In a flourish of engineering esprit de corps, a presenter from MEMs oscillator startup SiTime got a spontaneous burst of applause. A questioner from the floor put SiTime's Aaron Partridge on the spot to figure out the impact of putting a quartz crystal watch in the freezer. Partridge quickly rattled off the math on his feet to come back with a snappy answer—two seconds a day.
Engineers were not always so interactive. I'm going to hate myself for saying this, but organizers should shut off Wi-Fi in the auditorium next year so that those who attend must actually be intellectually present. I am sorry to report many of us were so deeply embedded in the Web that some Q&A sessions had no Q.
On an ironic and brighter note, industry consolidation actually spawned one of the more technical sessions at the event.
Now that the Wintel world has given up on the frequency race to get on the multicore train, it discovered it needed to shell out a few million dollars to academia to help define a parallel programming model before its 64+ core parts arrive in 2015. The Berkeley, Stanford and Illinois researchers on the receiving end of the money gave an update on their progress rewriting today's software stack for multicore.
One hot topic of discussion: how dumb software is sucking up many of the extra cycles microprocessor designers are providing. Over lunch, Intel and Nvidia engineers reassured some of us that Windows 7 will at least get us back to performance levels we remember from Windows XP days.
My fear for the future is the plaque of dynamic link libraries you cannot fully flush from your PC and the background Web services always snaking through your system will continue to choke any PC processor. I worry the rush to Internet app stores will create an even broader, thornier problems for tomorrow's smartphone SoCs.
Will we need to do daily hard resets on tomorrow's cellphones? That's a topic of lunch conversation for a future conference.