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

[VirtualLarry]

They were foolish for not licensing out x86 to more companies and in the end that's going to be their downfall.

I agree. What good is exclusive profits on x86, when the rest of the world has moved on to ARM?

 

[MarkizSchnitzel]

Is it possible that Intel is already internally working with other architectures? Sounds like it would be prudent, but I don't know the insider thinking. F.e., if that were to become publicly known, maybe it would have negative effect on their stock?

 

[jhu]

Yes but back then we didn't have a world filled with ARM tablets and smartphones. If you can produce a product that ports between console/tablet/desktop easily, that should be appealing to developers and gamers. Look at the current console designs, they're specialized PC parts and console ports are available on the PC at the same time now with very little changes needed. I don't see why that couldn't happen with ARM designs in the future. It's pretty evident that ARM is the future in the consumer market and x86's days are numbered. I don't see Intel overcoming this challenge despite how much money it has. In fact, it wouldn't surprise me to see Intel producing ARM CPUs in house at some point in the future.

Uhm... Batman: Arkham Knight perhaps? Great on console. Complete crap on PC.

 

[5150Joker]

FYI, the Samsung profit number is operating income not net profit; the comparable number for Intel here is $4.2 billion.

Much of Samsung's profit comes from the sale of memory and mobile devices; very little of that profit comes from selling logic semiconductors.

Where the Samsung Electronics division derives their profits isn't the point. It is that a division of this chaebol makes more money than Intel in a quarter, let alone the entire Samsung conglomerate which is far bigger than Intel could ever hope to be.

Also, NVIDIA is basically out of the race against Intel in its core MPU business (Tegra is 100% automotive now)

Someone should pass that memo on to NVIDIA because I don't think they know that Tegra is only for automotive: http://www.nvidia.com/object/tegra.html 🙄

Qualcomm is laying people off because it needs to bring its expenses into check (since the profitability of its mobile chip business has plunged into the abyss), and AMD is on its deathbed with people getting excited by a measly $370M cash infusion from selling off yet more assets.

AMD is definitely on the fast track to bankruptcy but that could change with Zen, K12 and their GPU refresh if they didn't screw up. But they are a minor player, Intel really needs to be worried about Samsung as they just picked up Jim Keller and it seems they are making the move to high performance SoCs, especially since they also control their entire supply chain like Intel. I wouldn't be surprised to see Samsung branch out from smartphones and tablets to notebooks in the future. NVIDIA also has a GPU upperhand and if ARM does successfully transition to notebooks and eventually desktops, they will be able to put up a fight against Intel, especially since getting to a smaller node is becoming more expensive and difficult and that's what Intel relies on.

Intel is in a very strong financial position and it can deliver these kinds of profits while at the same time investing at a multiple to what these other companies do in semiconductor logic-related R&D.

Nobody said it's weak financially, just that it's not the 800 lb gorilla it once was. Intel is faced with a declining desktop/notebook business, no growth in mobile and increasing competition in the server space.

Intel isn't invincible but the notion that it's just going to roll over and die because a lot of people are trying to compete with it doesn't seem that well thought out IMO.

It won't roll over but it's clear that it's facing a lot of pressure and it's going to get worse in the future.

You haven't really followed this industry for long, have you? Back in the heyday of the PC (when PCs were booming like phones are now), everybody had an x86 license: TI, IDT, Cyrix, VIA, AMD, IBM, etc.

One by one these players exited the market because it just wasn't economically viable; there's only so much x86 MPU revenue to go around and if you don't capture a large enough part of it, it makes no sense to continue to try to invest to build products.

Yeah and how many of them did Intel sue to oblivion or use underhanded tactics to force them out?

FWIW, it seems very counterintuitive that "licensing x86 to more companies" thereby increasing competition and diluting the value of the x86 compatibility advantage that Intel has would be anything but a bad thing.

Intel could have licensed it's atom design to counter ARM but it would rather keep that to itself for high margins. Problem is they're the underdog now and they still have that mentality of "margins margins margins". Tell me, how will Intel overcome all this competition from ARM vendors as convergence takes place between tablet/phone/desktop with ARM being the dominant processor? Also, it's a only a matter of time before Apple switches it's notebooks (or future convergence device) over to their in-house ARM SoC and that's going to be a large hit for Intel's bottom line.

