On 2/2/23 23:07, Tom Hunter via cctalk wrote: > Chuck, the CDC 7600 duty cycle integrator is really a work-around against > overheating and has nothing to do with core reliability and/or endurance. > > Core and the data stored in it lives "forever" if the operating constraints > of the medium are adhered to (temperature being one of the constraints). > The 7600 was able to push core access past these constraints, hence needed > the duty cycle integrator. Different physical design and/or better cooling > could have been alternatives, but the duty cycle integrator was an easy fix > once the machine was out in the field and core memory started failing due > to overheating.
Tom, I know all of that, having programmed a 7600. As a matter of fact, what's not mentioned in the manual is that a tight loop in a PPU (which did not have a duty cycle integrator) could throw parity errors. Just taking issue with your statement that "core memory is forever". Another problem with very old core is that its physical integrity is often an issue. Cores can crack, being essentially ceramics. I'm reminded of the CE fishing around in the oil bath of a 7090 with what amounted to a magnet on a broomstick to grab loose bits of core. --Chuck
