3 ms·
I have a few war stories about this from back in the day when I worked in hardware design: we made a chip that was supposed to have embedded SRAM but the cell d
by dboreham 8y ago
I have a few war stories about this from back in the day when I worked in hardware design: we made a chip that was supposed to have embedded SRAM but the cell design vs process was a bit off and it turned out to leak charge slowly, making it a DRAM. No problem : we modified the software to hit every row every so often which in effect implemented refresh. Also the converse problem: if you have a bug in your hardware that results in refresh being disabled it can be hard to detect because as noted above cells typically have a retention time on the order of 10's of seconds. Add in the effect of programs running, which causes accesses that as a side effect refresh the row targeted, and it can take a long time before you realize there is no refresh. The symptoms can be subtle and confusing, because for the most part memory _is_ being refreshed, just not in the way you expect.
- jacquesm 8y agoI think you have just finally solved a problem with a piece of hardware that I was asked to work on decades ago. Never could figure out exactly what caused the problem but reading this I suspect you are correct in that that particular system had a broken refresh design. It worked most of the time and then would crash in the weirdest ways, and it did contain 64K of dynamic RAM, I never thought to check the refresh circuitry. I was tasked with debugging the software end of things, and you know the proverb about programmers faced with a hardware problem. Extra silly of me because I really should have started with checking the hardware first, the assumption that the problem was software related clearly may not have been a valid one. The refresh circuitry was provided by a video chip and this is most likely where things went wrong, either in the configuration or in the actual wiring. Thank you!