Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Apple Silicon M1: Black Magic Fuckery

singhkays.com

Post reply on HN
Print view View original

981–990 of 1001 posts

Re: Apple Silicon M1: Black Magic Fuckery

#981

Post by ojosilva »

ed25519FUUU wrote: post #401
zamadatix wrote: post #360

Earlier quoted context omitted.

It'll be interesting what happens in a couple of years if server side gets more mixed with ARM as well.

Unless AWS starts selling their graviton CPU, whose going to create chips for the arm server market? The volumes are huge and the margins are tiny, which is not something Apple was ever interested in.

For a myriad of reasons, I don't think Apple would ever want to sell their chips standalone.

OTOH, Apple could consider releasing a line of Apple silicon cloud servers.

Even if more expensive than x86 servers, they could eventually become VERY competitive if they are more power efficient, which is a real issue for cloud providers. It would pay off any upfront cost if, on the long run, the server uses ie. 50% less energy than x86 alternatives for the same bang.

Now, the thing is that the typical cloud or datacenter machine today is more of a custom built piece of witchcraft than an off-the-shelf blade. So, meh, maybe not. But an ARM-based cloud, Apple or not, sound like the way to go, at least for the long term.

Re: Apple Silicon M1: Black Magic Fuckery

#982

Post by mikepurvis »

colejohnson66 wrote: post #936
mikepurvis wrote: post #803

Earlier quoted context omitted.

Sure, but there's a legit gap in the datacenter— not having a sanely, legally rackable OS X machine is a pretty big problem for a lot of organizations. Not everyone wants to do their Jenkins builds or generate homebrew bottles on a Mac Mini under someone's desk.

Isn’t the newest Mac Pro available in rack form?

Ah, so it is, and the thermal story there is definitely much better than with the Mini, there being a clear intake/exhaust flow. OTOH, there's still likely a gap in terms of management features, and the starting price of $6.5k for a 4U system is definitely going to be a barrier for some. Good to know there's at least something, anyway.

Re: Apple Silicon M1: Black Magic Fuckery

#983

Post by elliottkember »

diveanon wrote: post #980

I've seen lots of M1 benchmarks, but has anyone done a side by side comparison of what it is like to actually get work done on one? Take a conventional dockerized local dev environment and just start building stuff. How much time do you spend working around M1 arch issues versus building your app? This is the key factor that is keeping me from being an early adopter. I don't get paid to figure out how to work on a ne…

Good questions. Docker doesn’t work yet. My node and rails apps work great. React Native works great. It’s weirdly fine.

VSCode was slow for a bit, but I think it hadn’t synced my file exclude settings.

Re: Apple Silicon M1: Black Magic Fuckery

#984

Post by CyberRabbi »

jamil7 wrote: post #529
CyberRabbi wrote: post #485

These ppl are so drunk on the kool aid they don’t even realize that 80% of their experience is common to every new laptop purchase. In 1 year these “revolutionary” computers will be boring and slow again, especially when their batteries start to deteriorate.

Do you really believe this is a common hardware release? There’s nothing revolutionary here? People love to hate on Apple which is fair for a lot of reasons but I think you have to give credit where credit is due. The M1 will push all laptops forward and force others to complete which is great for customers.

I didn’t say there is nothing different here, only that the tangible differences are marginal to the exaggerated good experiences people are reporting. Apple has not revolutionized how lithium batteries work and the power savings by the M1 are less than 100% improvement, so the claims about battery life are clearly exaggerated and mostly common to all new laptop purchases. Apple makes decent products but they make even better marketing.

Re: Apple Silicon M1: Black Magic Fuckery

#985

Post by kllrnohj »

saagarjha wrote: post #952
kllrnohj wrote: post #836

Earlier quoted context omitted.

> On recent Apple Silicon CPUs uncontended most atomic operations are essentially free - almost identical in speed to the non-atomic version of the same operation. They are fast for atomics but still far, far slower than the equivalent non-atomic operation. An add operation takes around half a cycle (upper bound here - with how wide the firestorm core is an add operation is almost certainly less than half a cycle). A…

> So you're still talking the atomic operation being >10x slower than non-atomics. How did you arrive at this number?

> How did you arrive at this number?

It's in the post. Half a cycle for an add or less, and cycles are every 1/3 nanosecond. So upper bound for an add would be around 1/6th a nanosecond. Likely less than that still yet, since the M1 is probably closer to an add in 1/8th a cycle not 1/2. Skylake by comparison is at around 1/4th a cycle for an add, and since M1's IPC is higher it's not going to be worse at basic ALU ops.

6 nanoseconds @ 3ghz is 18 cycles. That's on the slow end of the spectrum for a CPU instruction.

Re: Apple Silicon M1: Black Magic Fuckery

#986

Post by freeone3000 »

johncalvinyoung wrote: post #837
floatboth wrote: post #828

Earlier quoted context omitted.

