sawyl | Snell on the demise of SGI

Last weeks Guardian feature, Death of the Dinosaur, claims that SGI was brought down by PC clusters. I disagree. Rather, I think that Addison Snell, writing in HPCwire, correctly attributes SGI's great decline to their problems with Itanium:

By 2002, x86 clusters were the norm. Our standard pitches that explained why you shouldn't convert your code to MPI were moot. The codes were MPI, and they weren't coming back. Furthermore, Itanium had marginalized in the market - SGI picked the wrong chip. Back to the drawing board.

Exactly. The great decline seems to date from the early part of the decade, when SGI were still trying to flog systems based on the increasingly dated R10K architecture, while promising customers that everything would be great just as soon as they upgraded to Itanium — something they sold as a drop-in replacement, glossing over the fact that the change in CPU would also require a different OS and different tool stack.

So perhaps it's not all that surprising that people weren't willing to take a bet on SGI as their HPC vendor of choice.