Months ago, the Storage Review took a look at the Quantum Fireball SE, one of the rare drives that was available in both SCSI and ATA versions. For whatever reason (probably SCSI overhead plus poor implementation in the drive’s control electronics) the ATA Fireball proved to be significantly faster than its SCSI brother. At that time, another pair of mechanically identical drives, the Seagate Medalist Pro 7200, was "just around the corner." In truth, the ATA ST39140A did arrive only a couple weeks afterwards. The SCSI version, however, did not ship until late June. We’ve finally gotten our hands on an ST39140W and eagerly installed it, seeing how it would do against the ATA drive.

Again, the Medalist Pro SCSI is mechanically identical to the ATA version that we reviewed several months ago. The drive features a quiet but hot fluid-bearing mechanism with a spindle speed of 7200rpm. The drive packs 2.3 gigabytes of data on each of its 4 platters. A 9.5 millisecond average seek time and a 512k buffer wrap up the package. The drive is protected by a consumer-grade 3-year warranty.

One thing that’s rather disappointing right off the bat is the price of the drive. Instead of being positioned as a "SCSI drive at an ATA price," the ST39140W’s pricing is more along the lines of a "budget SCSI" drive. Granted, a $550 price places is a little bit lower than entry-level units from some competitors, but that’s still a 50% price premium over the ST39140A. Further one must once again factor in the cost of adding a decent Ultra-Wide SCSI controller to the system, tacking on at least another $150 to the total price.

Before we proceed on to WinBench and Threadmark results, let’s address some concerns about the validity of comparing SCSI and ATA drives with synthetic benchmarks. A valid claim can be made that a SCSI drive, electronics properly implemented, will show substantial improvement over ATA drives on heavy and/or server type loads. Such uses are probably beyond the scope of the two benchmarks used here at the Storage Review. In such cases, however, SCSI’s reliability in addition to its performance becomes important. While Seagate makes some vague claims about increased reliability with the Medalist Pro’s fluid-bearing motors, the fact is that the drive is still backed with a three rather than SCSI-standard 5-year warranty. We should also note that fluid bearings have not yet made it into the company’s industrial-strength mainstay lines. The target application seems to be mainstream, single-user usage, where the results of these benchmarks may be compared and used as a guideline of relative performance between these two mechanically identical drives.

ZDBop’s Winbench 98 along with Adaptec’s Threadmark 2.0 were both run on the unit in Windows 95 OSR 2.1 and Windows NT Workstation 4.0. The drive was partitioned into a single volume of maximum size. The average of 5 trials is presented below.

The situation with the pair of Medalist Pro drives is similar to that of the Fireball SE’s: the ATA version is a bit faster. The gaps under Windows 95 were slight, ranging from 5-8% in WinBench. Under Windows NT, considered by many to be a "SCSI stronghold" of sorts, the gap actually widened, in one case up to 17%. Threadmark, another supposedly SCSI-friendly benchmark, placed the SCSI Medalist 8% behind that of the ATA drive under Windows 95. It’s only under Threadmark/WinNT that the two drives finished on par with each other.

I’d like to take an opportunity here to once again highlight relative CPU utilization between the two interfaces. There’s a common belief still rampant that SCSI devices burden the CPU less than their ATA counterparts. This is untrue. It may have been true that busmastering SCSI was a much more mature technology than busmastering ATA. Enabling DMA (busmastering) transfers for ATA drives takes more work (especially under Windows NT) than using DMA SCSI. Even so, once enabled, busmastering reduces CPU utilization of both interfaces equally. Neither interface has an advantage here.

Let’s take a quick moment to compare the Medalist Pro SCSI to a more direct competitor, say, the Quantum Viking II. Though a tad more expensive, the Viking is backed by a longer, 5-year warranty. It also provides a bit better performance, leading the Seagate drive by about 11% under WinBench/95 and 6-18% under WinBench/NT. Similar results occur with an IBM Ultrastar 9ES – Medalist Pro SCSI comparison.

The SCSI version of the Medalist Pro operates almost as hotly as its ATA brother, which means it’s hotter than both the Viking II and the Ultrastar 9ES. You definitely want a roomy, well ventilated case to house this one. Drive cooler recommended (as it is for most SCSI drives). Acoustically speaking, the drive is just as quiet as the ATA unit. The fluid bearings keep idle noise to a  minimum, while seeks are fairly unobtrusive. It’s a bit quieter than the Viking II, but slightly louder when seeking than the 9ES.

I (and Anand Shimpi, who reprinted the article on his site) received quite a bit of flack from the Fireball SE article. Some of the protests were quite lame ("UltraDMA has a maximum transfer rate of 33 MB/sec, Ultra SCSI, 20 MB/sec. Of course the ATA version will be faster!", while some where quite valid ("What if you wanted to use more than 2 hard drives?". Most people who wrote in regarding the article seemed to overlook this quote: "Its very difficult to draw further conclusions to apply to ATA vs SCSI in general, however. The only thing we can infer from the results is that the SCSI version of the Fireball SE is a poor-performing SCSI drive." Similar situation here. The Medalist Pro SCSI probably performed poorly not because SCSI is a snail-like interface, but because its implementation of SCSI is not up to snuff. Regarding the Medalist Pro strictly as a budget SCSI drive, my opinion is mixed. Though its slightly less expensive than the competition, you simply get faster performance and a longer warranty with the Viking II or Ultrastar 9ES. Unless you can get the Seagate for significantly less, you’d be better off with the Quantum or IBM.