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

Ask HN: Why hasn't x86 caught up with Apple M series?

news.ycombinator.com

Post reply on HN
Print view View original

201–210 of 640 posts

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#201

Post by fafhnir »

I have the same experience here with my MacBook Air M1 from 2020 with 16GB RAM and 512GB SSD. After three years, I upgraded to a MacBook Pro with M3 Pro, 36GB of RAM, and 2TB of storage. I use this as my main machine with 2 displays attached via a TB4 dock.

I'm working in IT and I get all new machines for our company over my desk to check them, and I observed the exact same points as the OP.

The new machines are either fast and loud and hot and with poor battery life, or they are slow and "warm" and have moderate battery life.

But I had no business laptop yet, ARM, AMD, or Intel, which can even compete with the M1 Air, not to speak of the M3 Pro! Not to speak about all the issues with crappy Lenovo docks, etc.

It doesn’t matter if I install Linux or Windows. The funny point is that some of my colleagues have ordered a MacBook Air or Pro and use their Windows or Linux and a virtual machine via Parallels.

Think about it: Windows 11 or Linux in a VM is even faster, snappier, more silent, and has even longer battery life than these systems native on a business machine from Lenovo, HP, or Dell.

Well, your mileage may vary, but IMHO there is no alternative to a Mac nowadays, even if you want to use Linux or Windows.

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#202

Post by aurareturn »

noelwelsh wrote: post #199
aurareturn wrote: post #117

Earlier quoted context omitted.

On Cinebench 2025 single threaded, M4 is roughly 4x more efficient and 50% faster than Strix Halo. These numbers can be verified by googling Notebookcheck. How many iterations to match Apple?

> How many iterations to match Apple? Why are you asking me? I'm not in charge of AMD. Yes the Strix Halo is not as fast on the benchmarks as the M4 Max, its bandwidth is lower, and the max config has less memory. However, it is available in a lot of different configurations and some are much cheaper than comparable M4 systems (e.g. the maxed out Framework desktop is $2000.) It's a tradeoff, as everything in life is.…

  Why are you asking me? I'm not in charge of AMD.
Because you claimed this so I thought you knew:

  A few iterations of this should be comparable to the M series

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#203

Post by svantana »

sys_64738 wrote: post #166
kangs wrote: post #136

Earlier quoted context omitted.

unfortunately, contrarily to popular belief, you cannot run Linux natively on recent macbooks (m4) today.

Depends what "natively" means. You can virtualize Linux through several means such as Virtual Box.

...but you won't get similar power efficiency, which was claimed.

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#204

Post by KingOfCoders »

1. Memory soldered to the CPU

2. Much more cache

3. No legacy code

4. High frequencies (to be 1st in game benchmarks, see what happens when you're a little behind like the last Intel launch, the perception is Intel has bad CPUs because they are some percentage points behind AMD on games, pressure Apple doesn't have - comparisons are mostly Apple vs. Apple and Intel vs. Amd)

The engineers at AMD are the same as at Apple, but both markets demand different chips and they get different chips.

Since some time now the market is talking about energy efficiency, and we see

1. AMD soldering memory close to the CPU

2. Intel and AMD adding more cache

3. Talks about removing legacy instructions and bit widths

4. Lower out of the box frequencies

Will take more market pressure and more time though.

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#205

Post by yaro330 »

aurareturn wrote: post #167
ben-schaaf wrote: post #101

Battery efficiency comes from a million little optimizations in the technology stack, most of which comes down to using the CPU as little as possible. As such the instruction set architecture and process node aren't usually that important when it comes to your battery life. If you fully load the CPU and calculate how much energy a AI340 needs to perform a fixed workload and compare that to a M1 you'll probably find s…

All that to say: M1 is pretty fast, but the reason the battery life is better has to do with everything other than the CPU cores. That's what AMD and Intel are missing. This isn't true. Yes, uncore power consumption is very important but so is CPU load efficiency. The faster the CPU can finish a task, the faster it can go back to sleep, aka race to sleep. Apple Silicon is 2-4x more efficient than AMD and Intel CPUs d…

> Apple Silicon is 2-4x more efficient than AMD and Intel CPUs during load while also having higher top end speed.

This is false, in cross platform tasks it's on par if not worse than latest X86 arches. As others pointed out: 2.5h in gaming is about what you'd expect from a similarly built X86 machine.

They are willing due to lower idle and low load consumption, which they achieve by integrating everything as much as possible - something that's basically impossible for AMD and Intel.