 

[VirtualLarry]

Intel could have licensed it's atom design to counter ARM but it would rather keep that to itself for high margins. Problem is they're the underdog now and they still have that mentality of "margins margins margins". Tell me, how will Intel overcome all this competition from ARM vendors as convergence takes place between tablet/phone/desktop with ARM being the dominant processor? Also, it's a matter of time before Apple switches it's notebooks over to their in-house ARM SoC and that's going to be a large hit for Intel's bottom line.

This. Mobile is where it's at, and Intel is the underdog, and not getting any better at it. Cherry Trail was a big disappointment. Where is Intel going to go for profits, once their x86 ecosystem is "worthless"? Sure, maybe they'll retreat to becoming a server-only company. But then again, how long can that last? Will they end up like IBM? It's really starting to look like that.

Edit: It's not a PC-centric world out there any more. Some staunch Intel defenders that are PC enthusiasts forget that. Intel seems like it's slowly but surely losing its grip on the market, except for Server.

 

[Nothingness]

Is it possible that Intel is already internally working with other architectures? Sounds like it would be prudent, but I don't know the insider thinking. F.e., if that were to become publicly known, maybe it would have negative effect on their stock?

Intel has had ARM licenses for many years and though they likely don't have an ARMv7/8-A architecture license, they probably inherited an M or R architecture license from Infineon. They are also used to fab ARM IP (as an example Altera latest 14nm chips use ARM Cortex-A53).

There are two steps beyond that:

1. fab a high-profile high-volume application SoC based on ARM
2. design their own ARM CPU.

I think 2 is unlikely (at least certainly not now), but 1 is still a possibility.

 

[cbn]

Intel could have licensed it's atom design to counter ARM

Well there is the Intel-Rockchip agreement:

http://www.anandtech.com/show/8061/...gic-agreement-with-chinese-soc-maker-rockchip

In return, Rockchip becomes the first company to be able to more or less license Intel x86 CPU IP for use in this combined, Intel-branded SoC.

And a collaboration (involving investment) with Spreadtrum as well:

http://newsroom.intel.com/community...nt-and-adoption-of-intel-based-mobile-devices

Under the terms of the agreement, Spreadtrum Communications, Inc. will jointly create and sell a family of Intel Architecture-based system-on-chips (SoCs). Initial products will be available beginning in the second half of next year and will be Intel Architecture-based SoCs sold by both companies.

"The adoption of Intel's architecture technology will enable us to accelerate the development of mobile SoCs that expand the breadth of our portfolio, benefiting handset makers addressing both China and the global market," said Dr. Leo Li, Chairman and CEO of Spreadtrum. "We are pleased to embark on collaboration with Intel around these new product offerings."

 

[knutinh]

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?

I think that Microsoft/Apple/Google are concentrating on higher-level languages/libraries, and that the bulk of "Apps" (as seen by consumers at large) have transitioned from projects where much of the complexity lies in low-level things (i.e. printer drivers, extended memory management), to projects where the complexity lies in network behaviour ("cloud" etc). This means that the local client can run some software abstraction that allows a pretty and smooth-rendered GUI enabled by relatively high-performance hardware.

The key here is that e.g. Apple wants developers to not only rely on their libraries and avoid low-level code, they even want their hands on source-code/byte-code that allows _them_ to recompile software on the fly whenever needed. This should make it a lot easier/faster to simply swap hw at some point in the future, than it used to be in the "good old days".

While ARM certainly have been successful at attacking Intel from their soft side (low performance, low energy SOCs with low prices/flexible licensing), it remains to be seen how much of an advantage their model/instruction sets have if used for workstation-class cpus. Will they hit the exact same limitations as Intel? Will the (typically somewhat) separated architecture/manufacture model used for ARM SOCs be better due to economic reasons or worse due to lack of integrated development compared to Intel? Why are not the promised server-class many-core ARM products materializing? These ought to be the poster-boys on why ARM is inherently more efficient at amount-of-work per watt, per dollar and per m^2 than Intel for tasks that are easily split into many parts.

