[Q] ARM vs x86 in consumer space in 10 years

[MarkizSchnitzel]

If you had to place a bet on which will dominate the consumer market in, say, 10 years, what would you say?

I guess it's tied with Windows closely, and what will the fate of windows be.

I know it's all wild speculation, and I am a noob when it comes to all of this, but I am curios as to how will the computer, software and SOC landscape look like in the future.

Are there any relevant trends to look at to try and predict?

Will we all be rocking iphones and android phones with amazing performance that have continuum-like capabilities? Or will windows remain on x86 and people would still be buying laptops?

 

[Zodiark1593]

Consumer apps are perfectly happy on either ARM or x86. However, production and other heavy tasks doesn't seem to be moving away from x86 at all. Unless someone decides to go and cram 50 ARM cores on a workstation chip and fit it with PCI-e, we'll probably be keeping this split for awhile longer until the next big thing hits. What that big thing is, I don't know.

 

[cbn]

Unless someone decides to go and cram 50 ARM cores on a workstation chip and fit it with PCI-e, we'll probably be keeping this split for awhile longer until the next big thing hits. What that big thing is, I don't know.

That sounds about right (some kind of Workstation/NAS/Console/Virtualization host/server device with PCIe for expansion.....aka "the don't call it a desktop, desktop"), but I think there will less cores than 50. Maybe 8 good cores with SMT + iGPU.

 

[MarkizSchnitzel]

Consumer apps are perfectly happy on either ARM or x86.

But if Windows fails, and google figures how to make android work on more productive form factors, and Apple completely switches to ARM, how would game industry react?

Would it be a smooth transition without a drop in quality for AAA games, or would it be a rough transition?

However, production and other heavy tasks doesn't seem to be moving away from x86 at all.

Is that because back compatibility? Why is corporate market so hard to change?
I remember when I worked in this company for making furniture, the application they used for design/CNC looked like it has been made for windows 95 and than never ever updated. Price has obviously in the same time frame risen by a lot. Why is that?

 

[cbn]

But if Windows fails, and google figures how to make android work on more productive form factors, and Apple completely switches to ARM, how would game industry react?

Would it be a smooth transition without a drop in quality for AAA games, or would it be a rough transition?

I don't think Windows will fail, but I do believe some gaming could move to Linux.

Example: SteamOS (or Steam Client for Ubuntu or Linux Mint)

P.S. CDProjekt (owners of Good old Games and makers of the Witcher series already offer games in Linux, and their Galaxy client will be coming soon to Linux). I think if we did see another development house move over to Linux it would probably be EA/DICE with the Battlefield series as the first title (reason: BF4 has Mantle as a renderer, so Vulkan and compatibility with Linux seems logical as future next step)

 

[Zodiark1593]

But if Windows fails, and google figures how to make android work on more productive form factors, and Apple completely switches to ARM, how would game industry react?

Would it be a smooth transition without a drop in quality for AAA games, or would it be a rough transition?

There is a severe lack of control on the software side of things and the massive scale that will make such a transition brutal, or at least, extremely slow.

The software ecosystem plays a major role here. One of the biggest strengths of the PC as it is today is it's ability to run applications from decades ago without emulation. To be rid of x86 altogether kills off that advantage. As Apple had learned, you need a very strong ecosystem already in place on the new architecture to start with for people to actually jump aboard as not everyone wants to wait for the software to be developed when already existing solutions exist on the old arch.

For ARM to take over the desktop PC, a comparable software ecosystem must be in place for early adopters to jump aboard, otherwise they will simply turn to the platform that already has the software that meets their needs. However, at the same time no one wants to develop expensive production apps for an ARM desktop/workstation if the user base is currently non-existent due to the risk involved. Quite a formidable (not insurmountable) challenge. Apple can probably do this relatively soon as it holds significant influence over such companies, and can probably even fund this development themselves. They just need a fast enough chip. There is no such unity in the Wintel space.

This doesn't count the fact that as of now, we have yet to see an ARM cpu scaled and designed to take on an x86 workstation cpu. We do not even know how the architecture will scale at 4+ GHz clock speeds, or if they can cram as many cores as needed to match their x86 competition and still keep power use down. There is too much we don't know here to say ARM will replace x86.

 

[NTMBK]

Personally I reckon the DEC Alpha will make a comeback.

EDIT: Or possibly MIPS.

 

[Thanatosis]

ARM is clearly dominant in mobile and that's where the consumer space is headed. x86 still totally dominates server and datacenter markets, but with Qualcomm launching 24 core server SoCs with similar i/o capabilities I think we will see a shift to ARM in that space too.

If intel sorts out its 14nm issues and somehow makes a comeback at 10nm (which is looking less and less likely every year) there is a possibility that they will retain there dominant position in the server space. The problem is most news is pointing to TSMC launching 10nm in 2H2016/1H 2017 while intel will still be launching Kabylake and giving the same core counts and same performance as they did before, plus 2.5-5% maybe. With the huge jumps in performance we see in A9X and Exynos 7420/8890 there is reason to believe that Qualcomm or another ARM vendor could continue to improve performance at that cadence, which mathematically makes it impossible for intel to keep up. Intel isn't a company that can pivot quickly and start making something different so they will likely just continue paying off vendors and using x86 lock in to suck revenue out of their server and datacenter customer. The only people who still buy x86 for consumer crap today are people who either need extreme performance to play games or people who are just too cheap to buy Apple. That doesn't look good for the future.

