Hacker News Reader

Comment by QuixoticQuibit | Hacker News Reader

I really wonder what this says about the x86 platform going forward.

Mobile completely passed it by.

I’ve been seeing more hype about ARM servers for a while with AWS Graviton instances, the new #1 supercomputer in the TOP500, etc.

And of course today we see that Apple plans to transition their Macs to their own ARM chips. Even Microsoft made an ARM-based Windows/Surface product but it didn’t seem to amount to much. I wonder if they’ll want to make another stab at it seeing Apple’s direction with strong vertical integration.

While I don’t think x86 as a platform is going away anytime soon, I feel like its market share and by extension its relevance will slowly dwindle over the next decade or two. Interesting times.

View on HN

I too don't see x86 disappearing soon but it feels like the world has changed and that change is not positive for Intel.

We've been used to x86 dominance on the desktop and servers for so long that I think a future where there are two architectures with critical mass and one of the architectures can be licensed by a number of firms is hard to imagine.

There will be some short term effort and pain but it must surely be a better competitive environment than we have now.

The historic / current Intel and AMD duopoly has surely not been healthy.

Intel has either squandered the tech and we will see drastic improvements soon or possibly they’ll fall closer toward the dustbin of history. I suspect the later since they’ve been stuck at various points of specifications for a decade plus. Or they do something novel for once. I wouldn’t count them out of a breakthrough into new architectures/fab processes entirely.

My comment was meant to be more general than Intel. I’m even including AMD here. x86 has no presence in mobile (that ship has sailed) and again I’m seeing quite a lot of innovation on ARM side wrt server-class processors. Not sure how traditional PCs will play out but Macs are going to split off from x86 within years.

I totally agree, I think it would be a good thing if x86 were to be replaced by something newer. Intel could even adapt and continue to play a leading role, that is another question.

They couldn't even execute on Edison. Intel will be gone in 20 years. They couldn't manage their way out of a bag.

Well they have spent the last 4 years trying to fix wave after wave of side-channel vulnerabilities. A handful of those affected ARM and AMD as well, but a very large number of them were Intel-only, and were a direct side effect of Intel cutting corners on safety to get advances in their performance numbers.

It's astonishing to me how far Intel has fallen in so short a time. I feel like it was only a couple years ago that Intel was understood universally to be the "heavyweight champion of the world" so to speak.

They completely missed the boat on mobile and AMD has leapfrogged them very recently on desktop. On top of that there were the Spectre vulnerabilities which shook confidence even further. This announcement is another huge blow given the extent to which the entire consumer electronics industry tends to follow Apple. I would be interested to hear an insider's perspective on such a rapid decline.

Not an insider, but as far as I know, it's mostly their fab that failed. They're still on 14+++nm when AMD is on 7nm, with 5nm coming soon.

You can have the best architecture in the world (no idea if they have) and the best engineers, it's hard to compete when you're so far behind in transistor size.

Do you have any good resources on why their fab process failed so hard? I've seen it mentioned frequently, but not the reason why.

They tried to jump too far with intel 10nm. The process node names no longer represent actual transistor size, they just indicate a new generation. With intel 10nm they took a risk and tried to shirk the die more then would normally happen in generation jump. If it worked it would put them a whole generation ahead of TSMC and secure process leadership. But physics bit them in the ass. It turned out that shirking that small was much harder then they thought, and the cells they were using were not robust enough to handle it.

Instead of putting them way ahead it cost them years of recovery and let AMD sneak up from behind.

Intel’s CEO, Brian Krzanich, was forced to resign in the middle of a major chip and manufacturing transition. [1]

That had to be a heavy blow because the politics in this company are ugly.

I feel like sometimes people look past the most obvious signs of why a company is struggling.

For example, Apple’s move to the mothership had a major impact on the company. It was one of the company’s biggest “product” releases. No one mentions this as a reason for anything.

[1] https://www.reuters.com/article/us-intel-ceo-idUSKBN1JH1VW

I agree with you, but keep in mind that those numbers represent the generation of the process node, rather than specific physical characteristics. Intel 10nm is roughly comparable to TSMC's 7nm.

The problem (in my non-insider view) is that Intel's 10nm just hasn't delivered. It was delayed substantially, and faced several problems even after rollout.

They still make more money than ever. It's mostly DC driven and many companies are probably looking at alternatives.

It was about time. The architecture has so much technical baggage. I don't even think it was intended to last that long.

x86 is only dead because CISC is dead. Now we just have to wait for Intel to go "full IBM" and try to ban people from writing emulators...

I don't think they can ban emulators without just straight up banning competitors like AMD that make compatible chips. (which presumably they would have done if they could)

They were forced to allow AMD early on to have a second source to sell to .gov, and since then AMD has gained enough patents that they need to cross licence.

So they could try to kamikaze AMD with the execs parachuting to some 3rd competitor at the last second?

Not really.

Amd64 is twenty years old, patents are expiring.

Reply on news.ycombinator.com