-k

 

[5150Joker]

There is the Intel-Rockchip agreement:

http://www.anandtech.com/show/8061/...gic-agreement-with-chinese-soc-maker-rockchip



And Spreadtrum:

http://newsroom.intel.com/community...nt-and-adoption-of-intel-based-mobile-devices

Interesting, now I wonder how much Intel charges them and if the products will be price competitive with ARM? Could be too little too late.

 

[Arachnotronic]

I agree. What good is exclusive profits on x86, when the rest of the world has moved on to ARM?

Yeah, what good is exclusive profits on x86 indeed. It's only the most profitable semiconductor franchise on the planet. Intel's server chip group makes more in operating profit than the top two merchant mobile AP vendors take in. :biggrin:

 

[Nothingness]

Yeah, what good is exclusive profits on x86 indeed. It's only the most profitable semiconductor franchise on the planet. Intel's server chip group makes more in operating profit than the top two merchant mobile AP vendors take in. :biggrin:

Indeed, and when something is that profitable, it remains so forever :biggrin:

 

[Arachnotronic]

Indeed, and when something is that profitable, it remains so forever :biggrin:

When something is that profitable, it gives the company profiting a lot of room to invest in future products/technologies 😉

 

[ShintaiDK]

People only see ARM as a growth segment still with its prime money coming from Smartphones. But the smartphone segment is very likely to contract next year. And then ARM companies is in the same position with shrinking consumer revenue.

I think people forget the economic situation. US for example so far this year got deflation. Eurozone close to as well. QE programs got no effect. There is no growth. The only inflation there is, is stock prices and housing prices. Making sure company sale volumes will keep declining.

Smartphone sales is already declining in China. And the ASP for smartphones is going down fast.

 

[5150Joker]

When something is that profitable, it gives the company profiting a lot of room to invest in future products/technologies 😉

Yeah and that's working out really well for Intel in the mobile space. :biggrin: :whiste:

People only see ARM as a growth segment still with its prime money coming from Smartphones. But the smartphone segment is very likely to contract next year. And then ARM companies is in the same position with shrinking consumer revenue.

I think people forget the economic situation. US for example so far this year got deflation. Eurozone close to as well. QE programs got no effect. There is no growth. The only inflation there is, is stock prices and housing prices. Making sure company sale volumes will keep declining.

Smartphone sales is already declining in China. And the ASP for smartphones is going down fast.

People see ARM as a growth segment because it is. Global shifts in the economy will keep happening, that's a cyclical trend that has been going on since the dawn of commerce. It doesn't take a PhD in business to recognize that Intel is surrounded and it's current monopoly is in a market that is dwindling rapidly and becoming obsolete thanks to convergence devices. On the business side, it is facing increasing pressure and won't be able to sustain the type of margins it enjoys currently for the foreseeable future unless it drastically changes course. But even if it does change course, I don't see those high margins it loves so much coming back once ARM chips away at the volume low/mid range end of notebooks/desktops.

I was reading an article where Apple could theoretically spend $25 on it's own custom ARM SoC for macbook where as it pays Intel around $150 per chip. Companies like Apple are all about squeezing every dollar they can and it's obvious that's where they will be heading. The same goes for Samsung, they're pouring a lot of resources into Exynos and now with Jim Keller on board, it is evident they want to move into the desktop/server SoC space since that is his specialty. Samsung is coming out swinging with it's total control of the supply chain and unlimited funds--Intel should be very worried as they can't pull their usual shady tactics with a giant like Samsung as they did with AMD and other smaller companies.

Read over that Harvard link I put up about Intel, you realize what a scummy company it is and I can't wait to see them put in their place one day.

 

[ShintaiDK]

ARM will shrink just as much as x86. The reason x86 began is that its a higher segment.

Look at China, negative smartphone growth ALL YEAR.

That 25$(more likely much more) Apple chip is around the same as a 20$ Intel chip. Now spot the missing part.

