[Bill Davis] "I want Apple to surprise me."
I used to want surprises from Apple, too. Now I'm more specific -- I want
good surprises.
[Bill Davis] "Every other manufacturer on the planet starts from what the new "generation" of components (busses, processors, drives, etc) do - then simply boxes them up with different name badges, case designs, and maybe an eye to the failure rating of the particular batch of parts they pick out to use."
To do otherwise means you have to be better at microprocessor design, fabrication, and integration than Intel is. That's a tall order. Intel won in the marketplace by providing better-performing products at reasonable prices, and no other company, including AMD, DEC, IBM, Motorola, and Sun, have been able to dethrone them in 30 years of trying.
Further, sticking with Intel designs and other industry standards provides compatibility. If Apple were to step away from Intel hardware, not only would they lose Thunderbolt, they'd also force third-party developers to recompile (at best) and rewrite (at worst) their code for a new architecture, or be forced into emulation a la Rosetta.
Apple did buy PA Semi a few years ago, but their expertise seems to be optimizing for low power consumption. While the performance they get at low power levels is impressive, that doesn't necessarily make them a good candidate for competing with Intel's Xeon processors, and performance is presumably still important to anyone considering a workstation.
[Bill Davis] "Apple is unique since they, more than most, invent, spec, and manufacturer MORE of their own solutions in more areas than most companies. So they have a chance to innovate rather than exclusively increment their machines."
Apple is an outstanding designer and marketer. They do have quite a bit of highly innovative in-house engineering. However, they also do a lot of integration -- buying technologies off the shelf, or buying out companies for their technologies -- and using them in their products.
Does Apple actually manufacture anything anymore? Don't they use contract manufacturing exclusively now?
[Bill Davis] "Thunderbolt has the promise to change everything. If we ever get to TB level 3 (all optical) - with 100 times the throughput of current buss speeds - then everything changes. Anything hung on the TB chain acts as if it's bolted directly to the processor. So the requirement for one box to hold everything disappears."
I've argued this point with others, too. Thunderbolt is a very cool technology, but it's still just an expansion bus, and with Intel's current architecture and roadmaps, it will remain an expansion bus. It's not at all like bolting anything directly onto the processor.
I think another flaw in this reasoning is that it assumes that all progress on all other interconnects will stop while Thunderbolt alone progresses. Intel's QuickPath Interconnect technology -- the system that actually connects the CPU(s) to the I/O controllers or routing hubs on the motherboard -- is immensely faster (by design and necessity) than Thunderbolt. PCIe 3.0 is here, and PCIe 4.0 is under development. I see no reason to believe that development on these other fast interconnects will stop.
I don't mean to try to minimize the impact that Thunderbolt will have. Fast external expansion is a big deal. However, as long as internal interconnects remain faster than external ones and remain segregated by design within the system, there will be performance advantages to keeping some things within the box.
These things will be less and less important for straight-up video editorial, which is now becoming less reliant on the CPU and more reliant on disk and GPU throughput. My world includes 3D rendering and compositing, where things like CPU performance and memory bandwidth are serious constraints.
Walter Soyka
Principal & Designer at
Keen Live
Motion Graphics, Widescreen Events, Presentation Design, and Consulting
RenderBreak Blog - What I'm thinking when my workstation's thinking
Creative Cow Forum Host: Live & Stage Events
