I wonder how the HPC community is going to respond; there is a not insubstantial
community who heard "150Gflop/$400" and "Linux" and decided to build clusters from PS3s. Those machines can probably just have updates held back, but it makes replacement a problem.
To forestall the inevitable "that isn't a serious use" argument, US Airforce owns
Something like 2,500 PS3s for compute work.
Killing Linux on the PS3 also presents something of an issue for the other Cell "partners", who seem to be looking at the PS3 as a low-cost Cell development starter kit. The other Cell machines on the market are *much* more expensive (an IBM QS22 blade is $8-20k, depending on configuration, and Mercury Computer Systems doesn't even like talking about how much their Cell boards cost). Given that Cell is an enormously difficult architecture to target, having relatively inexpensive systems to test and train on is very desirable for the other vendors, especially now that so many of the HPC folks are fixated on GPGPU, which is also terrible to program for, but has a far lower cost of entry. It could be that
IBM's decision not to pursue Cell in the HPC market is how it became politically tenable for Sony to kill off Linux on the PS3.