 

[5150Joker]

ARM will shrink just as much as x86. The reason x86 began is that its a higher segment.

Look at China, negative smartphone growth ALL YEAR.

That 25$(more likely much more) Apple chip is around the same as a 20$ Intel chip. Now spot the missing part.

China's market will eventually stabilize and no longer grow at the rates it has, that's a given as they switch to a consumer driven economy and become saturated. But smartphone growth will just switch to another emerging market like India, Brazil and later on perhaps Pakistan:

Where does Intel fit into any of this? Doesn't seem like it does but ARM will keep growing everywhere globally.

 

[NTMBK]

Where does Intel fit into any of this? Doesn't seem like it does but ARM will keep growing everywhere globally.

By selling modems and collaborating with LG, Rockchip and Spreadtrum.

 

[Nothingness]

When something is that profitable, it gives the company profiting a lot of room to invest in future products/technologies 😉

So basically you're saying that once a company has grown enough to generate enough money, it won't fall down or fail? You know it's not how things work 😉

As an example, and as others pointed out, Intel have proven that despite their profitability and huge investments in the smartphone market, they could fail.

 

[simboss]

People only see ARM as a growth segment still with its prime money coming from Smartphones. But the smartphone segment is very likely to contract next year. And then ARM companies is in the same position with shrinking consumer revenue.

Looking at ARM's results: http://www.design-reuse.com/news/38504/arm-q3-2015-financial-results.html it does not seem that obvious: Cortex-A only accounts for 17% of the unit shipment, obviously it has a much higher ASP, but it is still a balanced product portfolio, there are many places that are growing faster than the smartphone market is slowing down.
The ASP for ARM is probably increasing as well with the move to ARMv8 and more complex SoCs as well.


To be more on topic, maybe the question you should ask is: what will the consumer devices will look like in 5-10 years?
Microsoft is doing Continuum, and you are getting more and more web based and streaming based devices (20M chromecast sold, that is not insignificant), so there is a good chance the "thin client" model will finally be used.
Basically the client and the network are now (or will be in 5 years) good enough to provide the user experience people expect, and anything that requires more computing power can be done somewhere else more efficiently.

Will these "thin clients" be laptops, tablets, smartphones, TVs, HDMI dongles, something else... probably a mixture of everything, but the need for big CPUs in these machine is decreasing and will continue.

The only exception to that is gaming, but from the reviews it seems NVIDIA Grid is not too far off already, I can easily imagine that in 5 years this will be a solved issue as well.

I guess the big moment will come with the next gen of consoles:
If they stay on x86, there will be no reason to push all the SW to ARM as there will always be a significant gap in the offering, hence no incentive for anyone to move to ARM.
If they move to ARM, a big chunk of the most demanding SW will have moved to ARM already, getting the few major professional apps (Photoshop and friends) moved over is not such a big hurdle, especially because it would also open the tablet/smartphone to them.

Whether or not you would loose a bit of performance by doing it is not that important, for the users that are performance limited, a cloud based solution powered big a BIG cpu will almost always win.

If you go one step further, you can now have CPUs tailored for hosting specific apps with accelerators tightly coupled (FPGA or ASICs) which will give a massive win, something that is not possible with a consumer grade solution.

Once you get to that point, being x86 or ARM is mostly irrelevant, it will be slightly easier with x86 but you will have to recompile to use these accelerators anyway.

Maybe this will not happen because it is too complex, because consumers are always going to stick to the same "install my apps and run locally" model, but this seems to be where the industry is heading, Intel included (they would not have paid that much for Altera if they only wanted to fill their fabs).

 

[Nothingness]

The ASP for ARM is probably increasing as well with the move to ARMv8 and more complex SoCs as well.

Yes the ASP is increasing: royalty revenues were up 37% YoY while the number of units "only" increased by 20% in mobile and enterprise, and 30% in microcontrollers.

 

[NTMBK]

Looking at ARM's results: http://www.design-reuse.com/news/38504/arm-q3-2015-financial-results.html it does not seem that obvious: Cortex-A only accounts for 17% of the unit shipment, obviously it has a much higher ASP, but it is still a balanced product portfolio, there are many places that are growing faster than the smartphone market is slowing down.
The ASP for ARM is probably increasing as well with the move to ARMv8 and more complex SoCs as well.


