ARM v Intel Debate In One Act

By: BoredJuvenileTechnophile (noone.delete@this.nowehere.com), November 1, 2012 10:04 pm
Room: Moderated Discussions

Setting: Coupa Cafe, Palo Alto, over a couple Lattes...


ARM: Intel is entering its period of terminal decline. The whole PC industry has passed its time; PC unit volumes and ASPs will deteriorate over the next few years. The mobile revolution will succeed and surpass it, and the slowing but inexorable advance of Moore's law means most users can no longer clearly discern differences from one generation of PC to the next. CPU price deflation will ensue and consign the x86 monopoly to the ash heap of history. Intel has no real mobile play and has a track record of not only failing to fundamentally diversify its product portfolio, but in particular has tried to “do” mobile on multiple occasions and failed in each instance.

In addition, ARM partners are beginning to assail Intel's strongholds, delivering more than sufficient performance coupled with an insanely great power advantage and attractive price points. Low end client systems along with cloud servers will increasingly fall into the ARM camp – a trend to be turbocharged once Apple ejects Intel from its product lines. In the following years, ARM will begin to move up the food chain and devour the remnants of a weakened Intel's revenue streams.

Moore’s law – the principle that gave rise to Intel, the drives Intel, that was defined by Intel – will extinguish Intel.

Intel: Intel will decimate the ARM ecosystem. The PC is stronger than you think and the tablet and smartphone are not the end state of computing. Might I remind you that twenty-five years ago Intel bet the company on the microprocessor and not only survived against long odds but crushed the competition. Motorola limped away, IBM retreated to their soft-services domain, and AMD was kept alive as a foil to fend off regulatory scrutiny.

We will push the technology so hard and so fast that the world order you seek to conquer will be upended before you can seize it. Even now – even after announcing that our fabs are at 50% capacity – we are not only continuing with the construction of two new fabs, but just disclosed plans to build an additional two – massively expanding our already underutilized infrastructure from a current base of eight to twelve 300mm fabs. You should ask yourself: for what purpose? How will we fill our factories?

We are now focused on mobile with the precision of a silicon laser. When mobile products move to our latest node, our in-house architects, designers, and developers will co-optimize our platforms to bring performance, power, and cost advantages that yours cannot possibly match. To the best chips go the spoils – and to the rest, well they trod the well-worn paths of Transmeta, Via, PowerPC, and – since it’s now convenient for us – AMD.

We will be ready. You have no idea what is coming.

ARM: I’d say Anand Chandrasekhar has a pretty good idea of what is coming.

Your business model is eroding, your ASPs are falling – you are staring structural decline in its face. You cannot sustain your brute force R&D investments much longer. Our mobile ecosystem is growing at multiples of your rate. Need I remind you Apple’s Q4 profit will be much larger than your ANNUAL take? To the extent that fab tech truly matters, Apple, Samsung, and others (Qualcomm? TSMC?) will invest the resources necessary to take your crown.

And let’s look at history – let’s look at what Vin Dahm did with the Athlon and far inferior resources.

But even more importantly: SoC performance and power are only one piece of the puzzle. Even allowing for a magical world in which your fab technology is perpetually ahead of ours, ours will be good enough: no end user will desire what you think you can offer (Ultrabooks?) over what they perceive to be fashionable – or over what has a superior UI; or simply over what they are used to.

Apple has proven that culture trumps bleeding edge technology in the minds of today’s consumer -- and thus your technology is not the advantage you perceive it to be. You cannot overcome our entrenched market lead or the associated psychological switching costs: our readily licensed architecture, our mindshare with device makers, and our partners’ mindshare with consumers are more potent than you realize.

We and our partners understand our customers – or, better yet, we understand people. One only has to subject oneself to your “We are Intel” ads from a few years ago to see your sophistication on that subject.

Intel: Perhaps you have mistaken colloquy for calumny. At Intel we understand the rules, chief of which is Moore’s law – the full implications of which you do not appear to apprehend.

No matter how much your device makers bankroll their foundries, and even given a virtual IDM model, the necessity for developing a one-size-fits-all process shuttle and then working with every customer to adapt their needs to it, materially impedes ARM’s ability to be competitive with high yields at the leading edge of technology. Our design rules may be restrictive, but they mean we are a law unto ourselves.

As for consumer tastes, your race to deploy FinFETs proves that SoC performance and power are still very relevant. Awaking your partners to the near term FinFET advantage will perhaps be Anand’s most dangerous legacy. But two or three years from now when your foundries are struggling up the yield curve on their first high volume non-planar modules, we will have mature yields on a faster, lower power process with twice the density that will be spun on largely depreciated tools. Physics is physics; economics is economics. Consumer psychology didn’t prevent the ascendancy of the PC in the face of the almost cult-like Mac following.

And we are ahead not just where it counts, but when it counts.

