The ARM takeover begins

[Keyrock]

The potential of ARM chips replacing x86 has long been theorized and speculated upon, but I believe we're on the verge of it happening now. Qualcomm's upcoming 8cx (The Verge link) is spearheading ARM's move into the laptop segment. Initially, it's going to be ultra thin, low power laptops, but eventually ARM will invade more powerful laptop sectors and even desktops. ARM is already making a lot of headway in the server market and that trend will likely only continue.

 

The biggest hurdle to ARM getting into traditionally x86 dominated markets has long been native software support. If you look at benchmarks over the last few years of software running on x86 and ARM chips, the x86 often shows a massive performance advantage, but that is largely down to the ARM chip running an emulated version. When you compare a x86 chip and comparable ARM chip both running native versions *poof* the performance advantage largely (completely in some cases) disappears. Apple already has native ARM versions of most of its software, as they have begun their switch to ARM and now Microsoft is letting developers create native 64-bit ARM code for Windows apps as well as working on ARM versions of Microsoft Office. Adobe Photoshop and a host of other major productivity software makers are in the process of making native ARM versions of their programs or have already done so, and once the ball gets rolling others will follow, and quickly (they'll have no choice in the matter).

 

So what's driving this move toward ARM? There are 2 major factors at play: power consumption and profit margins. It's no secret that ARM chips run at a much lower power envelope than x86, it's largely why ARM dominates the mobile market. Famously, when testing the first ARM chip made in the 80s for power draw, they forgot to hook up the chip to the power supply AND IT STILL RAN simply off the electricity from the signal connection. Intel has spent years trying their hardest to push x86 chips into lower power envelopes in a desperate attempt to push into the lucrative mobile market and have largely been unsuccessful. The problem is the architecture itself and the bloated AF instruction set that forces x86 chips into extremely high amounts of transistor switching to achieve greater performance, which in turn requires a lot of power, plus x86 just doesn't scale nearly as well into higher core counts as ARM does (AMD has taken steps to allow their own x86 chips to scale better, in that regard they are way ahead of Intel, though still far behind ARM manufacturers). 

 

The other, and likely more important, factor is that ARM chips are easier and cheaper to manufacture, largely because ARM is a simpler chip design than x86. This is why ARM chips regularly beat x86 to every node shrink. Every OEM loves components that are cheaper to manufacture because it allows them to either pass the savings onto the customer or (more likely) increase profits for themselves.This is why so many companies are pushing ARM adoption, it means greater profits for them.

 

The days may be numbered for x86.

Edited February 10, 2019 by Keyrock

[injurai]

Yeah, windows world encourages per-compiled binaries, so software offerers will have to have a toolchain ready to offer new binaries. Problem is a lot of people rely on old software that hasn't been updated in years. I'm sure qualcomm will have an easier time with a younger generation, but there would still be that awkward division within the Windows platform.

 

While the days might be numbered for x86, I'm pretty confident that Intel and AMD can pivot into a new market and would probably have a lot more resources to draw upon.

[Keyrock]

AMD had already started working on ARM processors, though the AMD K12 that was initially slated for 2017 has yet to surface. I think AMD is still working on ARM chips but I'm not sure if the K12 will ever actually release or something else will release in its stead.

 

The change of x86 dominated markets to ARM won't happen overnight, but I'm very confident it will happen eventually. The profit margins with ARM chips are sufficiently higher than that of x86 to drive this forward. As for older software, that will be a growing pain for a while, but eventually I think we'll reach a point where emulated versions of those older programs that may never get native ARM versions will perform well enough to be used in the interim before they get replaced with newer native software.

Edited February 10, 2019 by Keyrock

[injurai]

I'd personally like to see AMD throw their support behind RISC-V.

[injurai]

I'll also say I wouldn't surprised if AMD and Intel salvage their better microcodes to form some new RISC architecture, essentially just tearing out the whole CISC layer and trimming the remaining features that existed to support that layer. I'd be nice to standardize around something, but I could also see a fight against ARM. Intel and AMD might even standardize around their own RISC architecture to fend off mobile players in the markets that they've traditionally owned. I can see Intel trying to go proprietary, but if AMD moves on RISC-V that might be enough to get Intel to circle that foundation as well. Of course they could then develop their own proprietary extensions, but hopefully they wouldn't have to. As RISC-V is supposed to be working on optional extensions like SIMD.

[Zoraptor]

AMD had already started working on ARM processors..

 

They already make at least one though it isn't an 'independent' chip but a coprocessor, it's their 'security processor'/ mandated NSA backdoor chip. Intel also made ARM chips via their purchase of the DEC part of the joint venture ARM/ DEC StrongARM chip (and, being Intel, some legal strongarming too), and StrongARM was a genuine potential competitor on desktop (up to 300MHz in 1997, and you could daisychain up to 8 (?) processors via Hydra, though for some reason that feature got killed off...)

 

StrongARM is a bit indicative. ARM on desktop is like mainstream VR gaming, game streaming services and the Year of the Linux Desktop; it's always just a few years off. My dad actually went to university with one of ARMs founders and got a tour around their facility in Cambridge in 1992, and we had a Acorn RISC PC with a StongARM in it- indeed, it's still in my shed.

 

No doubt nVidia will be very keen to get ARM into laptops though as a way to even the playing field vis-a-vis Intel and AMD; though their Tegras really aren't competitive as well as being hamstrung by, well, Jensen being Jensen. The Switch jailbreak has been described as a security flaw but it's actually a feature, documented by nVidia.

[AwesomeOcelot]

ARM isn't competitive in anything other than ultra low power devices. Intel is doing just fine in the Ultrabook segment. We will see what AMD brings with its Ryzen 3 APU.

 

A change from the dominant x86 architecture is bound to happen at some point, nothing lasts forever.

[Keyrock]

I'm not going to take Qualcomm's performance claims at face value, I never trust benchmarks from the manufacturer itself, but if the Snapdragon 8cx is anywhere close to what Qualcomm claims it is then it could really shake up the ultrabook market.

[Malcador]

https://www.realworldtech.com/forum/?threadid=183440&curpostid=183486

 

Linus comments on ARM servers,  figure it might be of some interest.