To be more on topic, maybe the question you should ask is: what will the consumer devices will look like in 5-10 years?
Microsoft is doing Continuum, and you are getting more and more web based and streaming based devices (20M chromecast sold, that is not insignificant), so there is a good chance the "thin client" model will finally be used.
Basically the client and the network are now (or will be in 5 years) good enough to provide the user experience people expect, and anything that requires more computing power can be done somewhere else more efficiently.

Will these "thin clients" be laptops, tablets, smartphones, TVs, HDMI dongles, something else... probably a mixture of everything, but the need for big CPUs in these machine is decreasing and will continue.

The only exception to that is gaming, but from the reviews it seems NVIDIA Grid is not too far off already, I can easily imagine that in 5 years this will be a solved issue as well.

I guess the big moment will come with the next gen of consoles:
If they stay on x86, there will be no reason to push all the SW to ARM as there will always be a significant gap in the offering, hence no incentive for anyone to move to ARM.
If they move to ARM, a big chunk of the most demanding SW will have moved to ARM already, getting the few major professional apps (Photoshop and friends) moved over is not such a big hurdle, especially because it would also open the tablet/smartphone to them.

Whether or not you would loose a bit of performance by doing it is not that important, for the users that are performance limited, a cloud based solution powered big a BIG cpu will almost always win.

If you go one step further, you can now have CPUs tailored for hosting specific apps with accelerators tightly coupled (FPGA or ASICs) which will give a massive win, something that is not possible with a consumer grade solution.

Once you get to that point, being x86 or ARM is mostly irrelevant, it will be slightly easier with x86 but you will have to recompile to use these accelerators anyway.

Maybe this will not happen because it is too complex, because consumers are always going to stick to the same "install my apps and run locally" model, but this seems to be where the industry is heading, Intel included (they would not have paid that much for Altera if they only wanted to fill their fabs).

FPGAs are much more expensive and difficult to develop for than traditional software. They're a tool that makes a great deal of sense in certain niches, they're not a general purpose device which is going to obsolete high performance CPUs.

As for interactive apps moving to thin client for everyone... I am highly skeptical. Not everyone has a good enough internet connection, and lives close enough to a major hub to get good latency. All it takes is a couple of random latency spikes at a critical moment getting you killed, and your experience is ruined.

 

[simboss]

FPGAs are much more expensive and difficult to develop for than traditional software. They're a tool that makes a great deal of sense in certain niches, they're not a general purpose device which is going to obsolete high performance CPUs.

You are using present tense, this is not the topic!
Of course 5 years to have a big shift like this is short, but things are clearly moving towards more dedicated logic, mostly because the general purpose CPU rate of improvement has been really low compared to what it used to be.
The point is also not to include FPGA in every CPU on earth, only to the server ones where you need them.
This is a niche indeed, but I am sure all the big SW providers of the planet would be happy to get a 10x boost and a way to lock the customers to their cloud platforms rather than selling a one-off license.

You could separate SKUs by number of gates available in the FPGA, then have proper libraries that do the acceleration job for you. Yes it is complex, but CPU design teams have been struggling for a while to get things moving, maybe these resources would be better spent on this area?

Maybe everyone will settle for the "good enough" and business will continue as usual, or there will be another disruptive technology, but so far FGPA and/or HW accelerators look like the most likely candidate to bring something new to the computing table.

As for interactive apps moving to thin client for everyone... I am highly skeptical. Not everyone has a good enough internet connection, and lives close enough to a major hub to get good latency. All it takes is a couple of random latency spikes at a critical moment getting you killed, and your experience is ruined.

Same thing, we are talking of the future. In particular the bandwidth is currently good enough to drive a 1080p@30fps in many cases, the focus of ISP should be to reduce the latency and improve reliability.
This is the the main barrier but does not seem bigger than the drive for bandwidth which has killed(*) the CD, then the DVD...
For now I don't see a huge push for that, ISPs still advertise mostly the DL speed, but I will be surprised if this does not change soon, mostly because the rationale to sell 100Mbps bandwidth is actually pretty bad, who would be streaming 2 (or even 3) 4K videos regularly, whereas latency is an easy sell to gamers first, then to the general public.