You underestimate how our present process lead shapes our future strategic mobility to push into new channels and our ability to tunnel through many of the economic barriers you think lie in our path. The industry is approaching an inflection point, an almost atomically abrupt junction in the progression of Moore’s law. One thing your process teams can validate for you is that moving at the pace of Moore’s law is challenging enough and is becoming increasingly more so; closing a multi-year gap in the near term to nullify our advantage is untenable for you or your partners.

Timing is everything. It determines your performance, it determines your costs, it determines your customers, it determines your power. Moore’s law is the ultimate time pacing rule for business, and Intel is the ultimate time machine – delivering a future beyond the reach of your arms.

ARM: You need to flush that product pipeline because that's a long latency miss if I ever saw one!

Time is on our side. You cannot Bohr a hole through your present impasse: this low cost, consumer electronics oriented, increasingly vertically integrated industry will never return to a purely horizontal model where any one supplier will be able to squeeze monopolistic rents from the rest of the value chain. You do not have control of the mobile distribution channels – and we have the key to opening your architectural lock on the computing industry.

Intel as we know it cannot survive in the realm of $20 ASPs … and to the very extent you push Moore’s law you drive us to the new reality more quickly. You are playing chicken with your own demise. There is no 125% Solution, there is no Operation CRUSH that can come to your rescue this time. Take your time, downshift your virtual factory, enjoy the good life while you have it – we look forward to hiring your best designers in about five years.

Intel:. Twenty years ago IBM decided to forgo the Pentium, planning to outflank Intel with their own proprietary architecture. The nascent broader PC industry was a small hodgepodge of players. Their components were not standardized and being divided they fought for scraps at the table of Big Blue – which at the time was the world’s largest tech company. They had R&D funding that we could not possibly match. So our biggest customer was about to lock us out of the market on which we had bet the company. We were about to lose our primary distribution channel and conceivably the entire PC microprocessor business. Yet today we have 80% of that business and more than 100% of its profits. What happened?

We pushed the process technology. We pushed the Pentium ramp – it ramped twice as fast as the 486 or any product ever had before. We formed Intel Architecture Labs, creating PCI and the other peripheral components that drove the standardization of the entire hardware side of the industry. We entered the motherboard business. We built the Dupont systems facility and began to manufacture entire PCs and sold them below cost to Dell, Gateway, Packard-Bell and others – complete with brand new, shiny Pentiums. In the end, leading edge technology offered at low cost won the day. IBM capitulated – they had to.

There is $50 or more worth of silicon content in every decent smartphone. When we put it all in one 3-D IC with a process that is both faster and lower power than yours and that has twice your density, our margins will not only persist – they could well grow. Without the middleman foundry our cost structure is the leaner.

Apple, Samsung – so what if they deny us the opportunity to distribute our product. Even if we lose our special relationship with Google, you forget that the carriers themselves are the ideal sharp end of the stick.

ARM: Is Intel going to contract with Hon Hai to produce 1 billion white label hand-sets per year? What relevance does ancient history have today? Twenty years ago you couldn’t make a properly functioning FPU! Where will your margins go? See my previous comment.

Intel: Where will Apple’s margins go! This is déjà vu; as it did with the PC industry, the horizontal model will triumph due to its cost advantages – and our insurmountable technological lead will be the hook that drives the industry to that model.

ARM: We’ve just discussed this. Even were we to agree with your premise, we are in an age when good enough technology can win the race. I have one word for Otellini: MICROMA.

Intel: Watch it ARM! In one respect, though, you are correct to bring up Microma – time will tell how right we are. Your misguided focus on "good enough" – your overstretched ARMs – are the embrace of death to the continuing rapid advancement of modern electronics. Everyone who strives to win paces themselves in all things. We have run this race with certainty for four decades. Consistent, reliable, expected exponential progress is a law that we will not let fade away.

ARM: No amount of pomposity can save you from racing to your own demise.

