Apple’s new M1 Pro and M1 Max processors • Hacker News Board

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

#731

Post by astrange » Tue, Oct 19, 2021, 1:51 AM UTC

sulam wrote: post #626

bla3 wrote: post #118

Earlier quoted context omitted.

I thought the memory was one of the more interesting bits here. My 2-year-old Intel MBP has 64 GB, and 8 GB of additional memory on the GPU. True, on the M1 Max you don't have to copy back and forth between CPU and GPU thanks to integrated memory, but the new MBP still has less total memory than my 2-year-old Intel MBP. And it seems they just barely managed to get to 64 GiB. The whole processor chip is surrounded by…

So, M1 has been out for a while now, with HN doom and gloom about not being able to put enough memory into them. Real world usage has demonstrated far less memory usage than people expected (I don't know why, maybe someone paid attention and can say). The result is that 32G is a LOT of memory for an M1-based laptop, and 64G is only needed for very specific workloads I would expect.

Measuring memory usage is a complicated topic and just adding numbers up overestimates it pretty badly. The different priorities of memory are something like 1. wired (must be in RAM) 2. dirty (can be swapped) 3. purgeable (can be deleted and recomputed) 4. file backed dirty (can be written to disk) 4. file backed clean (can be read back in).

Also note M1's unified memory model is actually worse for memory use not better. Details left as an exercise for the reader.

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

#732

Post by techdragon » Tue, Oct 19, 2021, 1:52 AM UTC

modulusshift wrote: post #434

zelon88 wrote: post #356

Earlier quoted context omitted.

Apple does nothing to improve front line technician procedures. They aren't even an engineering factor. If you happen to be able to replace something on an Apple product, it's only because the cost-benefit ratio wasn't in favor of making that part hostile to work with. Apple puts 56 screws in the Unibody MBP keyboards. They were practically the pioneer of gluing components in permanently. They don't care about techni…

Oh hey they went back to screws on the keyboards? that's nice, they used to be single use plastic rivets, so at least you can redo that. Also Apple's glue isn't usually that bad to work with. Doesn't leave much residue, so as long as you know where to apply the heat you can do a clean repair and glue the new component back in.

I think apple might have learned a painful lesson about keyboard repairability… with all those warranty repairs they are *still* stuck with.

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

#733

Post by znwu » Tue, Oct 19, 2021, 1:55 AM UTC

zX41ZdbW wrote: post #672

dragontamer wrote: post #427

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…

> The question is: what kind of problems are people needing that want 400GB/s bandwidth on a CPU? Well, probably none frankly. It is needed for analytic databases, e.g. ClickHouse: https://presentations.clickhouse.com/meetup53/optimizations/

This is some seriously interesting stuff.

But they are demonstrating with 16 cores + 30 GB/s & 128 cores + 190 GB/s. And to my understanding they did not really mention what type of computational load did they perform. So this does not sound too ridiculous. M1 max is pairing 8 cores + 400GB/s.

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

#734

Post by Jimpulse » Tue, Oct 19, 2021, 1:55 AM UTC

gjsman-1000 wrote: post #61

For me, think about that memory bandwidth. No other CPU comes even close. A Ryzen 5950X can only transfer about 43GB/s. This thing promises 400GB/s on the highest-end model.

The AMD chips in the PS5 and new XBox reach 448GB/s and 326GB/s bandwith respectively with their unified memory.

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

#735

Post by eointierney » Tue, Oct 19, 2021, 1:59 AM UTC

Great design, abolute nope, human rights are more important

Don't feed the monsters

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

#736

Post by hbn » Tue, Oct 19, 2021, 2:01 AM UTC

breiti wrote: post #610

hbn wrote: post #377

These things came at a great time for me. My late-2014 MBP just received its last major OS upgrade (Big Sur), so I'm officially in unsupported waters now. I was getting concerned in that era from 2015-2019 with all the bad decisions (butterfly keyboard, no I/O, touchbar, graphics issues, etc.) but this new generation of MacBooks seems to have resolved all my points of concern. On the other hand, my late-2014 model is…

You could easily upgrade macOS using https://dortania.github.io/OpenCore-Legacy-Patcher/MODELS.ht... . On the hardware side, you could open it up, clean the fan, re-apply thermal paste (after so many years, this will make a big difference) and maybe even upgrade the SSD if you feel like it. That way, this laptop can easily survive another 1-3 years depending on your use-cases.