 

[ShintaiDK]

In 10 years I think it will be exactly like today more or less.

The ecosystems also prevent any movement. Else we wouldn't use ARM or x86 today.

 

[Phynaz]

ARM already dominates the consumer space. I have far, far more devices with an ARM CPU in them than anything x86, and I'm willing to bet most people do.

 

[cbn]

ARM already dominates the consumer space. I have far, far more devices with an ARM CPU in them than anything x86, and I'm willing to bet most people do.

Yes, but beyond that area you mention (mobile) how would they attack?

I think they would probably go for the corporate data center first (ie, VDI). Then scale back some lesser core count non-mobile SoCs for the consumer space.

They could also scale up to non-mobile consumer from NAS first (bypassing corporate data center), but die sizes being what they are I am thinkng this probably makes less sense.

 

[Arachnotronic]

ARM already dominates the consumer space. I have far, far more devices with an ARM CPU in them than anything x86, and I'm willing to bet most people do.

This.

 

[jihe]

Personally I reckon the DEC Alpha will make a comeback.

EDIT: Or possibly MIPS.

what about itanium?

 

[Phynaz]

Yes, but beyond that area you mention (mobile) how would they attack?

I think they would probably go for the corporate data center first (ie, VDI). Then scale back some lesser core count non-mobile SoCs for the consumer space.

They could also scale up to non-mobile consumer from NAS first (bypassing corporate data center), but die sizes being what they are I am thinkng this probably makes less sense.

I didn't say mobile. My television has an ARM CPU in it. Heck, my refrigerator does. My car has about a dozen.

 

[cbn]

ARM is clearly dominant in mobile and that's where the consumer space is headed. x86 still totally dominates server and datacenter markets, but with Qualcomm launching 24 core server SoCs with similar i/o capabilities I think we will see a shift to ARM in that space too.

If intel sorts out its 14nm issues and somehow makes a comeback at 10nm (which is looking less and less likely every year) there is a possibility that they will retain there dominant position in the server space. The problem is most news is pointing to TSMC launching 10nm in 2H2016/1H 2017 while intel will still be launching Kabylake and giving the same core counts and same performance as they did before, plus 2.5-5% maybe. With the huge jumps in performance we see in A9X and Exynos 7420/8890 there is reason to believe that Qualcomm or another ARM vendor could continue to improve performance at that cadence, which mathematically makes it impossible for intel to keep up.

With Qualcomm having a pretty tight relationship with Windows (via their phone) I have to wonder if we eventually see some emulation software (from MS) allowing x86 apps to run on ARM Servers.

If so, I would have to imagine this emulation would hurt software performance (and who knows whatever else) to some degree , but if Qualcomm is able to size their cores properly maybe they eventually come out favorably enough that it is competitive?

So ARM loses on software performance, but wins on performance per watt?......but overall, this balances the situation out for VDI in some ways.

Then (if this happens) what could we expect Intel hypothetically to do as a response? Maybe develop a middle size x86 core (perhaps 3 wide) that yields better performance per watt on high core count/large die size Xeons than the 4 wide core they are using today?

Then after this a battle going back and forth between x86 and ARM for dialed in performance per watt/software performance on these high core count/high die size servers.

 

[VirtualLarry]

Then (if this happens) what could we expect Intel hypothetically to do as a response? Maybe develop a middle size x86 core (perhaps 3 wide) that yields better performance per watt on high core count/large die size Xeons than the 4 wide core they are using today?

With the "race to idle" principle, I don't know if that design choice would be a win. Then again, I don't even know if "race to idle" even applies to server CPUs.

 

[Zodiark1593]

what about itanium?

I believe Itanium was meant to replace x86 at one point, though that ambition never materialized. An 800 MHz Itanium chip was apparently capable of emulating a x86 Pentium of an effective 100 MHz speed. Keeping in mind that at the same time frame, 1.1 GHz Pentium chips were available.

They guys working on Xbox emulation have run into some pretty difficult issues concerning x86 emulation as well, and this is with modern Haswell chips running a Coppermine emulator.

Considering Intel is the one in the best position to make an x86 emulator, unless Intel fully adopts ARM themselves, it's a safe bet to assume absolutely no compatibility with x86 apps, or at least, anything speed critical.

 

[cbn]

With the "race to idle" principle, I don't know if that design choice would be a win. Then again, I don't even know if "race to idle" even applies to server CPUs.

One reason I brought that up is because (for example) on the E5 2699 v3 Xeon (Haswell 18C/36T, 145W TDP) the base clock is only 2.3 Ghz. This, of course, is for a 22nm processor

Successive nodes will increase the xtor budget and thus the number of possible cores for any given die size, but the typical power reduction isn't going to be enough to keep up with the potential increase in cores. For example, a doubling of xtors (and a doubling of CPU cores) requires a 50% power reduction on the node to keep clocks the same....but this won't happen as process tech is at best giving us something like 30% power reduction.

