Hacker News Reader

Comment by outside1234 | Hacker News Reader

x86 is dead - first in consumer, then in cloud.

It is hard to see to me how this ends any other way. The creative class (us) will quickly have largely all ARM computers within 4 years.

Its not hard to see from there how software will be even more optimized for ARM variants than x86 and that the scale of both mobile and consumer computing will push x86 out of the datacenter slowly as old software that relies on x86 is retired over the next decade.

People won't want to develop on x86 and deploy to ARM. ARM is more power efficient which is important in the data center too. We already scale by the core in the cloud, so why not just heap a few more cheap cores on if we need more to match x86 (which right now looks like we might not).

Tell me how I am wrong.

View on HN

Are there examples of Arm designs OTHER than the M1 which are suitable for consumers? Yes, the M1 is a remarkable product and it will certainly make inroads against x86 on the desktop but it is from a single company. Will Apple M1 (2, 3 etc) replace all x86 devices in a decade? That’s hard to swallow. Now, if we see another Arm product released that ALSO kicks x86 butt, from another player, then maybe I’d believe a change is happening

if you are not using an apple phone, your smartphone likely uses an ARM-licensed chip, and chromebooks/windows RT devices ran ARM cores

the only thing holding back wider adoption is legacy x86 software

The smartphone ARM and chromebook ARM processors are nowhere near the M1 or x86 CPU in terms of performance. You need an M1 kind performance in these devices outside of Apple.

Also Raspberry Pi’s are ARM.

ARM Macs run x86 software great (my experience) and ARM Windows machines run x86 software adequately (I have read). Don’t know how well QEmu works on Linux - haven’t tried it but know it’s available.

Apple's Rosetta 2 that allows x86 code to run on M1 is a different beast than QEmu though. It translates x86 machine code to ARM code. That's why it is much faster than VM solutions.

QEmu also translates x86 machine code to ARM code, or vice-versa, or many other combinations. It's more optimised for portability and ease of adding translation support for new architectures than performance though but it's a really nifty bit of engineering.

Gaming is still x86 territory and that includes 2 of the 3 main competing gaming consoles isn't? Hard to think of a PS5 as legacy stuff.

Is not the majority of gaming occurring on mobile ARM devices, not dedicated consoles or rigs?

“Mobile gaming has fast become the largest gaming market in the world with industry revenue expected to hit $76.7 Billion by the end of 2020 and 2.2 million mobile gamers worldwide. It’s become so popular that 72.3% of mobile users in the US are also mobile phone gamers. To put this into perspective when compared with the wider video game market, by 2022 the global game market is set to reach $196 Billion, and the mobile gaming market will account for $95.4 Billion of that alone.”

https://www.wepc.com/statistics/mobile-gaming/

>Is not the majority of gaming occurring on mobile ARM devices, not dedicated consoles or rigs?

Yes, the "I'm taking a 5 minute toilet break" gaming or the "I'm sitting in a bus and I have nothing else to do" gaming.

It's making money off of whales who like that there is f2p plankton to gobble up.

I played a bunch of mobile games in 2014 and they were all trash or cash grabs. Often with grinds worse than MMORPGs.

That's only because no one other than Apple has put up a serious competition in ARM CPU space.

Earlier consoles used a boatload of different CPU and GPU architectures (PowerPC, SuperH, IBM Cell, other proprietary stuff, x86), nowadays consoles have converged to essentially a locked down, glorified x86 PC on one side (Xbox One, Xbox S/X, PS4, PS5) and ruggedized ARM-based tablets (Nintendo Switch) on the other side.

On the mobile gaming space, ARM already has achieved utter dominance with Switch + iPhone + Android... all it needs for ARM to conquer the console market is for someone to tape out an extremely powerful chip and actually sell it - Apple won't.

Is the dominance of x86 due to any inherent properties of x86, or historical legacy?

I think the xbox-s/x and ps5's choice of CPU vendor was driven more by business factors and the integrated graphics, than the virtues of x86 per se. The Switch does pretty well despite the less powerful CPU in the tegra.

Qualcomm is also entering the laptop segment with its 8cx platform [1]. While not a serious contender to M1 yet, it is comparable to an Intel i5 [2].

