Opinion: Power imperative favors ARM's client-to-server play

9/10/2010 9:34 AM EDT

Configurable core ... this has echoes of Tensilica! That leaves one to wonder if ARM has licensed some technology from Tensilica for the A15 ... :)

9/10/2010 9:47 AM EDT

Configurable in the sense that: it can have one, two, four or many cores; it can have small single-level caches...or larger caches including L2; it can have multiple high-speed parallel buses or not....

I don't mean configurable in terms of instruction set additions or omissions.

So I doubt Tensilica, or Synopsys (ARC) licensing is required...unless you know better?

9/10/2010 8:04 PM EDT

If a single core A15 can deliver five-times the processing power of an A8 or A9, we could be looking at a significant change in the low-end computing landscape. The current generation ARM is powerful enough to power the iPad and I've run a minimalist Linux on an A8 powered Beagleboard.

Jump the processing capability up as is expected and we have a viable engine for the next generation of Netbooks. Especially with ARM style low power and low heat.

The biggest problem keeping ARM out of the consumer computer arena might just be the OS, and not the capabilities of the chip. Linux may be close, but it still has far too many rough edges to be considered a viable main-stream OS.

9/11/2010 5:28 AM EDT

Since Intel has had several iterations of RISC architecture (Itanium, Alpha/StrongARM (ex-DEC), ARM (ex-Level 3) and finally decided to dispose the entire product set to Marvell, do you think the A15 is going to get the real difficult job done, i.e. running x86 MFlops/Drhystone and I/O intensive, enterprise critical systems for corporate clients.

In other words, how much do you think A15 could dent Intel's future, as the stock is languishing badly these days because of the reverse paranoia among investors from the ARMed threat?

Many thanks,
alpine

9/12/2010 3:56 PM EDT

Thoughtful article, Peter. I believe ARM is moving into Intel's territory in a pragmatic way, and I think they will get there eventually.
I would like to know your opinion about whether you think ARM should make its own chips, cut the middle-man, and enjoy high profit margins as a result? or should it stick to its licensing model?

9/12/2010 10:47 PM EDT

This announcement signals ARM is moving from being a mobile processor company to the Processor company and could herald a sea change in the domains traditionally dominated by Intel and AMD. Now if I am not wrong they have cores from the very low end deeply embedded micro-controllers to the high performance stand alone computing units. Interesting to see how big brother will react to this news. With virtualization they are hinting at where they want to be down the road.
This could open a whole new space for start ups wanting to do some non-intel enterprise stuff.

9/13/2010 9:58 AM EDT

KB...

ARM should not make its own chips. It should not be a fabless chip company either.

ARM was set up with an intellectual property business model in 1990 partly because, as Sir Robin Saxby tells it, no bankers or backers would put up the money for anything more.

As it was ARM was started with backing from Acorn, Apple and VLSI Technology Inc.

By being modest in the licensing fee and royalty demands ARM has made it has been non-threatening to just about all of the semiconductor IDMs and foundries. And over 20 years has been adopted by nearly all of them. As such ARM can leverage off all of the investments the Samsungs, TSMCs and GlobalFoundries make in R&D and capital expenditure for manufacturing.

Why would ARM turns its back on those partners now?

If it did it would stimulate the rise of a rival processor architecture to replace its own with those abandoned chip makers.

Secondly many of the systems and service companies -- the Apples, Nokias and Vodaphones -- really appreciate that ARM is a humble and independent supplier of architectural intellectual property. They don't like the way Intel has run the PC sector and they would be less inclined to buy if ARM was a similar chip company.

Thirdly, ARM would have to go head-to-head in a capex battle with Intel, and I think we know who would win that.

9/13/2010 12:18 PM EDT

KB....

ARM is a relatively small company but I don't think it is at risk of collapse in the near future. It is consistently and enviably profitable, does not have the massive capital requirements that a leading-edge digital IC manufacturer needs -- which are of the order of billions-of-dollars every couple of years -- and does not suffer from the the boom-bust cycles of those further up the supply chain.

Basically only about a dozen companies, maybe 20, can afford to stay in leading-edge digital manufacturing IC. Starting with Intel, Samsung and TSMC the list peters out not long after.

The last reason given for ARM to make its own chips is that ARM is so small as to be susceptible to a take-over. I think there is something in that weakness, but entering manufacturing would not make it more resilient against take-over, but rather would make it more susceptible.

Perhaps the ARM ecosystem could provide a defense with the likes of Apple, Samsung, Texas Instruments, STMicroelectronics clubbing together to buy stock on the hint of an unwelcome take-over attempt on ARM. All have an interest in keeping ARM independent.

On the other hand they would get fed up of buying ARM stock expensive and selling it cheap when the danger had passed.

9/16/2010 11:36 AM EDT

@KB3001

Well I guess ARM could go fabless and try selling direct, but I don't think they would know how to produce what the OEMs want.

Remember the ARM cores are a fraction of the die area on most chips that ARM ships in. ARM has some but not ALL of the rest of the stuff that goes on those chips.

It would certainly disrupt the semiconductor partners who would almost certainly turn to another architecture as quickly as they could.

It is those semiconductor partners that are talking most intensely to the OEMs and integrating intellectual property from multiple sources to create the SOCs.

Apple doesn't want an ARM processor. It wants an applications processor, or a baseband processor with a lot of bells and whistles that the likes of Broadcom and Qualcomm know how to provide. Or indeed, that Apple may start designing themselves.

Similarly in microcontrollers chips for particular applications. It is the semiconductor partners that know the applications. They just decided to off-load the processor architecture to a third-party so they could focus on the stuff nearer to their customer. Like getting the chip working, the software running and all the I/O.

So KB, I think most people would say if it ain't broke don't fix it.

9/16/2010 11:49 AM EDT

I'm with Peter on the question of ARM making chips with or without their own fab. From my perspective. ARM is a great example of a company finding a core competence and sticking with it. By doing so, they've become one of the more influential semiconductor companies and probably someplace in the top four influential processor companies.

Some companies in some markets need to diversify or broaden to stay healthy and competitive, but not all do. For some, doing so is the start of a death spiral. I'm so tired of great companies with great products deciding that they need to be all things to everybody and within five years, we've lost them to the resultant financial and mismanagement mess.

ARM may never be the next Intel, but ARM has a very important and very beneficial role to play in the processor market. It's a for profit model that carries some of the advantages of a commercial endeavor as well as some of the best advantages that would come with an open source business model. They charge money and they deliver a great product to a large number of companies/end users on an equal basis. It's a refreshing clear and concise purpose.