Thus Intel has to make compromises (either provide a smaller increase in the amount of cores than what was actually possible or lower the clocks on a higher core count increase) to keep TDP in check.

And even with conservative core count increases, I am still thinking the clocks might even be lower than the E5 2699 v3. If this happens race to idle isn't going to help as much as it could because the clock really isn't that high to begin with.

So I am wondering if instead of a potential 2Ghz 4 wide core (or whatever the future brings), Intel sees a 3Ghz 3 wide cores as a better compromise. If true, then I am thinking race to idle would also work better and the performance per watt at load might be better as well. This assuming the uarch (pipeline, width, etc) is optimized for the narrow work range intended.

 

[5150Joker]

If the next generation consoles make the logical leap to an SoC that uses an ARM processor then that could pave the way for a Windows version that runs ARM. If that happens, we could see a lot more players in the desktop APU business like NVIDIA, Qualcomm etc. Quite frankly, if Steam OS evolves + Vulkan and developers throw their weight behind it, I don't see why a lot of gamers wouldn't switch to an ARM based desktop in 5-10 years. Anything to break from the Intel/AMD duopoly for processors would be nice and I'd like to see NVIDIA in the ring with some powerful SoCs made for the desktop. As for the concern of software ecosystem, that could be mitigated via online apps accessed through the browser and maybe even x86 emulation. I think gamers would be a good catalyst to target for the transition away from x86 to ARM.

 

[teejee]

There were 10 billion ARM enabled SOC's/CPU's sold 2013. Intel sells less than 500 million CPU's per year. So ARM camp is more than 20 times bigger today (volume wise).
My guess is that ARM will continue to dominate over Intel also in 10 years.

 

[Zodiark1593]

There were 10 billion ARM enabled SOC's/CPU's sold 2013. Intel sells less than 500 million CPU's per year. So ARM camp is more than 20 times bigger today (volume wise).
My guess is that ARM will continue to dominate over Intel also in 10 years.

Do realize we are talking about high-performance desktop PCs, and not cheap mobile, nor microcontroller parts.

Aka, your appliance ARM microcontroller has no bearing here.

 

[teejee]

Do realize we are talking about high-performance desktop PCs, and not cheap mobile, nor microcontroller parts.

Aka, your appliance ARM microcontroller has no bearing here.

well the open post only talks about consumer space, not your definition

 

[Thala]

If the next generation consoles make the logical leap to an SoC that uses an ARM processor then that could pave the way for a Windows version that runs ARM. If that happens, we could see a lot more players in the desktop APU business like NVIDIA, Qualcomm etc. Quite frankly, if Steam OS evolves + Vulkan and developers throw their weight behind it, I don't see why a lot of gamers wouldn't switch to an ARM based desktop in 5-10 years

Microsoft had the opportunity with Windows RT to extend the Windows ecosystem to ARM. However they decided to:
1) Artificially lock-out desktop app execution unless properly signed
2) discontinue Windows for ARM (at least from product perspective, internally they certainly still are compiling Windows for ARM)

In particular 1) was a stupid move, as you just had to recompile Win32 Apps to run under Windows RT. CLR/.Net apps were even running as is. Microsoft had a full featured Windows for ARM but stumbled over their own stupid decisions. Would not surprise me if 10 years down most would be running desktop apps under ARM, but it would not be Windows because Microsoft missed the boat and locked itself into x86 territory.

 

[NTMBK]

If the next generation consoles make the logical leap to an SoC that uses an ARM processor then that could pave the way for a Windows version that runs ARM. If that happens, we could see a lot more players in the desktop APU business like NVIDIA, Qualcomm etc. Quite frankly, if Steam OS evolves + Vulkan and developers throw their weight behind it, I don't see why a lot of gamers wouldn't switch to an ARM based desktop in 5-10 years. Anything to break from the Intel/AMD duopoly for processors would be nice and I'd like to see NVIDIA in the ring with some powerful SoCs made for the desktop. As for the concern of software ecosystem, that could be mitigated via online apps accessed through the browser and maybe even x86 emulation. I think gamers would be a good catalyst to target for the transition away from x86 to ARM.

We had MIPS consoles and PowerPC consoles, didn't change a thing in PC. Picture this- there is an ARM port of SteamOS, but it offers 1/10th of the games of the x86 version due to lack of compatibility. Which do you buy?

 

[ShintaiDK]

Microsoft had the opportunity with Windows RT to extend the Windows ecosystem to ARM. However they decided to:
1) Artificially lock-out desktop app execution unless properly signed
2) discontinue Windows for ARM (at least from product perspective, internally they certainly still are compiling Windows for ARM)

In particular 1) was a stupid move, as you just had to recompile Win32 Apps to run under Windows RT. CLR/.Net apps were even running as is. Microsoft had a full featured Windows for ARM but stumbled over their own stupid decisions. Would not surprise me if 10 years down most would be running desktop apps under ARM, but it would not be Windows because Microsoft missed the boat and locked itself into x86 territory.

Windows RT never had a chance. No matter what MS would do or not.

Its all about the ecosystem.