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

Ask HN: Can competitors catch up to Apple Silicon?

news.ycombinator.com

Post reply on HN
Print view View original

211–220 of 323 posts

Re: Ask HN: Can competitors catch up to Apple Silicon?

#211

Post by modeless »

throw34 wrote: post #208
modeless wrote: post #188

Earlier quoted context omitted.

Obviously you can choose not to aim for the best possible performance in order to reduce costs. But for the companies that are competing for the absolute best performance at any cost, which doesn't include Qualcomm, by far the biggest factor determining that performance is the fabrication technology.

That’s an interesting take given this thread because Apple has explicitly stated they don’t prioritize performance, they prioritize performance per watt, which is not the same thing. And that also shows there’s a whole design space here and just focusing on one dimension, process node, is overly simplistic.

Top end performance is power limited, so optimizing for performance per watt is almost the same thing as optimizing for performance.

Re: Ask HN: Can competitors catch up to Apple Silicon?

#212

Post by wffurr »

wmf wrote: post #190
seabrookmx wrote: post #181

Earlier quoted context omitted.

They don't have the RAM on package though.

This doesn't matter. I know no one will believe me, but there's no benchmark that can show the difference.

Isn't the packaging part of what makes the very wide memory controllers feasible?

Re: Ask HN: Can competitors catch up to Apple Silicon?

#213

Post by simonh »

PaulDavisThe1st wrote: post #56
simonh wrote: post #18

Yes, but the reasons are to do with business and investment as much as technology. Apple utterly dominates the mobile phone CPU landscape, and it’s hard to see that changing. The reason is that they capture a huge slice of the profits in mobile phones. Something like 90%. For a while they capture over 100% of profits, because so many of their competitors were operating at a loss. This enables them to make investments…

> Apple utterly dominates the mobile phone CPU landscape, That seems to be an interesting definition of "dominates". 14% of global smartphones from independents. 24% by Apple's claims. A curious kind of domination.

The thread is about performance and technology, and Apple dominates the phone CPU landscape in those areas. I thought that was clear as my comment mainly addressed technology and the driving forces behind it, but I’m sorry for the confusion.

Re: Ask HN: Can competitors catch up to Apple Silicon?

#214

Post by n7pdx »

modeless wrote: post #114

You're looking in the wrong place. The magic comes from TSMC, not Apple. Apple's major innovation is in using their unrivaled bank account to pay TSMC for exclusivity on their newest fabrication technology. As competitors gain access to that TSMC technology they will match or beat Apple's performance. And if Intel succeeds in reclaiming the fabrication technology lead from TSMC (we'll see) then Intel will beat Apple'…

lol this will age well. i remember designing CPUs at intel when they had like a ~2.5 gen process lead (literally 4-5 years) and the result was still barely better than competition.

what outsiders don’t realize is intel management never had to deal with competition. they spent their entire careers fighting each other in petty internecine political battles. if you think intel vs amd is a feud, you haven’t seen intel arch/design vs intel fabs. and in that kind of environment, loyalists and politicians are promoted, not people willing to critically inspect your own processes and improve things.

end result: intel has a massive competence gap. in the extremely unlikely scenario where intel can regain the process lead, they will still be behind the curve on innovation and execution.

Re: Ask HN: Can competitors catch up to Apple Silicon?

#215

Post by FireBeyond »

uniqueuid wrote: post #2

The picture you're painting is far too simple. There are many metrics that play a role for different applications. A modern Nvidia card is magnitudes faster than any CPU and the M1 GPUs. A Nvidia A100 card has HBM memory with ~2TB/s bandwidth. On the other hand, there are areas where a ~5GHz intel CPU is competitive, or even older server chips with AVX can compete. So let's see what happens when Intel and AMD step to…

Very simple. There's also several articles discussing that how certain M2 configurations are actually lower performance than the M1, so this is hardly a done race.

Re: Ask HN: Can competitors catch up to Apple Silicon?

#216

Post by pclmulqdq »

tambourine_man wrote: post #143
fxtentacle wrote: post #90

Earlier quoted context omitted.

"even CUDA can be replaced without much disruption in most cases" Tell that to AMD who's been missing out on the very lucrative government data-center market because they can't replace CUDA well enough. BTW, PyTorch and TensorFlow still ship without AMD support despite AMD's ROCm being around for 6 years now.

Is that more an indictment of ROCm immaturity or of PyTorch/TensorFlow unwillingness to support it?

ROCm is pretty mature, but harder to support than CUDA. Its BLAS library works well. PyTorch and TensorFlow are likely getting a lot more help from Nvidia than from AMD on compatibility. The ROCm ecosystem is a lot smaller, too, so there are many fewer examples to copy.

Still, if you want to do floating point math with high performance for a lot cheaper than Nvidia, investing the time in ROCm is probably worth it.

Re: Ask HN: Can competitors catch up to Apple Silicon?

#217

Post by pclmulqdq »

fxtentacle wrote: post #206
tambourine_man wrote: post #143

Earlier quoted context omitted.

Is that more an indictment of ROCm immaturity or of PyTorch/TensorFlow unwillingness to support it?

I believe ROCm isn't usable yet. I don't know anyone who successfully trained a state of the art neural network on anything that was not NVIDIA. AMD keeps offering their own forks of PyTorch and TensorFlow, but last time I checked they were years outdated, and nobody uses them. So either PyTorch/TensorFlow are secretly sponsored by NVIDIA, or AMD's and Apple's CUDA replacements aren't mature enough yet. Given how man…

It is the former, that Nvidia effectively sponsors ML libraries. If you're training a neural network, and you are hiring expensive ML people, you're likely just going to pay the Nvidia tax than work on ROCm support.

My use of ROCm for signal processing has been pretty nice so far, and I think the ROCm BLAS library is up to par now.

Re: Ask HN: Can competitors catch up to Apple Silicon?

#218

Post by wmf »

wffurr wrote: post #212
wmf wrote: post #190

Earlier quoted context omitted.

This doesn't matter. I know no one will believe me, but there's no benchmark that can show the difference.

Isn't the packaging part of what makes the very wide memory controllers feasible?

Not really. GPUs regularly implement 256-bit or wider RAM.

Re: Ask HN: Can competitors catch up to Apple Silicon?

#219

Post by simonh »

modeless wrote: post #114

You're looking in the wrong place. The magic comes from TSMC, not Apple. Apple's major innovation is in using their unrivaled bank account to pay TSMC for exclusivity on their newest fabrication technology. As competitors gain access to that TSMC technology they will match or beat Apple's performance. And if Intel succeeds in reclaiming the fabrication technology lead from TSMC (we'll see) then Intel will beat Apple'…

It’s true 5nm definitely gives M1 an edge, but Apple’s ARM chips have lead the pack by a mile in performance for a long time, at comparable nodes. To say all the magic lies with TSMC is to claim that with M1 Apple suffered a huge loss is leadership position in terms of design, saved only by TSMC. I don’t see how that’s a credible claim.

Re: Ask HN: Can competitors catch up to Apple Silicon?

#220

Post by gary_0 »

pclmulqdq wrote: post #210
gary_0 wrote: post #158

Earlier quoted context omitted.

And those bytes matter when most Intel CPU cores have only a few dozen KB of L1I cache. Interestingly, the M1 and M2 performance cores have a whopping 192KB of L1 instruction cache.

ARM instructions aren't as compact as their x86 counterparts, so they need a big cache to get similar performance on cache-bound applications.

AFAIK, AArch64 instructions are roughly 2x longer than x64's average, but that's 3x as much cache as a comparable Intel core.
Post reply on HN