Hacker News Reader

Comment by bayindirh | Hacker News Reader

x86 is way more complex when compared to ARM processors, as a result their TDP is way higher when you request performance from them.

Intel had to reduce the frequency of their processors when running AVX2 instructions and the AVX2 frequency of the processors were non-disclosable to anyone.

Also, benchmarking Intel processors and publishing these numbers were forbidden in some cases. I don't know whether this ban is still in effect.

x86 processors can't keep up with the ARM processors TDP and thermal profile wise. So they slow down a ton when running on battery. See Jeff Geerling's last video on Apple Neo vs. some Intel laptop. It's as "efficient", but slow as a newborn tortoise learning to walk when unplugged and trying to get the most endurance out of the battery.

My M1 Mac gets almost 2 days of low-intensity use after ~6 years of use, and it got warm once or twice because something ran away in the background for tens of minutes.

View on HN

It doesn't really matter, because x86 processors inside are almost similar to modern ARM processors, the instructions that most programs use are more or less the same, and more complex "legacy" instructions are just emulated by splitting them into simpler instructions inside the CPU.

To me people that say that x86 is slow etc. never used an x86 processor with anything other than Windows. Yes, Windows is shit and laptops that run Windows are for the same reason shit. I get it. But on an x86 system you can run other OS, and, for example, a Linux distro with well tweaked power consumption parameters can get you even 3 days of battery life, or even more.

The point is that Windows keeps almost always the CPU above the minimum frequency, because it's full of useless background services, because programs are not well optimized for the hardware (to maximize backward compatibility they don't compile target x86_64 v3 for example and thus don't leverage on features and instructions available on new CPU or fall in the emulation case I've mentioned initially), and other reasons.

With a Linux distro you can keep the CPU at 600Mhz while web browsing, you can even turn off cores that you don't need, and the battery with this configuration surely lasts ages (at that point it becomes more predominant the consumption of other peripherals such as the display).

Being mistaken on the CISC vs RISC debate (besides, modern x86 is closer to RISC via micro-ops then old school CISC) is understandable. There is a lot of misinformation out there and myths, plus, it just feels right to consider CISC overly burdened, etc. TLDR: Intel x86_64 is closer to RISC then you likely think and Apple Silicon arm is closer to CISC then you likely think. These lines are blurry and have been for decades, very much for good reason.

But talking this authoritatively on something without doing the reading, that's grating: https://chipsandcheese.com/p/arm-or-x86-isa-doesnt-matter

> But talking this authoritatively on something without doing the reading, that's grating...

Thanks for your prejudice on me without knowing anything about me. In short, I'm a HPC sysadmin and programmer who works in a HPC center, separated from the actual hardware by a couple of floors.

We can discuss how transistors' heat generation doesn't discern about ISAs or being in a DAC or a cutting edge microprocessor, and we can even discuss how implementation of some functional blocks generate heat regardless of the ISA being involved. If you want we can discuss how saturating memory controllers affect pipeline saturation in processors even...

But talking this authoritatively on something with that amount of prejudice, that's grating.

After that comment, why does it matter where you work?

Because being too confident about someone you don't know is a bad habit. Where I work doesn't matter though, but what I do is.

Pointing me to C&C is a nice touch though. Not only I read the site and very article you sent me before, I used to consume Anandtech before that.

As a mere mortal, I can make mistakes and gladly accept them, but I can't accept rude replies. Pardon my French, but being called a low-key liar or smoke blower gets me a little upset.

> Pointing me to C&C is a nice touch though. Not only I read the site and very article you sent me before, I used to consume Anandtech before that.

All I'll say is, that's worse then. Presuming you had not read up before promoting a long disproven myth, that was an assumption by me, I'll admit that and maybe I should not have done that, my mistake. But it was a gracious mistake, it was done in your favour, it was giving you credit.

But still, your point was AFAIK already dismissed long ago.

Reply on news.ycombinator.com