Live data from Hacker News

Apple’s new M1 Pro and M1 Max processors

apple.com

461–470 of 1001 posts

Re: Apple’s new M1 Pro and M1 Max processors

#461
post #407

Earlier quoted context omitted.

If Apple actually cared about sustainability, they would make their devices repairable.

They are repairable, but not by consumers in most cases.

You literally cannot replace the battery in that Surface Laptop without destroying the whole thing.

It's made to be thrown away, instead of repaired.

Re: Apple’s new M1 Pro and M1 Max processors

#462

I always thought it was strange that "integrated graphics" was, for years, was synonymous with "cheap, underperforming" compared to the power of a discrete GPU. I never could see any fundamental reason why "integrated" should mean "underpowered." Apple is turning things around, and is touting the benefits of high-performance integrated graphics.

> I never could see any fundamental reason why "integrated" should mean "underpowered." There was always one reason: limited memory bandwidth. You simply couldn't cram enough pins and traces for all the processor io plus a memory bus wide enough to feed a powerful GPU. (at least not in a reasonable price)

We solved that almost a decade ago now with HBM. Sure, the latencies aren't amazing, but the power consumption numbers are and large caches can hide the higher access latencies pretty well in almost all cases.

Re: Apple’s new M1 Pro and M1 Max processors

#463

Based on the numbers it looks like the M1 Max is in the RTX 3070-3080 performance territory. Sounds like mobile AAA gaming has potential to reach new heights :D

Elephant in the room is that an A15/M1 beefed up GPU is exactly the right chip for a Nintendo Switch/Steam Deck form factor device.

Or an AV/VR headset.

Re: Apple’s new M1 Pro and M1 Max processors

#464

Earlier quoted context omitted.

Don't know the clock speed but 8 cores at 3Ghz working on 128bit SIMD is 8 3 16 = 384GB/s so we are in the right ball park. Not that I personally have a use for that =) Oh, wait, bloody Java GC might be a use for that. (LOL, FML or both).

But the classic SIMD problem is matrix-multiplication, which doesn't need full memory bandwidth (because a lot of the calculations are happening inside of cache). The question is: what kind of problems are people needing that want 400GB/s bandwidth on a CPU? Well, probably none frankly. The bandwidth is for the iGPU really. The CPU just "might as well" have it, since its a system-on-a-chip. CPUs usually don't care to…

Doesn't prefetching data into the cache more quickly assist in execution speed here?

Re: Apple’s new M1 Pro and M1 Max processors

#465

Earlier quoted context omitted.

But the classic SIMD problem is matrix-multiplication, which doesn't need full memory bandwidth (because a lot of the calculations are happening inside of cache). The question is: what kind of problems are people needing that want 400GB/s bandwidth on a CPU? Well, probably none frankly. The bandwidth is for the iGPU really. The CPU just "might as well" have it, since its a system-on-a-chip. CPUs usually don't care to…

Doesn't prefetching data into the cache more quickly assist in execution speed here?

How do you prefetch "node->next" where "node" is in a linked list?

Answer: you literally can't. And that's why this kind of coding style will forever be latency bound.

EDIT: Prefetching works when the address can be predicted ahead of time. For example, when your CPU-core is reading "array", then "array+8", then "array+16", you can be pretty damn sure the next thing it wants to read is "array+24", so you prefetch that. There's no need to wait for the CPU to actually issue the command for "array+24", you fetch it even before the code executes.

Now if you have "0x8009230", which points to "0x81105534", which points to "0x92FB220", good luck prefetching that sequence.

--------

Which is why servers use SMT / hyperthreading, so that the core can "switch" to another thread while waiting those 50-nanoseconds / 200-cycles or so.

Re: Apple’s new M1 Pro and M1 Max processors

#466
post #221

Server Chips: if you removed the GPU, and added ECC - these would be dang nice server chips.

There is technically no reason Apple could not introduce a cloud computing service based on their silicon at some point. But would it generate the kind of profit margins they need? An interesting space to watch.

2 years when their contracts expire

Re: Apple’s new M1 Pro and M1 Max processors

#467
post #247

Earlier quoted context omitted.

It's a win for a worker who's compensated based on their work output, which is pretty much the opposite of what a salaried worker is.

Productivity is productivity, doesn't matter how one is paid.

…then why mention a salary at all?

Re: Apple’s new M1 Pro and M1 Max processors

#468
post #241

Earlier quoted context omitted.

Just today I got one 6s and one iPhone 7 screen repaired(6s got the glass replaced, the 7 got full assembly replaced) and battery of the 6s replaced at a shop that is not authorized by Apple. It cost me 110$ in total. Previously I got 2017 Macbook Air SSD upgraded using an SSD and an adapter that I ordered from Amazon. What’s that narrative that Apple devices are not upgradable or repairable? It simply not true. If a…

>Just today I got one 6s and one iPhone 7 screen repaired Nice, except doing a screen replacement on a modern iPhone like the 13 series will disable your FaceID making your iPhone pretty much worthless. >Previously I got 2017 Macbook Air SSD upgraded using an SSD and an adapter that I ordered from Amazon Nice, but on the modern Macbooks, the SSD is soldered and not replaceable. There is no way to upgrade them or repl…

Nice, except doing a screen replacement on a modern iPhone like the 13 series will disable your FaceID making your iPhone pretty much worthless.

Only if you go to someone who isn't an authorised Apple repairer.

Re: Apple’s new M1 Pro and M1 Max processors

#470

Based on the numbers it looks like the M1 Max is in the RTX 3070-3080 performance territory. Sounds like mobile AAA gaming has potential to reach new heights :D

The 3080M has something around 20TFLOPS of theoretical f32 performance. That's double that of Apple's biggest offerings.

In theoretical compute, it's closer to the 3060M which is nothing to sneeze at.

Post reply on HN