3/19/2003

boldly going forward, still can't find reverse

Brian has a post about Apple's updated innards for their latest batch of overpriced underperforming Macs. He isn't one of those insane Macheads that see Jobs as the Messiah (though, Jobs might be the only guy he'd vote for to oust Bush), and concedes several important points (which will likely make him a pariah).

The first point, about CPU bottlenecks, is really the most important:

[T]he innards have been re-architected based on the Xserve plumbing, hopefully with the result that data throughput is no longer as bottlenecked at the system controller as it was. (There's still the CPU bandwidth bottleneck, which won't be solved until the 970 is here, so things could still be plenty better.) But the PCI bus, the ATA/100 controller, FireWire 400/800, and gigabit Ethernet are all directly addressed by the system controller without any bridging, so that might be quite nice.


(emphasis mine) It's great that the system controller can access all the peripheral goodies he listed above directly, but ultimately that data has to go through a single bottleneck - the CPU. In fact, removing the downstream bottlenecks just makes the data pile-up at the CPU even worse. This CPU bottleneck is a crippling, fundamental flaw in the Apple system design.

The second has to do with price. Given that Apple thinks it can be the personal computer for the oppressed masses, it's ironic that even most die-hard Mac fans routinely invoke comparisons to Porsche and other elite consumer status symbols when gushing about why the 50-150% price markup is so worthwhile. Mac users take the superiority complex so seriously that it transcends computing and bleeds into their whole worldview. Brian writes:

There's still the cost issue-- $1500 for the base 1GHz single through $2700 for the dual 1.42 GHz (or more with the RAM and Bluetooth and such tricked out more fully), prices that require some creative thinking and/or drugs in order to place them in the same ballpark as comparable Pentium-class machines. (It's worth noting, though, that these are the lowest price points that PowerMacs have ever had.) But there are still the usual tradeoffs and justifications (system integration, ColorSync, short pipelines, FireWire 800, at least comparable performance);


Drugs indeed. But no, it's not worth noting that these are the lowest price points that the Power Mac has ever had, in fact that damages the case even more. Doesn't that imply that Macs have been overpriced from the beginning, even above and beyond subsidizing software development?

As for the standard laundry list of features that supposedly make the price Worth ItTM, look at each in detail:

1. System integration. The argument is that since Apple builds the whole box, you don't have to worry about finding obscure drivers and other system incompatibilities arising from having too much hardware choice. This is absolutely true for those few pathetic Win 3.11 for Workgroups users out there. There is no real advantage with today's systems - peripherals and hardware come with their own complete drivers for all Windows OS versions. Not to mention, if system integration is such a great advantage, why did Apple allow the CPU bottleneck problem to occur in the first place? System integration is much ballyhooed but to be honest there isn't a single example I can think of where it has actually been an advantage.

2. ColorSync. This is a nifty feature that is unused by 99% of Apple users. Brian himself admitted in an email that he doesn't use it. It's only essential to certain subsets of graphic designers for whom price is not an issue (and who are presumably rich and successful enough that time is not money) - but for the common user who Apple wants to persuade to Switch! , it's just added cost burden for unneccessary and unused functionality.

3. Short pipelines. I don't know exactly what Brian means, but I suspect he means RISC as opposed to CISC. That's a philosophical argument, and Ars Technica has had many detailed comparisons of both kinds of architectures which really demonstrate that it all depends on the software application, as well as the rest of the system's design (ie, CPU bottlenecks are a bad idea). See above.

4. Firewire. Almost all PCs now ship with Firewire ports, if not standard then at least as an option. Firewire was a good technical achievement by Apple but they botched its marketing, initially charging high licensing fees. As a result, USB won the connectivity market - when was the last time you bought a FireWire mouse or scanner? Now Firewire is cheaper to license and it makes an appearance, but there are few pieces of hardware peripherals that use it. And USB 2.0 is better anyway (plus backwards compatible!)

Apple's got benchmarks posted which claim superiority in Photoshop 7 tests over the 3GHz P4-based Dell, which we've all learned to treat with some skepticism-- but which at least show that Apple is confident enough in its machines' competence to make such claims in the full knowledge that independent testing labs will run their own numbers.


No, it shows that Apple knows that enough of it's diehard users are so brainwashed that they won't bother to educate themselves about how the last batch of Apple's Photoshop comparisons were meaningless, and how real and fair tests demonstrated the exact opposite of Apple's results.

Apple and Photoshop tests are like George Bush and Iraq - if they have such a great case, why do you need to lie about it? Even if you have a reasonable case to make, the lie damages the credibility. Except to the faithful.

I'm in the market for a new laptop, and it won't be an Apple.

No comments: