IBM MCA HARD DRIVES It is better to buy appropriate drives for the purpose: 50-pin types like the DORS for the desktops with onboard SCSI or "Tribble" / "Spock" / "Spock Prime", DCAS, DCHS in 68-pin for the Fast / Wide arrangement in Server 85 and 95 Non-Array and DFHS, DCHS, DDRS in 68-pin version for the Server 95 Array / Server 3xx, 5xx "Hot Swap" bays. These drives were designed to fully support the Hot Swap function - which is not the case on the "workstation drives" DCAS and DORS. DFHS drives Use them externally in an enclosure with sufficient room and additional fans with enough airflow in lenght over the drives. IBM made pretty bad first series DFHS. They were hot, loud and had problems with the internal controller microcode as well as with the surface coating which leads to "sudden death". There were "active cooled" 5.25" bays - which have two of these noisy "trouser button" fans ... which (in theory) should keep the drives cool enough. Practise shows, that these thingies are only nerving loud and fail occasionally (like CPU fans - they are similarly lousy). The DFHS are -like the Quantum Atlas II and III- intended for servers which supply active cooling. Parts of the DFHS heat up to 95°C .... The older DFHS series 1 (the "Fat Ones") get really hot - and are not recommended. Towards the end of the series the drives get better. DFRS Keep away from DFRS - they are "refurbished" (hence -R-) DFHS that had been sent back for repair and returned to the spares cycle. These "refreshed" drives have a large black stripe on the barcode label with white "RE" in it.atop. They are second choice.I had three of them - all are dead in the meantime. DCHS aka "Ultrastar 2XP" are related to the DFHS ... same mechanical chassis but a reworked electronics board. The -04U is the 4.55GB Ultra-Wide (or: Fast20-Wide as IBM named them at that time) version. It is a single-ended drive and should be used with active cooling or lots of airflow over the drive - like for the DFHS. They have EPRML (enhanced PRML - the DFHS introduced PRML before), ECC-on-the-fly and lesser data overhead ... therfore a tad higher capacity than the DFHS. The bigger capacity however forbids mixing them in arrays with "industry standard" drives, which have a size of 4.32GB only (the raster goes 1.08, 2.16, 4.32 ...). The DCHS is a bit faster (10.3 inner / 15.4 outer band vs. 9.59 / 12.58 MB/s media data rate on the DFHS). Average seek time 7.5ms and latency of 4.17ms is a bit better than the DFHS (7.8ms). The drive is not known for excessive malfunctions - has been produced for a year only and superseeded by the DNES/DGHS and DRHS series. On the WT/WB problem: don't do it. The drive cache should not be set to WB if you don't have a buffered Raid controller attached. There is a risk to write corrupted data-blocks if the adapter shuts down the drive before flushing the drive cache ... This feature can be permanently enabled with a particular software only. Therefore the value returns to the default WT after any power on. DCHS are okay. They are technically an intermediate between the DFHS and 0664. DDRS 7.200 rpm version of the DCAS. They're 7.200rpm - but don't make much noise. The DDRS finally was the 3rd generation drive (after DFHS and DCHS) where they learned to handle the problems. I have 7 of them in my Server 520 (DDRS-34560 UW, some LVD) along with 2 x DCHS and 2 x Quantum Atlas. The DDRS is available as 50, 68 and 80 pin version. For the Server 95A "non Array" with F/W "Corvette" I would pick the 68-pin version and tailorize an appropriate SCSI 68-pin cable. The DDRS superseeded the DCHS, which superseeded the DFHS. The follow-up to the DDRS was the DNES, which is a good 7.200rpm drive and viable at ebay in masses. Shortlived because 10.000 rpm DRVS came out. The DNES ran parallel with the DRVS for some time. DCAS Not a bad drive - but a 5.400 rpm type. The DCAS was designed for desktops / workstations rather than servers. But they were suitable for smaller servers as well - especially when they were "power-wise and thermically challenged" (mean: small power supply and bad internal ventilation). The DCAS runs pretty cool - only topped by its successor DORS. The DCAS is a nice, fast and low-noise drive. >How reliable are they? Depends on. Under normal circumstances the chance of running into a dead DCAS is low. *If* it had been shipped and handled with care. Electronics defects are rare - accumulated media defects kill it occasionally. I have 1 DCAS and 2 DORS with media defects - and fully functionally boards. If a DCAS runs through the Adaptec "Surface Scan" *without* needing to relocate a bad sector it will run for quite some time. (Same with the DORS and DDRS). If the drive is poorly mounted and dropped during running it will be dead within a short time. Once the heads damage the platter surface you will run into accumulated defects (due to spilled particles inside the platter cage) and you won't come far with it. If you have to relocate sectors during surface scan: mark the drive at the front with a red dot. You can bet that it dies on you within month latest if used in a server with long on-times and high drive useage. Especially in a RAID-5 systems that operates with interleaved disk accesses. Any time the head scratches over the defective spots on the platter more particles get loosened - and finally the drive come unuseable. Paper weight. DORS 5.400 rpm rated 5V/300mA and 12V/200mA, which makes a total of lousy 4Watts .....It is my favourite desktop drive in old PS/2s (56 / 57, 76 / 77) *because of* its low power consumption. DPES "Pegasus" drive is a 5.400 rpm drive designed for desktops. It has about 50% lower current draw and -therefore- runs significantly cooler than the DCHS. With the introduction of the DSAS and DPES series IBM returned to other conceptions of the R/W-amplifiers and different screening of the drive. These are rock-solid general purpose drives, which need no special treatment. DNES DSAS "Satsumas" (i.e. DSAS) drives are a) not hot-swap specified b) far too slow (4500rpm, seek 12-14ms, 96KB buffer) and c) the Id. pins are at the *wrong* end of the pin block to easily use the standard Id. cable, not to mention being halfway back down the drive's length (neither a "front* not a "rear" position, really).