Samsung is also planning to release an Exynos laptop [3].

[1] https://www.qualcomm.com/products/snapdragon-8cx-gen-2-5g-co...

[2] https://www.digitaltrends.com/computing/snapdragon-8cx-vs-co...

[3] https://www.engadget.com/samsung-exynos-amd-2200-windows-10-...

So what? Microsoft SQ1 has been there for a while [1]. Without a software support they are just a fancy stones that can't run desktop software.

1: https://en.wikipedia.org/wiki/Surface_Pro_X

This is a very late reply, but Windows 10 has its own x64 on ARM emulator [1]. It's not production ready as Rosetta but we'll see how it goes.

https://blogs.windows.com/windows-insider/2020/12/10/introdu...

Qualcomm are terrible, though; rather than doing an M1 and competing on performance, they rely on cornering the mobile market with a patent portfolio.

They also have (fortunately) zero control over the downstream software, so they can't do Apple-style vertically integrated improvements.

Qualcomm tried so many times, it is actually a strong argument against ARM dominance, if Apple is the only one who can make it work but there are two companies that can make x86 work then the answer is pretty obvious which architecture is better.

Lots of software have been ported to ARM architecture in the last half year, so the clock started ticking for great ARM based Windows laptops. But it's clear that ARM's memory ordering has won.

Creative class is a lot more than programmers. Look at boutique system integrators like Puget Systems... do you see Apple/ARM Hardware? Heck no you don't! These guys sell hardware to a LOT of the companies that comprise the credits in movie production, game production, laboratories, university science departments, ML researchers, big energy, and soooooo much more.

These guys are not rewriting their software stack for thin chassis, thermal limited, and non-repairable hardware. Raytheon doesn't buy Apple/ARM hardware for their simulations, design and development. Does Boeing, Airbus, Ford or Caterpillar run on Apple/ARM hardware? Are these companies just chomping at the bit to ditch their legacy stacks? I don't think so.

The parent comment didn’t say “we’ll all be using Apple”

He said ARM.

I edited my comment as I think in the context of the larger discussion, ARM and Apple are basically interchangeable.

I wholeheartedly disagree. Apple is probably the "first, best" case of ARM, but there's absolutely nothing stopping anyone from making a similar investment in ARM hardware. Indeed, Amazon is already doing so with Graviton, and we're seeing similar improvements in both raw speed and in perf/watt like what we're seeing with M1 chips vs. Intel[0]. And that's one of the best parts of ARM -- the design itself is available for licensing, unlike x86, so anyone can pony up the cash and build up their own customized chips that maintain compatibility with base ARM code (maybe needing a recompile to take advantage of new hardware features).

[0] https://dev.to/dnx/arm-vs-intel-a-real-world-comparison-usin...

ARM, not Apple. They are not the same thing.

Right now they pretty much are.

> Creative class is a lot more than programmers.

I'm not sure anyone thinks programmers when they think creative class. That's like saying the professional athlete class is a lot more than just Pokemon Go players...

Yes that's very much the more artistic side aka graphic designers, it smacks of doing an easy "olgy" degree or even a "PPE".

The sort who work in traditional professional jobs and who shudder at the thought of the "boffins" and greasy "engineers" .

Recall yesterdays conversation about European pay rates for developers.

Isn't creative class just people who make things without consuming raw materials? I think it includes consultants, "knowledge workers", etc. in its original formulation, not just artists.

Raytheon etc don't buy them yet.

But the M1 is already a huge hit in the music and design community, which is pro-performance oriented. If subsequent Mnx chips provide enough of a speed bump they're going to destroy the high end Xeon market.

You've named two industries where Apple have histrionically been really popular, mostly because of their software rather than for any particular performance reason. They'd all buy the latest, Apple product whatever was inside because that's the only one Apple is going to support going forward. The idea of real high-performance computing moving to ARM is entirely dependent on there being suitable good software support, which currently does not exist (at least in my industry there is nothing yet and the vendors move very slowly). This "destruction" might eventually come but it won't be any time soon and Intel will of course react in the meantime.

I agree with you in general, but disagree on the time scale. Particularly:

>The creative class (us) will quickly have largely all ARM computers within 4 years.