AMD would be able to do DRAM on package for the lowest wattage "ultrabook" chips, at the cost of producing a very different package for them vs. the bigger laptops that are expected to have upgradable SODIMMs. But I doubt that this "co-location" is that huge for performance. Whatever memory frequency and timings Apple is using are likely easily achievable through the regular mainboard PCB, maybe at the cost of slight…

Yeah, I'm no kernel developer, but I've been replying to anyone saying 'just stick n * M1 in it' that even AMD has been trying to move back to more predictable memory access latency, less NUMA woes.

But in general we're moving toward even less uniform memory, with some of it living on a GPU. NUMA pretended that all memory was the same latency, because C continues to pretend we're on a faster PDP-11, but this seems like a step in the wrong direction as for how high-performance computation is progressing.

Re: Apple Silicon M1: Black Magic Fuckery

#987

Post by DarkmSparks »

rthomas6 wrote: post #882
DarkmSparks wrote: post #523

One of the things I've noticed recently, and especially since the CPU space started finally moving again is how much of a divide there now is between the computer literate, and the computer illiterate. It probably creates a social divide at least that which existed when the majority of people couldn't read or write, and is just as "not OK". Example of this in the first paragraph of this article: > For everyday users…

Hasn't it always been this way? The only difference is now there are a lot more computer-illiterate people using computers.

Wouldn't say so. As Kovek so very nicely put it further up:

"Computer Literacy provides tools to interact with information."

Easy to forget the iPhone was ~2007, prior to this computers were still a fairly niche market and Android still in its infancy, computer literacy gave an advantage but computer illiteracy wasn't really a big disadvantage.

Now literally every industry from mom and pop stores to basic healthcare demand a pretty high level of computer literacy to the point they don't survive long without it.

Re: Apple Silicon M1: Black Magic Fuckery

#988

Post by zamadatix »

harpratap wrote: post #499
zamadatix wrote: post #447

Earlier quoted context omitted.

Apple isn't just now entering into this space with the M1 though, it's been this bad for 5 years vs Qualcomm already the change is now even the best power hungry x86 cores can't keep up with it. What Apple is delivering in the M1 continues to be 2 generations ahead of the Qualcomm's best in terms of performance an the +10% uplift on the usual +20% generational increase isn't changing that - Apple has been doing that…

I am not really a knowledgeable person when it comes to CPUs, I just follow what Anandtech posts. And reading his comments on X1 it looks like ARM (and thus Qualcomm) are just being too conservative because there was really no business need for a non-Apple company to make such a large chip. But now there is, so it won't be impossible for them to catch up. Huawei's Kirin 9000 is already scoring close to A14 in multi c…

Qualcomm's best phone chip (the 865) has 56% the single thread performance of the A14 as found in the iPhone 12. I'm just not following how suddenly with the M1 all Qualcomm needs to do is make a chip of the same size and it will be just as performant. The disparity in performance isn't new with the M1, Qualcomm ARM CPUs have been slower than Apple ARM CPUs for nearly a decade now. What was amazing about the M1 is that it beats even the best power hungry x86 chips in many single thread tasks as well, not that it suddenly jumped ahead Qualcomm (Apple was already ahead of Qualcomm). Also the business case isn't new, Qualcomm has been trying to display x86 laptop chips for years and has fallen short on performance every attempt. The M1 does not present a new challenge or new opportunity in this space, just better execution.

Re: Apple Silicon M1: Black Magic Fuckery

#989

Post by saagarjha »

kllrnohj wrote: post #985
saagarjha wrote: post #952

Earlier quoted context omitted.

> So you're still talking the atomic operation being >10x slower than non-atomics. How did you arrive at this number?

> How did you arrive at this number? It's in the post. Half a cycle for an add or less, and cycles are every 1/3 nanosecond. So upper bound for an add would be around 1/6th a nanosecond. Likely less than that still yet, since the M1 is probably closer to an add in 1/8th a cycle not 1/2. Skylake by comparison is at around 1/4th a cycle for an add, and since M1's IPC is higher it's not going to be worse at basic ALU op…

Where? 6 nanoseconds is pretty long, that’s about how long it’d take to do an entire retain/release pair, which is a couple dozen instructions I believe.

Re: Apple Silicon M1: Black Magic Fuckery

#990

Post by mlindner »

lazyjones wrote: post #924
mlindner wrote: post #666

Earlier quoted context omitted.

People are exaggerating a little bit, but this is the low end machine with 8GB RAM and only 7 GPU cores. So you can see it's pretty good in worst case. https://www.youtube.com/watch?v=UQoGPLO8zBI

It drops to 25fps with not much happening on the screen, no water and no crowded area, at 1440p. This guy also mentions dropping to 30fps: https://www.youtube.com/watch?v=2ubyXiY4N2o The comments here also say 25fps on max settings in Stormwind: https://www.youtube.com/watch?v=mIbEIhizoXY

Yes but again, this is missing one of the GPU cores, so that's quite a drop in performance.
Post reply on HN