Live data from Hacker News

Apple unveils M1, its first system-on-a-chip for portable Mac computers

9to5mac.com

711–720 of 1001 posts

Re: Apple unveils M1, its first system-on-a-chip for portable Mac computers

#711
post #271

Earlier quoted context omitted.

> I am very disappointed to see the design is identical to what we currently have Yeah one possible explanation I can think of is that they're not 100% confident in how rollout will go, and they want to avoid the situation where people in the wild are showing off their "brand new macs" which are still going through teething issues. There's less chance of souring the brand impression of M1 if they blend into the produ…

I think it could also be that a lot of people will be 'calmed' by the design they're used to when you're trying to convince them of a new chip the average consumer might not understand

I don't know about that, I think the confident move would have been to release the new chips along with a big design update.

As you say, I think the average consumer doesn't understand the difference between an Intel chip and an Apple chip, and will probably not understand what if anything has changed with these new products.

I would say developers would be the group which would be most anxious about an architecture change (which is probably why this announcement was very technically-oriented), and developers on average are probably going to understand that design changes and architecture changes basically orthogonal, and thus won't be comforted that much by a familiar design.

On the other side, average consumers probably aren't all that anxious due to the arch change, and would be more convinced that something new and exciting was happening if it actually looked different.

Re: Apple unveils M1, its first system-on-a-chip for portable Mac computers

#712
post #659
post #194

I'm interested to see the benchmarks vs. comparable AMD systems. Some of the claims, like 2x performance increase on the MBP are impressive, but intel laptops have been absolutely trounced by AMD 4000-series laptops of late. Also will be interested to see the benchmarks of the integrated GPU vs. discreet GPU performance.

Anandtech posted some comparisons of the A14 against Zen 3 today, which may be an interesting comparison: https://www.anandtech.com/show/16226/apple-silicon-m1-a14-de... Seems like the A14 is within 10-20% of the desktop 5950X in single threaded workloads. The M1 will probably close the gap with higher clock speeds. AMD will probably still be ahead on multithreaded workloads until Apple releases a chip with 8 high-pe…

Didn't they claim it's the fastest per-thread performance in the world?

Re: Apple unveils M1, its first system-on-a-chip for portable Mac computers

#713
post #638

Earlier quoted context omitted.

How many cores does that have, and how many cores does the i7 have though?

Why is this a mystery or rhetorical question? Look on the website? It's public: M1 > Apple M1 chip > 8-core CPU with 4 performance cores and 4 efficiency cores > 8-core GPU > 16-core Neural Engine Intel > 1.7GHz quad-core Intel Core i7, Turbo Boost up to 4.5GHz, with 128MB of eDRAM https://www.apple.com/macbook-pro-13/specs/ https://www.apple.com/shop/product/G0W42LL/A/refurbished-133... Apple's claim: > With an 8‑co…

The 2019 MBP 13 supposedly uses an 8th gen, 14nm Intel part (14nm is 6 year-old technology).

A more fair comparison would be Tiger Lake (20% IPC improvement) on Intel's terrible 10nm process. The most fair comparison would be zen 3 on 7nm, but even that is still a whole node behind.

Re: Apple unveils M1, its first system-on-a-chip for portable Mac computers

#714

Completely unconfirmed speculation incoming: There's a solid chance that the logic board is exactly the same on all of the Macs announced today and the only difference is the cooling solution. If you play around with the Apple Store configurator, the specs are all suspiciously similar between every new Mac. https://www.apple.com/shop/buy-mac/macbook-air https://www.apple.com/shop/buy-mac/macbook-pro/13-inch https://w…

One of the slides mentioned that the air is limited to 10watts though. I wonder if it does have the same soc but its nerfed beyond 10watts.

Re: Apple unveils M1, its first system-on-a-chip for portable Mac computers

#715
post #706

Earlier quoted context omitted.