I think good options might exist in 4 years, but still possess trade-offs, where legacy x86-64 platforms still have their place. This could be something as niche as maturity of IOMMU implementations in x86 workstations, but stack up a bunch of niches and I find it hard to see "largely all" in four years. Fifty-fifty in four years, "largely all" in ten for "the creative class". Then there is a long tail of industries may end up legacy-bound to x86 for decades to come.

Nothing is guaranteed. If starting in 2022 we were to see another 3-year performance plateau, ala 2016-2019, then anyone who bought into an mid- or high-end x86 platform today may not feel compelled to upgrade in just four years. Even if the ARM options exist, it might not be a compelling enough gap in four years' time. I'm not saying that a plateau is likely (competitive landscape suggests the opposite), but I don't have a crystal ball, so I won't discount it either.

Again I want to emphasize that I mostly agree, but I wouldn't bet any money on sweeping changes on such an aggressive timeline.

You overestimate the size and relevance of your peer group and you underestimate the inertia that will impede this migration.

"creative class" :-)

The m1 is cool but for heavy duty work its still got a number of issues memory multiple displays.

Rember there no such thing as to much Memory, IPC, Graphics performance - thee is sadly often a luck of budget.

> People won't want to develop on x86 and deploy to ARM.

It's still too early to say if people want to develop on ARM to deploy on x86.

The 90's with all their "better than x86" chips couldn't beat the fact that people want to develop on the same architecture to which they deploy. Back then that meant displacing all these other chips from the server side because there was no option but x86 for us mortals to have as PCs.

This dearth of alternatives on the server side took out their respective offerings on the high-end/niche workstation market as well. The PowerPC was one of these casualties, lets not forget about that.

Things are even worse today. The x86 is dominant in both ends and has no real alternatives: ARM server chips are weak and the inertia of building on x86 and deploying on x86 is too strong, and ARM desktop chips are also weak, except for a single luxury brand (Apple, which incidentally cares very little for any developers besides the ones that develop for their own ecosystem).

> ARM server chips are weak

For AWS cloud customers, I don't think this is true anymore. Graviton2 is quite capable with better performance in some instances for less money. I've already started moving some Java services to Graviton2 for the cost savings.

AWS is positioning the the Graviton2 in such a way that everyone using AWS will end up on them if they can.

Only if I can buy a desktop/laptop with a powerful ARM processor and plug-in a bootable usb and install operating system (arm version) of my choice with all drivers etc working fine.

I expect this will be the case with Linux on the M1 in a year or two, given the pace of marcan's work. Windows depends on Microsoft's willingness, of course.

That's the thing, reverse engineered drivers are not the solution and I don't see Apple providing drivers or any documentation of M1 SoC for other operating system. We need M1 equivalent from vendors that are NOT into "vertical integration" like Apple.

That's an argument against the M1 and ARM dominance.

Ok, it's time for the emulation anxiety meme (based on EV range anxiety), will this system emulate this extremely performance intensive x86 game at ultra with 4k graphics at 144hz? If not, then I will not buy it.

>Tell me how I am wrong.

It's already difficult enough to get Windows games to run on Linux, getting them to run on ARM with good performance isn't going to happen, especially when every ARM vendor insists on using an integrated GPU.

If your CI system builds for x86 and ARM from the first build, it will be simple to ship to both. For old games it’d be harder, but one could maybe compile from one ASM to another.

Intel and AMD aren't standing still in this.

Apple also isn’t just standing around, enjoying the M1

> as old software that relies on x86 is retired over the next decade.

seems very, very, very unlikely to me

a lot of people still rely on windows. maybe if microsoft gets a better emulator going, but I imagine with the amount of legacy they support that's a lot more challenging than for apple who are much more trigger happy when it comes to imposing changes on devs.

i think it'll definitely happen, especially with the web browser increasingly taking on the role of the operating system, but 4 years seems a little optimistic, even with your qualification of that statement

waiting for the gaming console announcement

Uh, Nintendo Switch?

Something a bit more modern and powerful, perhaps?

If you mean specifically an Apple Silicon game console, that's the Apple TV.

Reply on news.ycombinator.com