Quote:
Originally Posted by Koko
... The gyroscope was 'off' on the KDX (KT has none). ...
|
Actually, neither device as a gyroscope. The DX and K5 both have an accelerometer, which can be used to control screen orientation.
If you search Mobileread for "KT" you will find that the vast majority of hits are for the "Kobo Touch". In this forum, we have standardized on the "K5" label for the Kindle Touch, for both our forum prefixes and for the index wikis.
Quote:
Originally Posted by Koko
... But the devices where not file defragmented. I now post for the study checked the fragmentation on the devices it was 3% for the KT and 26% on the KDX. ...
|
Your FAT format fragmentation measurements do not show the REAL write wear-levelling fragmentation going on down at the flash media erase block level. Crossing erase block boundaries affects device timing (and search speed), even while reading. Over time, erasing and re-writing content on the USB drive would increasingly fragment the PHYSICAL storage erase blocks. On SSD devices, ATA secure erase and TRIM support can help "defrag" the wear-leveled media. But on a kindle, the only way to compare the speed of two devices fairly is when they are "factory fresh" (and not just reformatted).
However, as previously shown, the search result speeds seem to be primarily INDEPENDENT of hardware differences, including flash erase block fragmentation, and whether or not the K5 accelerometer was turned off.
I doubt that battery power consumption from the accelerometer (or even from the more power-hungry IR touchscreen) would affect search speed anyway, unless the battery were discharged enough to cause the firmware to change the CPU speed governor settings. Needless to say, these "modern methods" do no rely on SQL databases.
EDIT: So, I am staying with my previous conclusion that search speed is primarily limited by the choice of algorithm used to index the device content, and secondarily by the eMMC (USB drive) speed. There are modern indexing methods that can search "instantly" with virtually no overhead (other than the cost of initial indexing), with search times measured in microseconds instead of seconds, even when searching Petabytes of "already indexed" data.