limited to Ultra DMA 2 (UDMA/33)a note from author, automobus wrote:When (many years ago) I learned about PATA, I learned host controller can sense whether a UDMA-capable (80 conductors) or non-UDMA (only 40 wires) cable is in use. I understood, operating system and its PATA driver can limit mode speed to 33 Mo/s. I recently (2014) learned also drives can detect 80-wire cable. I want you to be equipped with this knowledge, when playing with other PATA systems. Happy hacking!
PATA Devices (parallel AT attachments) can detect and report 80-conductor (80 wires) cabling. This is reported to host in IDENTIFY DEVICE Word 93 bit 13. For SATA devices, word 93 is zero.
COPYRIGHTED CODE AND PARTS CONTAINED HEREIN
ThinkPads do not contain the cable which can be found in conventional AT PCs. Crowd favourites X22 and T23 are both engineered for UDMA. But programming host to permanently report 'wired for UDMA' was not good enough: IBM redefined controller cable detection used in ThinkPad. Under subject ThinkPad firmware, PATA controller parrots what drive reports.
SATA does not report parallel cable. At power-on, if IDENTIFY DEVICE Word 93 is zero, then ThinkPad firmware configures PATA controller to report '80 conductor cable not detected'. Every sane operating system will then limit performance to Ultra DMA 2 (UDMA2 33 Mo/s).
machines with this bugfeature
X22 genus: X22 X23 X24 1DET70WW (1.32)
X30 (Tested by kobalt.)
(possibly VAIO TZ (I do not have patience to research), Latitude D)
what your drive thinks of its cable
For 80-conductor, output from
hdparm -I ;includes "CBLID- above Vih" in section "HW reset results".
Centrino Carmel ICH4
T40 T41 T42 R50 R51
I suspect, firmware releases from earlier three years than 3.20, contained a routine/program, similar to that in earlier ThinkPads. I consider this a performance bug.
In effort to prove my suspicion, I ask:
If you use BIOS revision between 3.14 and 3.20 inclusive, please try SATA BEFORE flashing newer. If you use BIOS older than 3.20, please flash 3.20 and then try SATA BEFORE flashing final 3.23.
"1.05 (1RET34WW) 2003-05-15" doing it wrong.
3.13 configures silly. 3.21 configures optimum.
3.23 also works
X40 BIOS versions 1.55 and 2.08 do not exhibit this bug. (Tested by kobalt.)
mounting screw holes
LyCOM ST-173 is imported and whitelabelled or rebranded by several companies. Sold by Addonics has sixteen screw holes, in side & bottom, modern & pre-1998 locations. Sold by Aleratec has eight mounting holes, only side and bottom modern locations.
error in manufacturing
Anonymous stuff found in feeBay is often too cheap to have passed quality control. In ThinkWiki article "How to put SATA in old ThinkPads" is photos of an adapter board which did not work until a blob of solder was applied to fix it.
I surmise, delay is inherent to ATA-ATA translation bridges. It seems some bridges delay longer than others. If some drives need more time (than other drives) to carry-out commands, then ATA translation delay can cause timeout errors. Sonoma Alviso T43 contains PATA to SATA bridge, which is problematic with some drives. (Other ThinkPads based on Sonoma:Alviso are R52 and X41.) Ultrabay Slim SATA HDD Adapter contains hidden bridge to PATA, which is problematic with some drives. Lenovo X301 contains Marvel 88SA8052 SATA to PATA bridge, which is problematic with some 2nd HDD assembly.
ATA command trim (which is an overhyped solution to an artificial problem) commonly takes a long time to complete. With no ATA-ATA translation bridge, TRIM operation can timeout (performed by SATA drive wired to SATA host). With one ATA-ATA translation bridge, TRIM operation often timeouts. With two ATA-ATA translation bridges, TRIM operation is very likely to timeout.