> The faster the CPU can finish a task, the faster it can go back to sleep, aka race to sleep.

May have been true when CPU manufacturers left a ton of headroom on the V/F curve, but not really true anymore. Zen 4 core's power draw shoots up sharply pass 4.6 GHz and nearly triples when you approach 5.5 GHz (compared to 4.6), are you gonna complete the task 3 times faster at 5.5 GHz?

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#206

Post by RossBencina »

jonwinstanley wrote: post #185
ben-schaaf wrote: post #101

Battery efficiency comes from a million little optimizations in the technology stack, most of which comes down to using the CPU as little as possible. As such the instruction set architecture and process node aren't usually that important when it comes to your battery life. If you fully load the CPU and calculate how much energy a AI340 needs to perform a fixed workload and compare that to a M1 you'll probably find s…

A huge reason for the low power usage is the iPhone. Apple spent years incrementally improving efficiency and performance of their chips for phones. Intel and AMD were more desktop based so power efficiency wasnt the goal. When Apple's chips got so good they could transition into laptops, x86 wasn't in the same ballpark. Also the iPhone is the most lucrative product of all time (I think) and Apple poured a tonne of t…

I thought they just acquired P.A. Semi, job done.

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#207

Post by daeken »

chasil wrote: post #176
aurareturn wrote: post #73

Earlier quoted context omitted.

Per core, Apple’s Performance cores are no bigger than AMD’s Zen cores. So it’s a myth that they’re only fast and efficient because they are big. What makes Apple silicon chips big is they bolt on a fast GPU on it. If you include the die of a discrete GPU with an x86 chip, it’d be the same or bigger than M series. You can look at Intel’s Lunar Lake as an example where it’s physically bigger than an M4 but slower in C…

Is it not true that the instruction decoder is always active on x86, and is quite complex? Such a decoder is vastly less sophisticated with AArch64. That is one obvious architectural drawback for power efficiency: a legacy instruction set with variable word length, two FPUs (x87 and SSE), 16-bit compatibility with segmented memory, and hundreds of otherwise unused opcodes. How much legacy must Apple implement? Non-ke…

There's no AArch32 or Thumb support (A32/T32) on M-series chips. AArch64 (technically A64) is the only supported instruction set. Fun fact: this makes it impossible to run Mario Kart 8 via virtualization on Macs without software translation, since it's A32.

How much that does for efficiency I can't say, but I imagine it helps, especially given just how damn easy it is to decode.

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#208

Post by KingOfCoders »

PaulKeeble wrote: post #60

I tend to think its putting the memory on the package. Putting the memory on the package has given the M1 over 400GB/s which is a good 4x that on a usual dual channel x64 CPU and the latency is half that of going out to a DRAM slot. That is drastic and I remember when the northbrige was first folded into the CPU by AMD with the Athlon and it had a similarly big improvements in performance. It also reduces power consu…

Yes, this saves a lot of power and adds performance. But destroys your eco system and annoys a vocal user base. Apple has no eco system and lots of fans, so they are playing their cards right.

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#209

Post by zillow »

mnw21cam wrote: post #184
mkl wrote: post #138

Earlier quoted context omitted.

Sure you can. There are a bunch listed in this article: https://www.ultrabookreview.com/6520-fanless-ultrabooks/ Fanless x86 desktops are a thing too, in the form of thin clients and small PCs intended for business use. I have a few HP T630s I use as servers (I have used them as desktop PCs too, but my tab-hoarding habit makes them throttle a bit too much for my use - they'd be fine for a lot of people).

Do you have a version of that web page for people who want to run Linux? That'd be particularly helpful.

its x86, they all run linux. x86 (as in amd64) is standardized

Re: Ask HN: Why hasn't x86 caught up with Apple M series?

#210

Post by KingOfCoders »

aurareturn wrote: post #117
noelwelsh wrote: post #99

Like a few other comments have mentioned, AMD's Strix Halo / AI Max 380 and above is the chip family that is closest to what Apple has done with the M series. It has integrated memory and decent GPU. A few iterations of this should be comparable to the M series (and should make local LLMs very feasible, if that is your jam.)

On Cinebench 2025 single threaded, M4 is roughly 4x more efficient and 50% faster than Strix Halo. These numbers can be verified by googling Notebookcheck. How many iterations to match Apple?

On one of the few workloads where massive parallelism makes sense, why quote a single threaded number? I'm curious.
Post reply on HN