Interesting, thanks for the link

I actually ended up pulling the trigger on the base model 14 inch that was announced today, but I'll probably still keep this laptop around as a tinkering device in the future. If not only because it's still fairly capable, but I've got some good nostalgia for it!

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

#737

Post by fomine3 » Tue, Oct 19, 2021, 2:06 AM UTC

GeekyBear wrote: post #326

Unklejoe wrote: post #211

Earlier quoted context omitted.

The overall bandwidth isn't affected much by the distance alone. Latency, yes, in the sense that the signal literally has to travel further, but that difference is miniscule (like 1/10th of a nanosecond) compared to overall DDR access latencies. Better signal integrity could allow for larger busses, but I don't think this is actually a single 512 bit bus. I think it's multiple channels of smaller busses (32 or 64 bit…

> The overall bandwidth isn't affected much by the distance alone. Testing showed that the M1's performance cores had a surprising amount of memory bandwidth. >One aspect we’ve never really had the opportunity to test is exactly how good Apple’s cores are in terms of memory bandwidth. Inside of the M1, the results are ground-breaking: A single Firestorm achieves memory reads up to around 58GB/s, with memory writes co…

It just said that bandwidth between a performance core and memory controller is great. It's not related to distance between memory controller and DRAM.

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

#738

Post by daxfohl » Tue, Oct 19, 2021, 2:06 AM UTC

haberman wrote: post #415

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.

Still confused how 32 core M1 Max competes with Nvidia's thousands-of-cores GPUs. Certainly there are some things that are nearly linear with core count, or otherwise they wouldn't keep adding cores, right?

Edit: Found answer here. GPU core is not the same thing as a CUDA core. https://www.reddit.com/r/hardware/comments/73i3ne/why_do_app...

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

#739

Post by noahtallen » Tue, Oct 19, 2021, 2:10 AM UTC

boardwaalk wrote: post #501

out_of_protocol wrote: post #320

Earlier quoted context omitted.

On mobile devices - definitely On desktop - missing what, 2% or less? Checked Steam stats - yep, about 2%

That Steam stat is probably a chicken and the egg situation. I know I don’t run Steam on my Macbook because there’s nothing I want to play — but I would if there was. Still the Mac marketshare is not that high (~15%?) but might start looking attractive to developers looking to “get in first” when hardware that can actually run games becomes available (cough).

I mean, it’s similar to Linux, right? Linux has about 2% on Steam, and that’s with compatibility layers like wine allowing many windows games to run cross platform. This [0] puts Mac at 9.5% of operating systems overall and Linux at 2.4%.

But games with native Linux support are not very common compared to Windows, even though it’s mainly a matter of supporting Vulkan, which many modern games already do. My point is that even though Linux should be relatively easy to support natively (compared to mac not supporting cross-platform graphics APIs out of the box), devs aren’t putting the effort in.

I really hope this changes, and hopefully mac “gaming-level” hardware could help push cross-platform work along.

- 0: https://netmarketshare.com/operating-system-market-share.asp...

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

#740

Post by stirlo » Tue, Oct 19, 2021, 2:17 AM UTC

willis936 wrote: post #309

sunaurus wrote: post #35

I would like to upgrade to Apple silicon, but I have no need for a laptop. I hope they put the M1 Max in a Mac Mini soon.

There were rumors that it was supposed to be today. Given that it wasn't, I now expect it to be quite a while before they do. I was really looking forward to it.

Yep, If it didn't launch today it will launch alongside the M2 or Mac Pro update probably around WWDC next year.