March 02, 2004

Of Chips and Businesses

Steven Den Beste is back and feeling a bit beset at some of the response he's getting on the XBox's use of the PPC chip. In some small part it's deserved as he's missing a few points but he obviously got an overload of viciousness he certainly does not deserve. Don't the flamers know to ease a guy back into the swing of things?

Regarding the PPC, chip pricing and Apple, things do need to be cleared up, and not just with Steven Den Beste's post but with the original that started things up, an article over at Brian Tiemann's Peeve Farm.

First of all, the original article is too limited to reliably transmit much emotional content at all. Thus I don't think either characterization of "overjoyed (and oversmug)" or "tongue-in-cheek and 'isn't that [w]eird?'" By word count it's a 90% cut and paste from the underlying Inquirer article. It's a real linker not thinker article so it's very thin gruel that SDB is hanging his article on.

But things are a little weirder than the Peeve Farm's noticed. It is quite probable that the custom NT kernel is just updated enough that it'll handle the new hardware. NT for PPC was discontinued at the SP 3 level and according to Microsoft Windows NT 4.0 Workstation went completely out of support on June 30, 2003 so if you have any problems with security, functionality, etc. you're just out of luck.

Microsoft Support Lifecycle policies nowadays are pretty well set out and NT 4 Workstation is just not going to be supported. This is not tolerable for a major Microsoft product line so either Microsoft is going to have to tailor an exception to their support policy or they are going to reintroduce multiplatform compatibility to Windows.

I think that NT 4 for PPC is going to be a very limited release and will be replaced by a more modern and secure version of Windows for the PPC as soon as Microsoft can crank one out. This is a major development effort that requires an awful lot of code cleanup and recompiling so why is Bill Gates doing it? You don't just restart a major platform line just for one product, an SDK to a game machine that you're adopting an inferior chip for as an anti-trust exercise to financially keep alive a semi-moribund competitor (Apple) that will not change their way of doing business so are going to die in the long run anyway.

SDB seems to have come to the conclusion that Apple has not evolved recently. In fact Apple has, and in very positive ways. In product after product Apple is demonstrating that it has squared the circle and learned to love open systems without sacrificing the anal retentive control freakiness that makes end users put up with all the disadvantages of a closed system.

Apple's Safari product is a case in point. Safari is actually just a user interface shell wrapped around KHTML. This is the reason that Apple can tolerate the expense of supporting their very own browser, they aren't really spending the money. They find partners in the open source world with whom they can work, who have engineered good products, they move in and create a secret skunk works version to improve things up to Apple standards.

While Steve Jobs is still on the keynote stage creating his latest reality distortion field presentation, the engineers are in the back rooms doing a massive code dump back into the upstream project, confident in the fact that their code is good and will be subsequently accepted and maintained by the open source geeks, creating both lower costs and a positive network effect. Konqueror wasn't widespread enough to have HTML coders specifically test for compatibility and code to it but Konqueror + Safari + whoever joins in is a different story.

The result is lower costs for development with much higher volume because the engineering interfaces that others interoperate with are not the same as program names. If your threshold for rewriting a site to a browser is 10% of the visitor traffic and Apple has 7% of your hits and Konqueror has 4% you write to the common KHTML engine and both groups are happy.

This is not the Apple I first encountered in the mid-80s or even the Apple of six or seven years ago. It is an Apple that is concentrating on engineering aesthetics as its core product and it expresses that through its operating system and application software as well as its Human Interface Guidelines. Apple's best value proposition has always been its UI and aesthetics and the enhanced productivity that flowed from those choices. They've recently figured out how to purify their offerings so that they don't have to give up control but can radically reduce cost and increase volume via these open source partnerships.

But back to the heart of the discussion which is computer hardware. The decision of a chip line ripples extensively. There is tremendous inertia and a temptation to stick with what was tried and true. This is why chip manufacturing companies create extensive roadmaps so their large customers can make their own long range plans to satisfy their own customers over the next decade. Even if a chip line is currently in first place, if your roadmap stops 3 years out and your message is go remake your own entire product line, a lot of customers are going to jump ship, heading for the exists as soon as they can.

This is the true heart of the problem for Microsoft. Intel has been saying for years now that 32bit will stay on the desktop and 64 bit will be for servers only and very highly specialized low volume workstations. Their solution was to keep on with the Pentium line and to switch to an entirely new architecture, Itanium. The entire Xbox 2 saga is Microsoft's delivery of a message to Intel rejecting its strategy because of roadmap concerns.

