What is the Intel Switch Costing Apple?

SenseOfHumor writes "A Business Week article says that it costs Apple $898 for an Intel iMac before loading it with software and packaging. From the article: 'But for Apple, the switch to Intel chips is less about saving money in the short term, and more about hitching its wagon to Intel's longer-term product road maps, particularly in the area of notebooks. IBM's chips are power-hungry and generate a lot of heat, and therefore not suitable to notebook computers.'"

Comments

When did this change?

[Anonymous Coward]

In the mid 1990s, Apple showed the famous picture of a Pentium grilling a hot dog and claimed Intel's chips were power hungry and ran hot compared to the nice cool sleek PowerPC. That was one of the supporting reasons that Apple ostensibly switched, according to all the engineering presentations at WWDC. So when did this change?

The main reason of course was that RISC processors were on a much faster performance incline than the fuddy duddy old CISC processors like the x86 line. The graph comparing the two in the period 1995-2005 showed CISC acceleration continuing to slow and RISC acceleration continuing with, I believe, a skyrocket attached to the top of the graph. We all know how that turned out.

Re:When did this change?

[dasil003]

No one at the time expected the changes in CISC processors. CISC processors still do have a "complex" instruction set in that they allow multiple forms of adddressing and varying length opcodes. However, internally these chips have become much more RISC-like. The current generation of Pentiums actually does an internal version of dynamic translation from CISC to RISC-micro-ops (which may be 1 or more per CISC instruction) and executes the micro-ops using a different instruction set internally. This internal RISC instruction set is used so central to the design that the L1 I-Cache is not actually a verbatim data cache of the CISC instructions but actually a trace cache of the translated RISC-like micro-ops.

It really just goes to show the error in the view that RISC and CISC are considered opposite approaches to processor design. The dichotomy was more pronounced in the early days of chip design, but the fact was that proponents of both approaches had good points, and so it was inevitable that modern chips combine the best of both philosophies.

I think the progress made on the PowerPC architecture is a testament to its viability. The fact that it's even managed to stay anywhere close to Intel/AMD is remarkable given the difference in R&D dollars (I'm just guessing). But the timing of the Intel switch makes perfect sense.

Consider the switch to the PowerPC in the 90s. It was a time when Microsoft was rapidly catching up to the Mac in terms of UI, and computers were generally underpowered for the common applications that people needed. Gambling on a more promising architecture could have paid off huge if the performance panned out. That never happened, and Apple was in pretty bad shape by the late 90s.

Now, however, computer performance has reached adequate levels for all the things the common people want... audio, video, web surfing, word processing. We can always use more power, but performance is not such a big deal as it used to be. Since they're not seeking a competitive advantage in performance, it makes sense of Apple to at least assure commodity performance by going with the dominant CPU architecture. Apple has contiunously struggled with supply problems from chip vendors for years, hopefully this will now be behind them, and they can focus on the creative part of their business which is where they've always excelled.

Don't We Know this already?

[patman600]

Wasn't this the publicly stated reason for switching when Steve announced the move last summer? They said IBM makes great server chips, but the future of personal computing is laptops, something Intel is putting more R&D into than IBM, and thus provides a better solution.

why is this news?

Re:Don't We Know this already?

[LurkerXXX]

I'm wondering, why Intel?

Where have you been?

#1) It's been discussed about a million times that Apple has had issues chip suppliers before producing enough of the desired chips for them. Intel has the fab capacity to handle any requests Apple makes. AMD doesn't. If Apple went to AMD, they would instantly become AMD's biggest customer. That puts a huge strain on production. AMD is pushing their production capacity as is. If I recall, there were recent shortages of the 3800+ dual core chips. That's without AMD taking on a bigger customer than they've ever had.

#2) Laptops. Laptop sales are growing, and have been higher than desktop sales for the past two years or so. While Intels desktop chips are hot and slow, the Pentium M is a nice fast low-power chip, and slightly better than any of AMD's current laptop chips.

In a few years when AMD has more fab capacity and maybe a better laptop chip than Intel, I'm sure we'll see Apple thinking about moving over to AMD, or at least offering those as optional chips. Or at least threatening to like Dell to make sure they get a sweet deal on chip prices from Intel.

All Intel, All The Time?

[SoupIsGood Food]

This is why I'm confused about the push to "All Intel, All the Time!" Apple, with Mac OS X's Unix and NeXT roots, should embrace a multi-platform strategy to get the most bang for its buck wherever it can. The PowerPC-derived Cell will rock for workstation and servers, and the Meron will kick major butt for home user kit. Best tool for the job, and just compile for the famous NeXT "Fat Binary." Back in the day, the same NeXT executable would run on 68040, Sparc, PA-RISC and Pentiums. Why not now? Why tie yourself to x86 alone, when there are better alternatives to fit the niche you're targeting?

Too much politics, and not enough engineering.

~ SoupIsGood Food

Brainiac design

[argent]

The G5 is a "brainiac" design, a big complex chip with a long highly parallelized pipeline. This is a relatively new approach for RISC chips, which have typically concentrated on a small core, short pipeline, and simple design with a lot of "close" cache.

Intel's Pentium chips have all been "brainiac"s to some extent, but none so much as the P4... which they've backed away from. The new chips in the new Macs are less like the G5 or P4 and, while not exactly as clean and tight as the G4, are closer to it than they are to the real brainiacs.

But there's nothing wrong with the G4 core as a core. Taking the G4 core and giving it a faster bus, the way Intel's taken the PII/PIII core and given it a faster bus in Yonah, would have made a lot more sense. And Freescale's got one like that in the pipeline. They could have called it the "G5 Mobile". :)

Re:Pentium-M

[aaronl]

The P3 design is old (dates back to Pentium-Pro), but it provides more per clock than the travesty that is the P4. It doesn't have constant issues with branch predication failures causing a potentially 24 cycle execution halt, from flushing the pipeline, for example. Compared to P4, it also requires substantially less power to do the same work *and* it's much less expensive to manufacture. The P4 is a bad design, and it's amazing the amount of money they've thrown at it, just to have to keep working on P3 because of the P4 shortcomings.

Re:Cool

[BrookHarty]

Well, to be fair, at the time Intel was pushing the heat pump P4 against the PPC. Now intel dropped back to the P3, modified it, called it the pentium-m, now the duo-core. They had to do a 180 and rethink their roadmap.

I think the greatest thing will be virtualized intel cpu's running multiple copies of OSX for servers. Or even Windows and OSX. Apple xserves will look very attractive now when can do anything, have apple quality hardware, and have true migration to any OS or software that you need. Brilliant move.

Re:the real costs

[Jerry Smith]

porting operating system $30,000,000

OS X derives from NextStep/OpenStep, and has been developed for n86 from day 1. They ported every release to PPC. Yes off course it needed aftercare, but still: first for n86, then to PPC.