Based on the info I've read at Andre's site and the SatCP tutorial, it's my understanding that EAC secure mode works in the following way regarding caching:
In secure mode, multiple attempts are made to read the data, and multiple data values from these attempts must agree before EAC accepts the data and moves on to the next data. For this to work, each attempt must result in a physical re-read of the disk's data. When caching is present, the multiple read attempts end up just returning what was already in the cache multiple times. Thus it's possible to get matching data even if the original data was in error. This basically defeats the purpose of EAC secure mode. The purpose of the EAC "Drive caches audio data" setting is not to disable or enable the cache (which apparently only the firmware can do), but rather to tell EAC that it must clear the drive's cache after each read, such that the data comparison occurs with data obtained from a physical re-read of the disk's data each time.
So there's four total combinations of drive caching and EAC caching configurations as follows.
1) Drive caches data, EAC is told that the drive does so. Cache is flushed after each read, slowing down the process. Result: degraded speed but full accuracy
2) Drive caches data, EAC is told that it does not. Data re-reads for error checking simply compare the same data from the cache with itself. Result: maximum speed with degraded accuracy.
3) Drive does not cache data but EAC is told that it does. Extra commands are sent to flush the cache, but there is no cache, so all this does is slow down the rip. Result: degraded speed but full accuracy.
4) Drive does not cache data and EAC is told that it does not cache. Each re-read command results in a physical re-read of the data. No extraneous cache flush commands. Result: maximum speed with maximum accuracy.
In short, having a drive that does caching on audio extraction but telling EAC that it does not, maximizes ripping speed, but defeats the purpose of EAC secure mode. People use EAC secure mode for the most accurate possible data. My opinion is that if the drive caches audio data, EAC should be configured to tell it that, so that the speed benchmarks are for the condition of highest data accuracy. Since maximum accuracy is the reason for using EAC secure mode in the first place, it seems that the fairest comparison is the configuration that gives the highest accuracy.
That said, I did some benchmarks of my 5239s by telling EAC that it does not cache. This only gave me a minor improvement in ripping speed. It still doesn't explain the huge difference in EAC secure mode benchmarks between the 5239s and the 5238s. I'm starting to think that these drives may in fact be very different from each other despite the similarity in part numbers. Aargh!