Apple know more than any OS company out there about UI, but they have a horrible track record for writing software that runs anywhere but on a tightly controlled system. Writing for Intel does increase one's problems exponentially, because now you also have to troubleshoot someone's hardware configuration. And you're right, to be efficient, Apple writes its OS with a deep knowledge of the hardware. This may not necessarily translate to the same performance against two different Intel boxes.
I can't say if Apple is as bad as Be when it comes to knowledge of the hardware, but I was extremely disappointed in BeOS. It was so cool and had so much promise, but I couldn't get it to install on one of my systems. I bought at a CompUSA 2 Toshiba Infinias on the same day. I installed BeOS on one, it worked like a champ. It would never install on the other one. I spent hours and hours and hours trying to figure it out- after consultation with Be, it turned out that there was a tiny rev difference in the chipset between the 2 machines. That was enough to make BeOS install on one Toshiba and not on the other. No wonder Be never made it! If Apple is that dependent on the hardware, it will never make it in the Intel arena.
Still, this is not rocket science. There should be a way to write an OS that can work in either Intel or Motorola environments - that is, assuming Apple wanted to do that.
going to Intel would probably burden Apple more than anything. First, the lack of chip competition resulting would hurt Apple more than it would benefit the overall scheme of balancing OS power, hardware power, chip power and internet/raw clustering power. Being slaved to Intel is not the solution, to harness it maybe, in, say, an OS capable of working with it, but not for it.
The ability to cluster Apple/PowerPc chips at the software level would be much better at exponentialy boosting Apple machines powers than investing all that $$ into a software retooling to AMD/Intel, making us slaves of chips instead of balancing ease of use, power expandability and scalability. Once you have a chip, all you need is to know how to use them. It does not matter how big they are, just as they are reasonably functionable and within the latest RISC technology.
http://www.architosh.com/news/2002-09/2002c-0913-rop-clustering.phtml
Anthony Frausto-Robledo (
afrausto@architosh.com)
13 Sep 2002
Reader Opinion: Apple needs OS-level clustering, G5 and more bandwidth
One of our more astute and 'busy bee' readers wrote in to comment on follow-up emails we have been plowing through. It seems our discussions have come full circle on a number of critical issues to Apple and its goals in the high-performance markets, such as 3D and animation. I have assured this reader -- like many others who are very critical of Apple's progress at this moment -- that our beloved platform will once again (clearly) reign supreme over x86 in sheer application performance.
The G5 and G4 Debacle
Our reader wrote in to say: "And you are right in saying that Apple REALLY NEEDS the G5 and it needs it yesterday. Apple MUST leapfrog the competition with respect to speed." "They need a fast machine that will be able to compete with any of Intel's future offerings."
As it turns out now, we have been able to confirm that our information on the G5 test boxes late last year were more or less right. At that time Motorola was a bit ahead of schedule on the G5, and there were over-clocked versions running as high as 2.5Ghz.These were early test versions. Unfortunately, for reasons we can't touch upon here, the G5 progress was stalled. It should be noted that, most sites -- including ours -- were dead wrong in predicting when the Mac would get a genuine G5. The G5 development team had planned for a 2003 volume delivery, despite published PowerPC road maps.
We understand that Motorola was absolutely right in stating a ways back that there was plenty of life left in the G4 architecture. And we will likely see two more iterations of the G4 over the next year or year and a half. We understand that there is a Motorola MPC7460 and MPC7470 in the works. Those processors may involve a 0.13 micron process, support for faster bus frequencies and other changes in addition to greater overall chip frequencies.
The bottom line is that while Intel (especially) may be pulling way ahead in megahertz frequencies, Apple is making significant motherboard architecture improvements that will allow the G4 architecture to reach its maximum performance. The current crop of Power Macs involve these performance enhancing changes. For those waiting for the G5 before they upgrade their gear...you may be waiting a long time. Keep in mind that the new dual 1.25Ghz model has a peak performance of 18.3 gigaflops. That's more than 490 percent faster than a 500Mhz G4.
Clustering in Future Mac OS X Versions
There is currently a notion going around that the next core after Jaguar will be clusterizable (if that's a word) at the system level. Our reader reports: "If you recall, a couple of years ago I opened a dialogue with Dean Dauger (Think: Pooch / AppleSeed) and he did say at the time that Apple is considering using these technologies on the OS level. I'm certain that it's only a matter of time before we have plug-n-play clustering."
These notions play out very well with a number of items reported on this site and elsewhere. For starters, we too covered the Pooch technology and spoke to Dean Dauger. We were also informed of Apple's hush-hush Studio Summit, if indeed that is what it was called. There Apple solicited advice on what it would take to get major studios to adopt Mac OS X as their primary creative platform. From our source there were a number of key items requested, some of which have been met:
A Rack Mount Server which could allow for clustering for Render Farms
Greater Bandwidth on the motherboard
Faster Processors
Faster Graphics Cards (a workstation class card) and faster Graphics
easy plug&play clustering software built-into the OS
Industry leading OpenGL support: As in Jaguar OpenGL
According to our source the rack mount server was one of the most popular items requested because big studios need inexpensive render farm solutions. They also need storage. Both of those items are being or have been fulfilled by Apple (as in the new Xserve).
As for faster RAM and more performance on the motherboard? Well, the latest Power Macs demonstrate that Apple has the ability to be innovative and re-architecture the entire system to take advantage of inexpensive RAM. As for graphics cards, ATI came back with a vengeance with their Radeon 9700 card, and Apple has executed a major innovation in the way Quartz operates and is accelerated with Jaguar OpenGL.
What remains to be completed is an easy to configure clustering software built-into the Mac OS's API structure so that developers can take advantage of it. With such an addition, fx studios, 3D graphics professionals and architects can take advantage of multiple Xserves and the home brew clustering of Power Macs running the latest Mac OS. Pooche is a way to do this today, but it still needs apps to be aware of it...and Pooche is not really true plug & play technology.
The Motherboard Issue
Currently the Power Macs' motherboard bandwidth is limited, compared to AMD and Intel systems. As some have speculated, the current G4 processor is to blame there. Faster DDR RAM support will come with a newer G4 processor. The main system bus should improve over time, but Apple's innovative motherboard architecture doesn't necessitate Apple maintaining frequency parity with x86 systems.
In short, the requests of power users like our reader, that assume that a G5 is what is needed to keep Apple ahead, is partially wrong. Apple will stay ahead of the performance curve through balanced system design and innovation, with and without the G5.