They saw how Apple hemorrhaged customers and lost credibility when their chip supplier had a roadmap problem. Microsoft cannot afford to get into that same bind because they would lose decades worth of psychological positioning as the default choice for a wide variety of markets. Microsoft doesn't have to be the best, but it can't afford to fall behind Linux too much or it will lose enough customers that it will, once again, descend to the level of just another computer solution provider with a wide variety of products. It can't afford that because so many people utterly despise them for their past competitive behavior and would love to exit the Microsoft product ecosystem if only they could.

Now this is more than just my naked speculation because Intel, itself, has gone back on its roadmap and surprised everybody (but AMD) with the announcement of a Xeon processor that is x86/64, something that they swore they would never do because it would absolutely kill Itanium.

For years Intel declared that x86/64 would only be a slight improvement and would peter out quickly against alternatives like the Power/PPC combination. As recently as October, 2003 speculation was still rife about whether Intel's secret "backup" project called Yamhill really existed. By February of 2004, Intel was claiming that it meant to do this all the time. This is a orwellian revisionism of the highest sort with everybody expected to flush their old roadmaps down the memory hole and forget all of Intel's previous disparaging talk about x86/64 and its limitations. A good examination of the current uncertainties in the Intel camp can be found here.

So Microsoft going to PPC for its XBox 2 and restarting the Windows on PPC production ramp make sense. IBM's roadmap is clear, the 970 is a cut down Power 4 but Power 5 is already out and the 980 will soon be too with further improvements to the Power line to follow. IBM makes too much money from the Power line not to keep on developing it as a priority. With IBM in the AIM driver's seat instead of Motorola, PPC is a safe choice for Microsoft as a backup to a x86 roadmap disaster that is growing in probability.

Now this disaster is not going to happen in the near future. Intel might figure out a way to avoid problems but even a 5% chance of such an occurrence is enough for Microsoft to be scouting for backup plans and the XBox 2 and the adjustments to MS' OS product line to re-include PPC in their lineup are just as good business as Apple's Darwin-X86 project for the reverse challenge, a PPC roadmap disaster.

The low prices that IBM is giving Apple, though, are likely not from any Machiavellian payments under the table or other such things. The low prices stem from the fact that PPC has tended to have smaller die sizes than Intel chips at comparable levels of performance. For the uninitiated, chips are made in groups on a round silicon wafer that is later cut up with individual chips that have tested good ending up in conventional packaging and sold. The cost per wafer is the same whether you make a big chip or a small chip and simple geometry tells us that you can fit more small chips on a wafer than big chips, thus your cost per chip are lowered as the fixed wafer price is spread out among more chips. Each 'bad' chip, also has a lower cost as you are removing a smaller percentage of the wafer's surface from productive use.

So PPC has always had a unit price advantage at similar chip fabrication technology levels. And Microsoft is not the only company to notice that x86 has the potential to have a roadmap disaster. IBM sells an awful lot of x86 gear in competition with its own Power/PPC gear as well using a variety of operating systems. It has its own reasons to gear up PPC because it has its own customers who need to be satisfied irrespective of whether Intel can get past this rough patch or not.

Much of this is covered in SDB's post but he declares that "Apple probably isn't happy, but had no choice" which is really where I part company with him. Apple gets royalties out of every XBox 2 sold as part of the AIM alliance and it did not do so for every XBox. Apple also gets an easier migration off its own hardware if Microsoft supports PPC. Apple might have historically subsidized software production to sell more boxes but it is doing so less than it used to, witness the phasing out of free updates for its iLife suite of programs. It's still bundled with new hardware but it's also sold separately. That's not consistent with the 'old Apple'.

Apple's opening of their architecture came with the end of the hardware ROMs. That event happened many years ago, long before Steve Jobs came back. In fact it happened some time before Apple opened things up to clone manufacturers. It's unlikely that such an old concession to openness would be withdrawn unless Microsoft ponied up some PPC business to IBM. If Apple went back to proprietary ROMs, it would certainly scare off a lot of the open source advocates that are supporting Apple's software efforts for free these days and provide nothing but bad press. It would also create an awful lot of work for Apple and would kill the Darwin project, a key part of Apple's modern strategy where the core OS (Darwin) is an open source project. That core is where most of the interaction occurs between OS and ROM/firmware so making a proprietary ROM but leaving Darwin open would make reverse engineering relatively trivial.

No, Apple has got to be feeling pretty happy right now. They want more PPC customers because they gain profits on every PPC that ships. Their core value proposition would not be too affected by Microsoft's adoption of their preferred chip and the perception that there is Apple in their own world and then everybody else would be dealt a harsh blow. Hypothetically stranded Apple customers could just load Windows on their macs and replace them with other branded PPC machines in future. The great Apple bugaboo, what to do if Apple dies, would finally have a convincing and final answer.

Posted by TMLutas at March 2, 2004 02:50 PM