever come to you?
Free lunch means something for free. Where do you see anything for free?

 

[Jimmy Weirdarms]

I'm struggling to find one, but there must be one, otherwise Intel and AMD would have done it already.

Intel has sat on its laurels for years, and failed to get its smaller processes together. Everyone else is making great strides.

There's a good analysis here - https://jamesallworth.medium.com/intels-disruption-is-now-complete-d4fa771f0f2c

Going into the history of how intel disrupted markets and was a dynamic mover, but has recently become the slow behemoth.

The lunch has not been free - Apple is the world's most valuable company, they have ploughed billions into chip design, and as arm licensees they've used a great starting point to build something amazing.

Intel and AMD are not the only show in town, IBM still do good things with POWER, it's just that they cost ludicrous money.

 

[4K8KW10]

It's an open source supported platform, if you believe that any other chip provider be that intel, amd do not have experience and capability in fpga designs then you are mad. Didn't amd just aquire xilinx a market leading fpga maker?

AMD is just a copy-cat of Intel. Or at least has been for the vast majority of the 50-year history.
Intel is still struggling with 14nm and can't move its CPUs past it.

ARM is already doing 5nm and looking forward to 3nm. Yes, very incompetent Intel.

 

[Cenuij]

Intel and AMD have no technological competence to build ARM RISC architectures. If it was so easy, there would be dozens of companies doing it already.

Really man this is just embarrassing for you, every AMD CPU for about the last 5 or more years has an ARM core inside it.

 

[4K8KW10]

Really man this is just embarrassing for you, every AMD CPU for about the last 5 or more years has an ARM core inside it.

And why are these AMD CPUs 105 watt parts then? 105-watt is crazy high TDP.

 

[Cenuij]

And why are these AMD CPUs 105 watt parts then? 105-watt is crazy high TDP.

How is TDP in any way related to the fact you made a 100% incorrect statement?

 

[Vince]

AMD is just a copy-cat of Intel. Or at least has been for the vast majority of the 50-year history.
Intel is still struggling with 14nm and can't move its CPUs past it.

ARM is already doing 5nm and looking forward to 3nm. Yes, very incompetent Intel.

I think you need to go back over that history, the number of 1st in terms of CPU design Intel vs AMD is significantly in AMD's favour. Even if AMD was born out of a requirement from supply chains to be an alternative supplier of intel designs. The fact that ARM is on 5nm means nothing, not really. AMR can be on anything you want, it's apple that put arm on 5nm not arm that put arm on 5nm. As above it escaped me but the security processor in 1st gen Ryzen was an ARM Cortex A-5.

 

[HACO]

The m1 looks promising. Yeah I'd love to know how much its using when it's runn rendering. Is it reported as 31W?

Interesting to see where Apple take this, I have no idea how they will scale this into workstation machines, but i'll be very intrigued.

Anandtech estimated it to be at 20-24w while at full load.

I'm looking at this and I can't help but look for the flaw. There's no such thing as a free lunch, and I'm looking for the trade off. Anyone got any ideas where it might be?

I'm struggling to find one, but there must be one, otherwise Intel and AMD would have done it already.

The only thing I can think of is where do they go from here? Can they scale up to 16, 32 or more cores or is it limited by architecture.

How much of the performance is due to the on package memory? That clearly isn't cost effective to scale up, hence only 8gb and 16gb parts being released so far.

Ram limitation is likely due to LPDDR4X availability of 16GB packages (almost non-existent) than anything in the architecture. So they had to stick with 8GB max. Scaling memory is very easy compared to everything else that they've achieved.

The real trade-off however, is die space. These chips are wide and generally large. Apple doesn't need to turn a profit on the chips themselves, so they can go as large as they want. Intel and AMD need to maintain profit margins, so their hands are tied in a lot more ways than just what the technology allows.

Intel and AMD have no technological competence to build ARM RISC architectures. If it was so easy, there would be dozens of companies doing it already.

Nonsense. You keep repeating this but AMD is literally building ARM CPUs right now. Inside every Ryzen CPU, there's an ARM core. Intel had an ARM ISA license and used to build products but gave up because it wasn't profitable. Intel and AMD don't need to build ARM chips, that's why they don't.

I wonder if it's the size of the SoC and the cooling capacity? Then again the Mini has plenty of room inside it.

Bravo. That's it. Size & complexity of the chip is Apple's advantage, for Intel and AMD that affects margins, Apple doesn't care about margin on chips. They can absorb that with the savings they make from not needing third-party CPUs.

It's an open source supported platform, if you believe that any other chip provider be that intel, amd do not have experience and capability in fpga designs then you are mad.

It's an open architecture, but not open source.

 

[Cenuij]

Really man this is just embarrassing for you, every AMD CPU for about the last 5 or more years has an ARM core inside it.

And lets not forget there was a whole series of ARM Opteron server parts

 

[Vince]

It's an open architecture, but not open source.

Nice correction and that is true. There are different licenses available as far as I remember depending on what you want to make?

 

[HACO]

Nice correction and that is true. There are different licenses available as far as I remember depending on what you want to make?

Yeah, there's an ISA license, and then there are core design licenses (for every product), and there's also just a huge license that has everything. Each license has subcategories (some allow modification, some don't), and then they can be perpetual or not.

RISC-V ISA is fully open source and doesn't require any license.

 

[Cenuij]

I think you need to go back over that history, the number of 1st in terms of CPU design Intel vs AMD is significantly in AMD's favour. Even if AMD was born out of a requirement from supply chains to be an alternative supplier of intel designs. The fact that ARM is on 5nm means nothing, not really. AMR can be on anything you want, it's apple that put arm on 5nm not arm that put arm on 5nm. As above it escaped me but the security processor in 1st gen Ryzen was an ARM Cortex A-5.

He couldn't be more wrong if he tried, 64bit x86 computing is designated `amd64` in build architectures because AMD released the first 64bit x86 parts and Intel had to copy those instructions

 

[4K8KW10]

Intel and AMD don't need to build ARM chips, that's why they don't.

They will start if they want to survive.

Also, building competitive ARM architectures is different to what AMD does right now.

 

[HACO]

They will start if they want to survive.

Also, building competitive ARM architectures is different to what AMD does right now.

You moved the goalpost now, and you're absolutely wrong about everything you said in these last few pages. How many times do you need people to correct you before you stop and think that maybe you have no idea what you're talking about?