Intel: Pomposity or clear thinking? We have the experience to know the difference. We have put four new fabs down on this. We can do no otherwise!
 Next Post in Thread >
TopicPosted ByDate
ARM v Intel Debate In One ActBoredJuvenileTechnophile11/01/12 10:04 PM
  About time somebody wrote this :) NT (NT)Rohit11/01/12 10:12 PM
    Hah. This is an early Christmas gift from PaulChuck11/02/12 11:36 PM
      Hah. This is an early Christmas gift from PaulWiseGuy11/03/12 10:31 AM
      Ecosystem flexibility is overstated. Vendors want all-in-one solutionsanon11/03/12 11:00 AM
        Disagre. Evidence: Apple and Samsung (NT)anon11/03/12 04:39 PM
          Disagree. Evidence: Apple and SamsungWiseGuy11/03/12 07:45 PM
            Disagree. Evidence: Apple and Samsunganon11/04/12 05:15 AM
              Disagree. Evidence: Apple and SamsungWiseGuy11/04/12 07:34 PM
                Disagree. Evidence: Apple and Samsunganon11/04/12 10:37 PM
      Hah. This is an early Christmas gift from PaulRob Thorpe11/04/12 08:17 AM
  About time somebody wrote this :) (NT)Rohit11/01/12 10:13 PM
  ARM v Intel Debate In One ActCharles11/02/12 05:59 AM
    ARM v Intel Debate In One ActRohit11/02/12 08:24 AM
      Intel can/could pursue both ARM and x86 at the same timeMark Roulo11/02/12 10:01 AM
        Intel can/could pursue both ARM and x86 at the same timeLinus Torvalds11/02/12 02:19 PM
          Intel can/could pursue both ARM and x86 at the same timeRicardo B11/02/12 04:13 PM
            Intel can/could pursue both ARM and x86 at the same timeFoo_11/03/12 12:39 PM
          Intel can/could pursue both ARM and x86 at the same timeMichael S11/03/12 11:20 AM
          Intel can/could pursue both ARM and x86 at the same timeAlberto11/03/12 04:31 PM
            Intel can/could pursue both ARM and x86 at the same timeanon11/03/12 05:09 PM
              Intel can/could pursue both ARM and x86 at the same timeAlberto11/04/12 02:18 AM
                Intel can/could pursue both ARM and x86 at the same timenone11/04/12 02:25 AM
                  Intel can/could pursue both ARM and x86 at the same timeMichael S11/04/12 04:02 AM
                    Intel can/could pursue both ARM and x86 at the same timeAlberto11/04/12 04:17 AM
                      Intel can/could pursue both ARM and x86 at the same timeMichael S11/04/12 05:56 AM
                        Intel can/could pursue both ARM and x86 at the same timeAlberto11/04/12 06:26 AM
                    Intel can/could pursue both ARM and x86 at the same timeWilco11/04/12 06:25 AM
                      Intel can/could pursue both ARM and x86 at the same timeMichael S11/04/12 07:15 AM
                        Intel can/could pursue both ARM and x86 at the same timeWilco11/04/12 07:46 AM
                          Intel can/could pursue both ARM and x86 at the same timeAlberto11/04/12 09:28 AM
                            Intel can/could pursue both ARM and x86 at the same timenone11/04/12 09:43 AM
                            Clovertrail tablet from Dell - ridiculously expensiveMichael S11/04/12 09:43 AM
                              Clovertrail tablet from Dell - ridiculously expensiveRob Thorpe11/06/12 04:23 AM
                                Clovertrail tablet from Dell - ridiculously expensiveMichael S11/06/12 05:10 AM
                  Intel can/could pursue both ARM and x86 at the same timeAlberto11/04/12 04:11 AM
                    Intel can/could pursue both ARM and x86 at the same timeWilco11/04/12 06:45 AM
                Intel can/could pursue both ARM and x86 at the same timeanon11/04/12 05:20 AM
                  Intel can/could pursue both ARM and x86 at the same timeAlberto11/04/12 06:39 AM
                    Intel can/could pursue both ARM and x86 at the same timeWilco11/04/12 06:53 AM
                      Intel can/could pursue both ARM and x86 at the same timeAlberto11/04/12 09:10 AM
            Intel can/could pursue both ARM and x86 at the same timeWilco11/04/12 06:17 AM
        Everybody forgot XScale? (NT)EduardoS11/02/12 04:26 PM
          Intel aquired XScale ... not quite the same thingMark Roulo11/02/12 04:48 PM
        Intel can/could pursue both ARM and x86 at the same timeMichael S11/03/12 11:12 AM
    ARM v Intel Debate In One ActDoug S11/02/12 10:10 AM
      Some Intel pricing numbers ... they might even be correctMark Roulo11/02/12 10:18 AM
        Some Intel pricing numbers ... they might even be correctDoug S11/02/12 10:45 AM
          Some Intel pricing numbers ... they might even be correctsimno3411/02/12 12:28 PM
          Some Intel pricing numbers ... they might even be correctMark Roulo11/02/12 12:47 PM
  ARM v Intel Debate In One ActAndreas11/02/12 09:17 AM
    ARM v Intel Debate In One ActDoug S11/02/12 10:40 AM
      ARM v Intel Debate In One ActWiseGuy11/03/12 08:01 PM
        ARM v Intel Debate In One ActEduardoS11/03/12 08:40 PM
        ARM v Intel Debate In One ActFoo_11/04/12 06:37 AM
          ARM v Intel Debate In One ActWiseGuy11/04/12 07:00 PM
          ARM v Intel Debate In One ActKira11/04/12 08:34 PM
            ARM v Intel Debate In One Actanon11/04/12 10:46 PM
              ARM v Intel Debate In One ActEduardoS11/05/12 02:07 AM
                ARM v Intel Debate In One Actanon11/05/12 03:37 AM
                ARM v Intel Debate In One ActFoo_11/05/12 12:27 PM
                  ARM v Intel Debate In One ActEduardoS11/05/12 12:59 PM
            ARM v Intel Debate In One ActFoo_11/05/12 12:28 AM
        ARM v Intel Debate In One ActMichael S11/04/12 07:56 AM
      ARM v Intel Debate In One ActAndreas11/05/12 06:55 AM
Reply to this Topic
Name:
Email:
Topic:
Body: No Text
How do you spell blue?