(*) of course there are still people using CDs, DVDs, for many reasons, but I think everyone agrees this is not what the future will look like.

 

[imported_ats]

As for interactive apps moving to thin client for everyone... I am highly skeptical. Not everyone has a good enough internet connection, and lives close enough to a major hub to get good latency. All it takes is a couple of random latency spikes at a critical moment getting you killed, and your experience is ruined.

Even for those with a good connection, latency will basically kill this. Thin clients certainly have their niche, but they've never been able to expand beyond it and historically their niche has shrunk over time. The problem with thin clients is that they generally require significantly more compute muscle on the back end to work than just having a thick client to begin with.

And all the game via stream junk has basically failed for the same reason. Latency. That and it is generally MORE expensive to support a user via streaming than locally. In order to support a remote game stream, you need lower frame latency on the server than you would locally so you can try to claw back some of the latency differential.

 

[imported_ats]

You are using present tense, this is not the topic!

People have been trying to make FPGAs easier to program for decades with little to show for it. Very very low confidence that's going to change.

Same thing, we are talking of the future. In particular the bandwidth is currently good enough to drive a 1080p@30fps in many cases, the focus of ISP should be to reduce the latency and improve reliability.
This is the the main barrier but does not seem bigger than the drive for bandwidth which has killed(*) the CD, then the DVD...
For now I don't see a huge push for that, ISPs still advertise mostly the DL speed, but I will be surprised if this does not change soon, mostly because the rationale to sell 100Mbps bandwidth is actually pretty bad, who would be streaming 2 (or even 3) 4K videos regularly, whereas latency is an easy sell to gamers first, then to the general public.

You don't need the bandwidth to stream video, that's mostly a latency insensitive market and the bandwidth to do it has been there for years. But that doesn't do a damn thing for interactive use. Even doing interactive use over local networking has obvious degradation and once you add in geographic differences and switches/routers, it only gets worse. Hell, people notice a difference between game servers with 50ms and 80ms pings.

As far as the rational to sell 100Mbps or 1Gbps or 10Gbps, people aren't buying those connections for the constant data rate, they are buying them for the burst data rate. Less waiting to DL 50+GB is a significant driver. I know a lot of people with 1Gbps connections, and they wouldn't go back to something slower if you paid them. Hell I know 1 guy who basically keeps his Steam library in a ramdisk because he can just delete and redownload games so fast it doesn't matter how much storage he has for it. He's going to be upgrading to a 10Gbps within the next month, fyi.

 

[NTMBK]

You are using present tense, this is not the topic!

It's a fundamental fact of FPGA development. You're basically developing hardware, not software. It's not going to change any time soon.

I have no doubt that the hyperscale customers like Facebook, Google, and so on will have the volume to justify FPGA development. But it's only at those sort of scales where the economics make sense.

Same thing, we are talking of the future. In particular the bandwidth is currently good enough to drive a 1080p@30fps in many cases, the focus of ISP should be to reduce the latency and improve reliability.
This is the the main barrier but does not seem bigger than the drive for bandwidth which has killed(*) the CD, then the DVD...
For now I don't see a huge push for that, ISPs still advertise mostly the DL speed, but I will be surprised if this does not change soon, mostly because the rationale to sell 100Mbps bandwidth is actually pretty bad, who would be streaming 2 (or even 3) 4K videos regularly, whereas latency is an easy sell to gamers first, then to the general public.

(*) of course there are still people using CDs, DVDs, for many reasons, but I think everyone agrees this is not what the future will look like.

Bandwidth != latency! Netflix streaming is extremely latency insensitive- buffer a few seconds of stream, and you can smooth out all the fluctuations in bandwidth and latency spikes. With interactive gaming you have to feed in real time, so you can't tolerate that kind of inconsistency. You would need to re-architect the entire network infrastructure of the country around it.