4 ms·

SGI's story was somewhat similar. Some really bone-headed decisions were made at a high level, which resulted in the company losing contact entirely with the wi

SGI's story was somewhat similar. Some really bone-headed decisions were made at a high level, which resulted in the company losing contact entirely with the wide bottom of the market, instead retreating to the narrow top. It's pretty cool to be able to sell a mega-supercomputer for the price of a small castle once in a while, but it turns out it's not a viable business model long-term.

groby_b 12y ago
It is, until people can just click a button and buy a bunch of smaller machines with the same aggregate power for less. Neither SGI nor Sun were boneheaded, per se. They simply missed the signs of the Internet and cheap commodity hardware. Just as MS missed the Internet and Mobile. And I'm sure today's high-flyers are missing something too. I guess what I'm trying to say is that in an industry moving as fast as ours, the only viable long-term model is to change your model all the time.
- hga 12y ago
  Errr, I may not have made it clear, but Sun did make commodity hardware, as in AMD and later Intel server boxes that would run Linux: http://en.wikipedia.org/wiki/Sun_Microsystems#x86-based_systems http://en.wikipedia.org/wiki/Sun_Microsystems#x86-based_syst... Not "cheap", but of at least perceived high quality, and something that many startups found attractive. The latter just couldn't actually buy these boxes in sufficient quantity. Lots of companies suffer because they screw up their sales, fail in that sort of basic blocking and tackling.
- emmelaich 12y ago
  I'm pretty sure a lot of Sun were boneheaded. I had an IPX at work in the early 90s and the 486/66 I bought for home ran rings around it. Their off again on again attitude to Solaris on x86 and their closing down of their free C compiler ceded the market to Linux. Oh and about Java. When they brought in their Coolthreads machines Java ran abysmally without special settings. You would think the Java guys would talk to the Coolthreads guys to get nice settings out of the box. Sun were like three companies in one who didn't talk to each other. It's not just the hardware made Java perform like crap. Java made Solaris perform like crap too. http://www.brendangregg.com/DTrace/smc.html Java reading a single file 12million times increased the boot time from 10 minutes to 30 minutes!
  - hga 12y ago
    There's a story that the Sun486i, which went as far as a small pre-production batch, was vetoed by the internal SPARC types in part because it was faster than what they were delivering at the time, so the company didn't followup on the wildly popular Sun386i.... On the other hand, you're comparing a 1991 or earlier SPARC chip with the late 1992 486/66. Weitek offered their own after market clock doubled SPARC POWER µP, although I'm not sure it gained as much as the clock doubled 486. The Solaris story on x86 was poor, indeed: in 1999 I set up a x86 machine with a removable hard drive so it could run the various operating systems our customers used, Solaris was the only OS we couldn't get running on it (and I first bare metaled a UNIX(TM) system in 1981 (sic) so I don't it was our fault). One or two BSDs including the commercial one, Linux, forget what else, but not Solaris. Then again, as I note plenty of people were willing to run another OS on Sun x86 hardware, but couldn't buy it. And echoing your other Java comments, a 2001 project I was involved in, running on SPARC hardware, dropped the idea of using Java because the SPARC implimentation at the time was too buggy. As for their closing down their free C complier, I myself never found it difficult to get GCC running on a SPARC machine through the '90s. But, yeah, that added a barrier.
    - emmelaich 12y ago
      Re the IPX acknowledged, but I think Sun were still selling them at many times the cost of a 486/66. Re Sun C and GCC - it's not the difficulty of getting GCC on; its the fact that the whole culture of development started to revolve around GCC rather than Sun C.

groby_b 12y ago

It is, until people can just click a button and buy a bunch of smaller machines with the same aggregate power for less. Neither SGI nor Sun were boneheaded, per se. They simply missed the signs of the Internet and cheap commodity hardware. Just as MS missed the Internet and Mobile. And I'm sure today's high-flyers are missing something too. I guess what I'm trying to say is that in an industry moving as fast as ours, the only viable long-term model is to change your model all the time.

hga 12y ago
Errr, I may not have made it clear, but Sun did make commodity hardware, as in AMD and later Intel server boxes that would run Linux: http://en.wikipedia.org/wiki/Sun_Microsystems#x86-based_systems http://en.wikipedia.org/wiki/Sun_Microsystems#x86-based_syst... Not "cheap", but of at least perceived high quality, and something that many startups found attractive. The latter just couldn't actually buy these boxes in sufficient quantity. Lots of companies suffer because they screw up their sales, fail in that sort of basic blocking and tackling.
emmelaich 12y ago
I'm pretty sure a lot of Sun were boneheaded. I had an IPX at work in the early 90s and the 486/66 I bought for home ran rings around it. Their off again on again attitude to Solaris on x86 and their closing down of their free C compiler ceded the market to Linux. Oh and about Java. When they brought in their Coolthreads machines Java ran abysmally without special settings. You would think the Java guys would talk to the Coolthreads guys to get nice settings out of the box. Sun were like three companies in one who didn't talk to each other. It's not just the hardware made Java perform like crap. Java made Solaris perform like crap too. http://www.brendangregg.com/DTrace/smc.html Java reading a single file 12million times increased the boot time from 10 minutes to 30 minutes!
- hga 12y ago
  There's a story that the Sun486i, which went as far as a small pre-production batch, was vetoed by the internal SPARC types in part because it was faster than what they were delivering at the time, so the company didn't followup on the wildly popular Sun386i.... On the other hand, you're comparing a 1991 or earlier SPARC chip with the late 1992 486/66. Weitek offered their own after market clock doubled SPARC POWER µP, although I'm not sure it gained as much as the clock doubled 486. The Solaris story on x86 was poor, indeed: in 1999 I set up a x86 machine with a removable hard drive so it could run the various operating systems our customers used, Solaris was the only OS we couldn't get running on it (and I first bare metaled a UNIX(TM) system in 1981 (sic) so I don't it was our fault). One or two BSDs including the commercial one, Linux, forget what else, but not Solaris. Then again, as I note plenty of people were willing to run another OS on Sun x86 hardware, but couldn't buy it. And echoing your other Java comments, a 2001 project I was involved in, running on SPARC hardware, dropped the idea of using Java because the SPARC implimentation at the time was too buggy. As for their closing down their free C complier, I myself never found it difficult to get GCC running on a SPARC machine through the '90s. But, yeah, that added a barrier.
  - emmelaich 12y ago
    Re the IPX acknowledged, but I think Sun were still selling them at many times the cost of a 486/66. Re Sun C and GCC - it's not the difficulty of getting GCC on; its the fact that the whole culture of development started to revolve around GCC rather than Sun C.