It seems reasonable for me. As stated in TRG's web pages it is about 1/8000 probability that something goes wrong (I suppose when RAM is full) at each refresh cycle. If you have only 1M of 8M used, than it is 1/(8000*8M/1M)=1/64000 probability that your data will be corrupted at the refresh cycle. And when we compute long term probability for let's say 14 days: 8M used: 1/8000/cycle, 20160 cycles, 92%/14 days 1M used: 1/64000/cycle, 20160 cycles, 27%/14 days to get *at* *least* *once* crash
In a n minutes it is a 1-(1-1/8000)^n chance that you get an error. In two weeks (20160 min) it is a 92% chance. Calculations are (most probably) based on full memory. gokky
I'm not so sure. IMHO if DRAM doesn't do the refresh, then refresh must be done by CPU. And if CPU must be awaken, then it shortens batery life. Am I missing something? Gokky
Slashdot Mirror
It seems reasonable for me. As stated in TRG's web pages it is about 1/8000 probability that something goes wrong (I suppose when RAM is full) at each refresh cycle. If you have only 1M of 8M used, than it is 1/(8000*8M/1M)=1/64000 probability that your data will be corrupted at the refresh cycle. And when we compute long term probability for let's say 14 days: 8M used: 1/8000 /cycle, 20160 cycles, 92%/14 days 1M used: 1/64000 /cycle, 20160 cycles, 27%/14 days to get *at* *least* *once* crash
In a n minutes it is a 1-(1-1/8000)^n chance that you get an error. In two weeks (20160 min) it is a 92% chance. Calculations are (most probably) based on full memory. gokky
I'm not so sure. IMHO if DRAM doesn't do the refresh, then refresh must be done by CPU. And if CPU must be awaken, then it shortens batery life. Am I missing something? Gokky