This makes sense. Most likely this is why the CPUs are all limited to 16GB. It's likely when they unwrap the 16 inch MacBook Pro, it will open up more configurations (more RAM in particular!) for the 13" MacBook Pro and hopefully the mini.

RAM limits are pretty easy to explain. 16GB chips cost disproportionately more and use more power. I wonder if they use 2 4GB chips or 1 8GB chip in the low-end SKU?

It's even easier to explain than that. The RAM is integrated into the CPU. While there are a few SKUs here, Apple only designed and built one CPU with 16GB RAM. The CPUs are binned. The CPUs where all RAM passed testing are sold as 16GB, the 8GB SKUs had a failure on one bank of RAM.

There are no 32 or 64 GB models because Apple isn't making a CPU with 32 or 64GB of RAM yet.

Re: Apple unveils M1, its first system-on-a-chip for portable Mac computers

#716

Max 16GB of RAM on these new machines is really not that great to be honest. The mini supported up to 64GB before.

16GB is fine for mobile and low end. I'm a developer and struggle to fill that amount of memory, even with VMs running. I guess they made a cost trade-off for these machines, which will not be carried forward to the high-end. Perhaps a new "M2" chipset, with discreet RAM and graphics. Next year? It would be amazing if they could bring ECC for a decent price as well, time will tell.

I'm a rails dev and I constantly struggle with 16GB. Once you start up rails, background workers, webpack, vs code, MS teams, a database, plus your web browser you very quickly run out of memory.

Re: Apple unveils M1, its first system-on-a-chip for portable Mac computers

#717
post #498

Earlier quoted context omitted.

They probably beat AMD/Intel on perf/power efficiency, which is why it makes sense for MBA and 13" MBP. The smaller machines are also likely held back by cooling solutions, so if you have Intel beat on power efficiency in a tiny form factor, you can boost your clock speed too.

Considering how AMD also beats intel on power/perf by a wide margin and they compared their results against Intel CPUs I wouldn't be surprised if their power/performance was close to AMD (they do have heterogeneous CPU cores which of course is not the case in traditional x86)

AMD's Smart Memory Access would like to have a word. I'd note that in unoptimized games, they're projecting a 5% performance boost between their stock overclock and SMA (rumors put the overclock at only around 1%).

https://www.amd.com/en/technologies/smart-access-memory

Re: Apple unveils M1, its first system-on-a-chip for portable Mac computers

#718

Apple M1: 192KB I-Cache, 128KB D-Cache, 12MB L2 Cache AMD Ryzen 9 5950X: 32KB I-Cache, 32KB D-Cache, 8MB L2 Cache Isn’t this difference huge? What am I missing here?

The L2 cache on the M1 is the last level cache (LLC), where the 5950X has a 64MB L3 cache for LLC. Also I'm not sure we know yet how much of the chip is using that L2, it might be more than just the four high perf cores.

A closer comparison is probably Intel's 10900K which has a 20MB L3.

Re: Apple unveils M1, its first system-on-a-chip for portable Mac computers

#719

Earlier quoted context omitted.

I can't see how something that tiny can compete in any meaningful way with a giant nVidia type card for training. I'd imagine it's more for running models that have been trained already, like all the stuff they mentioned with Final Cut.

Isn't it better to rent a cloud with as many GPUs as necessary for a time needed to train the model? I don't know state of things in ML.

Not necessarily.

It can be surprisingly cost-effective to invest a few $k in a hefty machine(s) with some high-end GPU's to train with due to the exceedingly hefty price of cloud GPU compute. The money invested up-front in the machine(s) pays itself off in (approximately) a couple of months.

The "neural" chips in these machines are for accelerating inference. I.e. you already have a trained model, you quantise and shrink it, export it to ONNX or whatever Apple's CoreML requires, ship it to the client, and then it runs extra-fast, with relatively small power draw on the client machine due to the dedicated/specialised hardware.

Post reply on HN