Recently, it seems that two premiere SCSI drive manufacturers, Seagate and IBM, have abandoned their "three-tier" product line for a more consolidated dual-level offering. A former distinction between "entry-level" and "enterprise-class" drives has been blurred, with a single drive from both companies representing their 7200rpm offerings. A 10k drive (of course) rounds out the high end from each company.
Quantum went through such a change last year, dropping its rather lethargic Fireball SCSI line. The manufacturer presented two SCSI offerings, the entry-level Viking II and the enterprise-class Atlas III. Interestingly, the Atlas line never fared well in StorageReview.com's benchmark and usage tests, supposedly being optimized for server access patterns that didn't shine through. Thus it was with great interest we observed Quantum's announcement of it's 1999 SCSI drive line: The entry-level (presumably a Viking III) and enterprise class drives would be consolidated into a single 7200rpm offering, the Atlas IV (gulp). Even more interesting, Quantum announced that they'd jump into the 10k rpm battle with the Atlas 10k.
The flagship Atlas 10k is an 18.2 gigabyte drive featuring six 3 GB/platter disks, on par with the reigning champ, the Seagate Cheetah 18LP. The Atlas' rated seek time is just a tad more svelte, at 5.0 milliseconds, just a hair under the Cheetah's 5.2 ms. The drive features a 2 megabyte buffer, doubling that of Seagate's offering. A five year warranty protects the disk.
The Atlas 10k is the first drive we've tested that features the new Ultra160/m SCSI interface. A doubling of the maximum transfer rate will undoubtedly be the most ballyhooed facet of the spec; and who knows, if you somehow find yourself with 14 Atlas 10k drives running off of a single Ultra160/m controller, you may run into situations where the 80 MB/sec of Ultra2 simply doesn't cut it. For most uses, however, Ultra160/m compatibility at this stage is unnecessary.
In usage with our testbed's Adaptec 2940U2W SCSI host adapter, I ran into problems getting the drive detected on the LVD (U2W) segment of the controller. The drive would appear just fine when attached to the UW segment. After much hair pulling, a Quantum applications engineer determined that the 2940U2W had to be "hard terminated." In other words, setting host adapter termination in the BIOS (sufficient for every LVD drive thus far) was not enough: a jumper physically located just beneath the LVD connector had to be shunted. Isn't new technology great? ATA-66 and Ultra160/m both arrive, both (supposedly) completely backwards compatible, yet these problems crop up. Oh well, enough griping, let's take a look at some figures: