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

Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

semianalysis.substack.com

Post reply on HN
Print view View original

91–100 of 206 posts

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#91

Post by Sakos »

modeless wrote: post #27
daguava wrote: post #2

There's a lot of claims of poached talent in the article, basically claiming [paraphrasing] "Apple, maintaining their stressful work env and not paying to shore that up lost some rockstars" How true is this? If they're on the money it's an excellent example of a talent retention miss leading to a demonstrable mediocrity in delivery.

Seems silly to me. CPU designs are important, but these companies have more than enough engineers to make competent designs even with some people leaving. There's another factor that completely dominates. It's all about the fabs. Intel lost the performance lead, was it because of their designs? No, it's because they lost the lead in fabs. AMD passed Intel, was it because of their designs? No, it's because they use TS…

Key talent is incredibly important. If you lose enough senior engineers, it doesn't matter how talented the rest are. You've lost so much institutional knowledge that is either extremely difficult or impossible to regain. And Apple is notoriously under-staffed for a lot of their projects. With the staffing losses to Nuvia, I wouldn't be surprised if they lost enough key talent that it's going to take them a long time to recover and be able to deliver significant performance improvements again. That's what happens when you treat software developers/engineers like commodities.

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#92

Post by theplumber »

otikik wrote: post #73
riffraff wrote: post #69

Earlier quoted context omitted.

My dad used to tell a story when as a kid he worked at a gas station. Guy shows up with a Ferrari, asks for a full tank, pays with a large note and my dad asks him to keep the change. The guy replied - boy, I got a lot of money Really, a lot. You know how I got them? I never gave anything away for free. Hand that change over.

That just illustrates that a lot of rich guys are entitled assholes. The correlation between assholery and driving an expensive car in particular has been studied. https://youtu.be/1EHhFwGeQLc

The more we automate the better. Computers don't ask to keep the change.

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#93

Post by brigade »

inkyoto wrote: post #77
brigade wrote: post #60

Earlier quoted context omitted.

Android flagships are shipping with SVE2 as of this year, which I actually didn't realize until like two weeks ago because there's been nearly zero buzz about it. What's SVE2 being used for over NEON as of now?

Low level runtime optimisation that yields substantial performance gains in the user facing or system level software, ranging from cryptography through to data processing algorithms and very high throughput JSON parsing. Take OpenSSL as an isolated example. By simply fiddling with the C compiler flags to allow it to use NEON on M1, the sha256 calculation speed-up is 4x for 128 and 256 block sizes, with performance ga…

Yeah that's NEON. And there's instructions that literally calculate SHA256 so generalizing that is moot. My point was first, what real benchmarks are there of SVE2's benefits over NEON with mainstream CPUs that M2 would compete against? Unlike AVX-512, NEON was already pretty rich, so the new instructions have rather specialized usefulness.

Because outside of servers where little cores don't exist, 256b ALUs in big cores mean 256b registers in little cores, and Cortex-A510 is way smaller than Gracemont. And then you're giving Samsung another opportunity to screw up big.LITTLE...

And even the server CPUs with SVE are 2x256b except A64FX which is HPC exclusive, so no better than 4x128b.

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#94

Post by brigade »

inkyoto wrote: post #86
brigade wrote: post #76

Earlier quoted context omitted.

Soo... basically a 2x speedup in going from 4x128b to 2x512b ALUs, after discounting the frequency difference. But realistically, Intel's client configurations are 3x256b, which is only 25-40% faster in that paper. (I suspect any application doing enough quicksort that the 2x speedup is significant, would be even happier going slightly off-core to a coprocessor more specialized in vector processing, like Hwacha. Ther…

> Soo... basically a 2x speedup in going from 4x128b to 2x512b ALUs, after discounting the frequency difference. But realistically, Intel's client configurations are 3x256b, which is only 25-40% faster in that paper. 2.4x difference was, in fact, reported, however I still find it somewhat difficult to interpret the reported results. The processing unit size difference alone and the number of LU's can't account for su…

If something is bound by memory/cache bandwidth, then increasing ALU width wouldn't help in the first place.

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#95

Post by wikibob »

happyopossum wrote: post #49
exhilaration wrote: post #21

Earlier quoted context omitted.

I'm still amazed at how jumping to a rival firm like this is possible in California, those of us in pretty much the rest of the country are locked behind non-competes.

That’s quite an exaggeration - many states disallow or severely limit non-competes, and in many of the states that allow them, they are often unenforced, or easy to get around. So yeah - some people in some parts of the country are locked behind non-competes (if they aren’t willing to move), but it’s hardly everyone.

That is a gross exaggeration of the situation.

California has a total ban on non-competes.

A very small handful of other states put restrictions on non-competes, but even those generally allow non-competition agreements if time limited, and the employee makes over ~$100k.

It’s widely accepted that prohibiting non-competes has been a significant factor in the tech industry success in California.

As one example, it is well known that Amazon aggressively enforces non-competes, even against line engineers.

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#96

Post by inkyoto »

modeless wrote: post #89
inkyoto wrote: post #79

Earlier quoted context omitted.

«Whole system performance» is a meaningless term as it is a function of many, usually poorly controlled, input variables, and your whole system is different from my whole system. If my VPN tunnel allows me to have faster transfer speeds simply by virtue of having ISA assisted optimisations in the cryptographic library it uses, the net result will be very noticeable to me but perhaps not for you unless you also have t…

The idea that my overall browser performance might improve in a noticeable way because it is switching from NEON optimized Quicksort to SVE2 optimized Quicksort is simply laughable. On the other hand, switching to a processor fabbed on a better process node could easily have a noticeable impact on overall browser performance, or battery life while browsing, or both.

It is certainly not laughable to me. You have singled out Quicksort as the sole example of performance gains whereas I have used it as a single isolated example of very large performance gains that can be had. SIMD instructions have seen a lot of other mainstream use cases recently which also includes the memory copying or memcpy(3) optimisations amongst others. Your browser has a Javascript engine, and since it a Javascript engine, it has a garbage collector. Garbage collectors move memory blocks around all the time, and a SIMD optimised memcpy will yield substantial performance gains. Or faster JSON processing. Therefore, SIMD + an improved fab process will result in much larger overall performance gains for you and me as browser users as opposed to an improved fab process alone. It is also a realistic example of the «whole system performance» improvements if the browser is treated as a «whole system».

And an optimised QuickSort can also come in handy if one pokes around a large browser history or uses it as a knowledge base, which I do and use it on a regular basis. My browser keeps a uninterrupted record of all visited websites over the last 15+ years and being able to zoom in on a particular time span to find something within that temporal range quickly is important to me. I am almost certain that a sorting of sorts is involved somewhere behind the scenes.

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#97

Post by heurisko »

riffraff wrote: post #69
TylerE wrote: post #38

Earlier quoted context omitted.

Things don't become the "est" anything by being wasteful.

My dad used to tell a story when as a kid he worked at a gas station. Guy shows up with a Ferrari, asks for a full tank, pays with a large note and my dad asks him to keep the change. The guy replied - boy, I got a lot of money Really, a lot. You know how I got them? I never gave anything away for free. Hand that change over.

Reminds me of the Simpsons episodes when Bill Gates "buys out" Homer's internet business.

"Well I didn't get rich by writing a lot of cheques!"

https://m.youtube.com/watch?v=H27rfr59RiE

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#98

Post by danuker »

theplumber wrote: post #92
otikik wrote: post #73

Earlier quoted context omitted.

That just illustrates that a lot of rich guys are entitled assholes. The correlation between assholery and driving an expensive car in particular has been studied. https://youtu.be/1EHhFwGeQLc

The more we automate the better. Computers don't ask to keep the change.

Google doesn't ask whether you agree to account suspension either.

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#99

Post by otikik »

imwillofficial wrote: post #84
otikik wrote: post #73

Earlier quoted context omitted.

That just illustrates that a lot of rich guys are entitled assholes. The correlation between assholery and driving an expensive car in particular has been studied. https://youtu.be/1EHhFwGeQLc

Quoted post unavailable.

The assholery part here is not about who keeps the change. It’s about telling a gas station boy “that’s how you got rich”. What makes you rich is first of all being born rich, which most rich people tend to ignore. On the few cases were we are really talking about people with humble beginnings, it’s all about making your time count. If the guy was really rich, the time it would take the boy to get inside, take 2 dollars from the cash register and give them to him was just not worth it. It was all self aggrandising bravado.

Re: Apple M2 Die Shot and Architecture Analysis – Big Cost Increase and A15 Based IP

#100

Post by nicky0 »

otikik wrote: post #99
imwillofficial wrote: post #84

Earlier quoted context omitted.

Quoted post unavailable.

The assholery part here is not about who keeps the change. It’s about telling a gas station boy “that’s how you got rich”. What makes you rich is first of all being born rich, which most rich people tend to ignore. On the few cases were we are really talking about people with humble beginnings, it’s all about making your time count. If the guy was really rich, the time it would take the boy to get inside, take 2 doll…

I doubt it's even a real story. I've heard variations of it. The basic idea is "just because I'm rich doesn't mean I'm giving stuff away, how do you think I got (and stay) rich?". The ferarri, the $50 note, and other details are just embellishments in this version of the tale.
Post reply on HN