Upon reading several threads about this common issue (Furballi, Inertia, hellbitch among other ones ) I have decided to change back the NEC 3500 to MB secondary IDE as master, leaving the ‘data’ HD (Western Digital 160GB) as the slave. Of course, UDMA re-appears, as expected.
Transfer rate tests now exhibit a profile very similar to the Liteon 5232K, i.e., upto 16x read speed, 26+ MB/s burst rate, and CPU at similar lower values (directly related to DMA, as we know).
Performance of HD (normal operation) is not degraded, more considering NEC writer/reader is normally off
Determining (Nero) maximum speed for burning a DVD disc using that slave HD as the data source for the NEC master (same IDE bus), a value of 34x is shown, so NEC could easily burn at 16x, just ‘yawning’. Tested a Ridisc Ritek G04 (top silver / magenta) at 12x writing using V2.18 Liggy-Dea v2 (beta3, yet) and effectively the burning is cleaner (no 10x-11x clamp) and final PI/PIF is cleaner. Transfer rate test (Nero) = 16x, both NEC reader or Liteo 5232K reader.
I assume the master / slave affair is almost negligible, except in the case (not the current one for me) when 2 optical drives are in the same bus, one as a data-source, the other as the writer. In that case (own experience with CD’s / VCD’s, old PII mobo), the data-source CD drive [potentially slow?] should be the… master; the writer (slave), sporadically would use the seamless-link, burn-proof or whatever brand-name is assigned to this function
By the way, only to add that connecting the master device to the outer connector of the 80-pin cable is a ‘best practice’ because in the case you disconnect the slave unit (e.g, for repair), the master [leaved temporarily ‘alone’] will not suffer electrical signaling problems if connected otherwise
Bye (& thanks, Furballi)