question for those in the know - intel and arm?

[tommo123]

assuming i have this right, the patent for x86 is over (or near enough) but is useless as other aspects of modern cpus (sse, x86-64 etc) are still patented or whatever.

could intel get an arm licence, throw everything they have at getting the best phone/tablet soc out there (non x86) and with their huge process tech advantage, totally outclass every other player? then, add extensions or whatever the correct term is to essentially lock out other players from the market? meaning everyone else could use arm, but not the custom designed parts intel would make?

if intel was willing to use arm, couldn't they totally dominate the phone/tablet market? i mean even if say in 2 years time they came out with a soc that was just equal to whatever qualcomm has at the time, their node advantage should mean that their socs would use far less power than the other players using TSMC/GF or whoever right? thereby having better battery life?

might even be enough to have apple running to intel to make socs for the iphone 6 or whatever.

or, is intel so adamant about sticking with x86 that it overrules every other consideration?

--

forgot to say, i wrote this in one go as it poured from my brain - and my brain is a messed up place - so sorry if it's a weird read

 

[Jawadali]

or, is intel so adamant about sticking with x86 that it overrules every other consideration?

I think this is the case. Intel used to make ARM CPUs (StrongARM, XScale) for PDAs and other devices. They sold this division to Marvell in 2006.

With how popular (and high-performing) ARM has become/is becoming, I am wondering how they feel about that decision now, but I guess they are pretty confident in their ability to scale down x86.

I don't know the details, but I'm sure others here will be able to shed some more light on this.

 

[386DX]

I doubt Intel will ever bother with building ARM CPU's. The profit margins is just to low for Intel's standard. In the recent breakdown of the iphone 4s, the material costs for Apple to build the phone was around $188 for the cheapest model. The most costly components were the display, wireless radios, and the NAND flash memory. The A5 CPU (one of the high end ARM CPU) had an estimated cost of $15. That's $15... not $15 profit for Samsung (the CPU maker) but $15 is what they charge for the CPU. If you consider it might take them $10 in materials and labour cost, then paying licensing fees to ARM, they are making a few dollars on each CPU. Those are not margins Intel has traditionally gone after when they can make tens, hundreds or even thousands of dollars on one CPU.

 

[grkM3]

you can bet intel is getting into the arm sceen,profits too low? they can make there own chips and sell 5 million chips just to the iphone 5 etc.

people will pay the premium also,you can bet anything if intel had a quad core arm running 3x faster than the comp and consuming 2x less power they will charge a premium for there chips.

the new droid razr is going on sale for 650 bucks,thats the most expensive cell ever sold here in the states and people will pay for it.

I can see them charging up to 80 for just the cpu if its got the performance and power consumtion to back it up.

they will also get a bunch of these chips out of a single wafer since they wont be that big in size.

just imagine the new 1.2 ghz arm thats going in the samsung prime if it were built on intels 22nm 3d process.It would consume alot less power and probably clock a lot higher aswell

 

[tweakboy]

This arm will never come to the consumer PC desktop. I mean Windows 8 at end of 2012... what arm cpu ? NOPE!!!,, regular quad core or 8 core cpu. gl

 

[cbn]

the new droid razr is going on sale for 650 bucks,thats the most expensive cell ever sold here in the states and people will pay for it.

Of that $650, how much do you think is SOC cost?

 

[grkM3]

Of that $650, how much do you think is SOC cost?

30 bucks? people would pay another 50 for a better intel cpu,I know I would if the power draw was better.

 

[cbn]

30 bucks? people would pay another 50 for a better intel cpu,I know I would if the power draw was better.

I don't know the price myself, but that number you are quoting sounds about right.

I do agree with you about the value added benefit of a $50 CPU.

That $50 CPU would result in a 67% higher price to the SOC maker, but it would increase the price of the handset only 3%. ($670/$650= 1.03)

That sounds like a great trade-off to me!

This got me wondering why ARM SOC chip makers don't do more "binning" (to compensate) for the Intel Process tech you are talking about?

I think this has to do with Smartphone SOCs being contained in a controlled market held by a very limited number of device makers and Wireless Venders. This very small number of device makers markets only a few handset models. Having special "Binned" SOCs is probably too niche for this kind of Mass Market...there might not be enough to go around! (unless the entire supply goes to a single very low volume handset model) Once we see the ARM Tablets/Tablet Laptops (not sold through Wireless Contracts) open up and become more prevalent I think we will see the emergence of more "binned" SOCs (with the appropriate mark-ups).

