I have left SiCortex. The reasons are not terribly relevant: I still believe in what SiCortex is doing, it just became clear that there were other places where I could have greater impact.

What now?  Well, the industry is still as screwed up as it was when Jud Leonard, John Mucci, and I started SiCortex with backing from Bill Strecker and his associates at Flagship Ventures. In fact, it is even more screwed up: SiCortex and others woke up the sleeping system giants, alerting them to the fact that their hardware was bloated and consumed far too much power for the amount of work they were doing. But the industry reaction has been disappointing, to say the least.

A healthy industry would have responded with dramatically improved designs. The computer industry responded with a barrage of PR events proclaiming “breakthroughs” like 20% less money spent on cooling for just 100% more money spent on plumbing, or 15% reductions in power consumption by letting everything run hotter.

At the same time, procurements still focus too much on matrix multiplication and not enough on the rest of the performance picture. We are at a point in the industry where very little differentiates the systems from the dominating vendors, and we aggravate the problem by applying a metric (LINPACK) that ignores the small amount of differentiation that does exist.

So, even though I’ll be “spending more time with my family” as the announcements always say, I’ll also continue to research, write about, and explore this odd corner of the world that continues to produce fascinating stories and intriguing puzzles.

I’m working on some PETSc experiments for the next post. I’m not sure how long it will take to finish them, but the early results are interesting.