Regarding your post which is repeated in its entirity below. As opposed to getting religious about all of this, or arguing with you about the goodness of one company versus the evil of another, I just wanted to note why I keep reading this newsgroup since I don't own any SGI shares. I read it because I want to invest in SGI. I think that they have a valuable franchise. I don't think that the franchise is based on vertically integrating the microprocessor with the platform with the operating system with the application with the sales with the service with the whatever -- instead, I believe that the franchise is valuable because SGI is the best 3D company in the world.
In another post you mention something about "50,000 versus 5000" engineers. I bluntly don't grasp what this has to do with your original argument. The subject of smaller engineering teams was covered exhaustively 20+ years ago by Fred Brooks in "The Mythical Man Month" -- no one is going to argue with you that smaller is better -- but the fact is that larger is a reality because of competitive schedule pressures.
Getting back to your original point, I thought I'd submit to you the type of airplane I'd like to fly in if I could in terms of the design environment used by the engineers.
The Boeing airplane I would like to fly on would be designed by several companies, each of which would have a much larger R&D budget than any one company. I'd like the workstation used to design this airplane to be based on an Intel microprocessor -- and since I see the R10000 every day and I see the Pentium Pro every day I can tell you that I don't think that this is going to hurt anyone. I'd like to see the workstation based on Windows NT -- but since this would limit the scalability you could achieve in a highly-shared multiprocessor environment and would also hurt your availability I could understand limiting NT to workstations only and using Unix (or MVS, or whatever) as the server operating system as as the computational farm for tasks that can't be easily partitioned among workstations (e.g., not like rendering).
The 3D graphics system would be designed by SGI. We can argue about the fact that you need greater design coupling between your microprocessor and the 3D graphics system than is available over PCI or over Intel's AGP or whatever -- but I'd simply respond that you can have direct access to the memory bus and that I'd use the loading normally used for one of the microprocessors in a uniform SMP-based multiprocessor to locate the 3D graphics system (for more on how you'd actually implement this, look at all the companies putting "8-way" Pentium Pro-based systems together with ASIC's.) The bottom line is that it's technically feasible.
The software would be designed by Autodesk, or Microsoft/SoftImage, or SGI Alias/Wavefront, or Adobe/Photoshop, or whatever (I realize I'm mixing a lot of software here) -- the key would be that the airplane designer would be able to pick the best software and not be limited by application availability.
If Boeing ever uses the workstation I just described, I'll invest in SGI. And I'll keep looking for any and all signs that SGI is headed in that direction. At this point, however, I agree with one of your statements below -- SGI is not showing an inclination to rapidly move in this direction.
>>Regarding this NT stuff, please post only valid opinions that you can back up. I don't believe for a second that SGI is even contemplating such suicide. They have more to lose than to gain by adopting Intel/NT based PCs. Many of their devoted customers will certainly question their focus and longevity when competing with brain dead H/W from Compaq et al. Do you really believe DEC will ever become the glorious company they once were now that they have to help pay for Gates' toys. Never in this lifetime. If SGI wants to compete in the low end world of the PC, they must make an O2 that people can buy at a competitive price. I for one would like to chuck out my PC for an O2 even if I had to pay a $500.00 premium, but I can't justify a premium of $4,000.
Software is another beast altogether. Their justification IMHO for spinning off the Cosmo software was so that they can port it to whatever platform made sense from a strategic and financial perpective. Absolutely nothing wrong with that. Actually, that is the best way to go!
Frankly, I would rather fly in Boeings being designed on a SGI workstation than on one designed on a Compaq. Just thinking about the constraints that an engineer would face designing such a complex piece of machinery on a limiting piece of hardware (Pentiums and NT) scares the shit out of me. << |