 

[386DX]

Of that $650, how much do you think is SOC cost?

According to the iSuppli teardown analysis of the iphone 4s the CPU was $15, its not gonna be much more for the Motorola Droid.

http://www.isuppli.com/Teardowns/Ne...8,-IHS-iSuppli-Teardown-Analysis-Reveals.aspx

People don't realize the costs of an ARM CPU isn't much in the overall price of a cellphone. Its nowhere near the price of the cheapest x86 CPU ($50-$60). The markup is on the entire phone not the individual parts. Chip makers like Qualcomm, etc are doing well in the mobile sector because they don't just provide the CPU but also the Radio's and other components. Its the reason why Apple profits are more then Samsung even though Samsung provides the CPU and other parts for the iphones.

 

[gevorg]

Intel will rather create a new architecture, than use ARM. Intel will milk x86 as much as they can. They might still get a piece of the mobile/tablet action by being at least one die shrink ahead of ARM (to make up for the power inefficiencies of x86), so stay tuned for mobile/tablet chips from Intel in 2012/13. Its so much easier to have just one competitor on the x86 arena.

 

[grkM3]

the droid has the OMAP 4460.dont know what that cost is but I think I saw 30 dollars on one link.Its built on 45nm and can be clocked to 1.5ghz

imagine if intel built that same chip for them at 22nm in there fabs,you would get 3x better power consumtion.

I would ask intel to build my chips and give them a small cut in the profits.apple would buy 10 million of them for the next iphone since they are having huge power issues with the 4s that just came out.

everyone would win from that,intel would make around 30 a pop and the cell makers would have a better setup for there phones at a small increase in price.

I could see intel selling 30 million chips to ALL the major cell makers

 

[Idontcare]

imagine if intel built that same chip for them at 22nm in there fabs,you would get 3x better power consumtion.

You might lower the power consumption of the ARM SOC itself by a factor of 3x but you aren't going to lower the power consumption of the LCD screen, nor the power consumption of the radio emitter either.

Remember Atom and and the Poulsbo chipset?

 

[grkM3]

You might lower the power consumption of the ARM SOC itself by a factor of 3x but you aren't going to lower the power consumption of the LCD screen, nor the power consumption of the radio emitter either.

Remember Atom and and the Poulsbo chipset?

yeah but the whole problem apple is in now with the 4s is from going to a dual core chip.They say its a gps bug but I doubt it will get fixed with an update,the battery is just too small in that cell.

The new droids are using a super amoled screen that drop power down from reg lcds that were used before.Not sure about apples retina display but there 4s is using the same screen with just a newer dual core chip.

knowing intel they wont build for someone else,they would rather compete heads up because how big the mobile market is.

If they can build a chip that gets 2x performance and 3x less power makers will choose intel over arm.

its all about battery life and performance now and thats whats driving all new cell phones.

When I bench my droid (overclocked) and even stock clocked the cell gets really hot and I can litterally watch the battery bar drain as its pushing the cpus.

right now a 22nm arm would be clutch in my celly.

 

[dealcorn]

Intel feels pretty good about the ability of its process technology and increasing transistor budgets (for gating) to address power consumption issues so that is no motive for them to do an ARM chip. The investment in x86 IP is big and much of the IP is world class. Considering the cpu workload, Intel finds the world coming to its strengths. As they are going against an entrenched competitor, Intel knows the old rule of thumb that you need double the performance or double the value to displace the entrenched guy. Intel likes its hand from a technical perspective and expresses confidence it can market an attractive chip. Conceptually, Intel's hand grows stronger as dimensions scale and Intel's experience grows. Every tick and every tock makes it stronger.

Already, you can see signs of ARM's hand falling apart. It is great that ARM64 is coming to servers but did you catch the 2 watt power requirement per core. The often denied statements that ARM will lose efficiency as it scales appear to be true. This chip will go against Atom 22nm. At 32nm the N2600 has 3.5 watt TDP with IGD. Xeon Atom should come in under 2 watts at 22nm. ARM is about the become the fat chick.

From a market perspective, Intel's potential customers generally will not support Intel unless it is able to directly connect with consumers. That's where Intel's challenge lies as it goes against an entrenched competitor.