 |
 |
Error Correcting Memory If this means what it used to mean (which cannot be guaranteed for acronyms, especially in computing) it is memory which stores additional bits alongside the actual data, in such a way that it is possible to correct some errors in the write/read process. The basic method is to group the bits, typically into a slab of 64 bits (ie 8 bytes) and then to compute a collection of parity bits over sub-sets of the 64 data bits. These extra parity bits are storerd alongside the original data bits. When the data are read back, the parity bits are recomputed and compared with those originally stored. By doing this over enough sub-sets, it is possibile to detect one or two bit errors; however the smart bit is that it is also possible to decide which bit is in error for the single bit error. Since you know which bit is wrong, you can correct the error by simply inverting that bit. There are extra costs. You need an extra byte of memory for every 8 bytes stored, and you need extra circuitry to compute and check the parity digits. Modern memory is pretty reliable, and my guess is that unless you are into very high reliability environemnts (eg direct control of a patient life support system) error correcting memory is not likely to be worth the extra cost. You might do better to spend the money on MORE memory, and improve overall performance. Mike Wells ========================================================== Professor Mike Wells Department of Physics, The University of Leeds Leeds LS2 9JT, United Kingdom Phone: 0113-233-2339 E-Mail [log in to unmask] ========================================================== %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%