Analysis: Perpendicular Versus Horizontal Drive Technology

Hard drives haven't really undergone any truly revolutionary changes in decades, possibly since the move to integrated drive electronics (IDE). Sure, they've received upgrades to areal density so that current devices can hold far more than even their close ancestors of a few years ago and gobs more than the 10MB of the top of the line drives from the 1980s. We've also seen them switch from parallel ATA (PATA) to serial ATA (SATA) and watched as the industry endowed them with faster data access speeds, longer life spans, and enhanced reliability. Still, these have been more an evolution than revolution.

The one upgrade that has been revolutionary is the move from a horizontal to a perpendicular recording scheme. I can hear the head scratching in the back rows

Picture a dinner plate covered in a thin layer of lobster sauce. You've eaten the shrimp already so all that's left is the white rice sitting the sauce and, by some cosmic intervention, all of the grains are lying flat (horizontally!) on the plate, end to end, a spiral pattern that stretches from the outer rim of the plate to its center. (Some intervention, huh?)

The plate is analogous to a hard drive's substrate, the lobster sauce to the slurry that's used on that substrate to hold the ferric particles (the rice, in this case) in place. Hard disk manufacturers spin the substrate to get the particles to align as noted. We relied on cosmic intervention to keep you from being ejected from the restaurant.

You can employ a variety of techniques to pack more and more rice onto the dish in that manner but, at some point, the length of the rice (or the ferric particle) becomes the real limiting factor. Bottom line: you can only fit ten 1-micron wide particles in a 10-micron space.But what if you could get the cosmos to intervene once again and, this time, when it's packing your rice on your lobster sauce-coated plate, it stands up the grains (perpendicularly!) instead? The particles are much thinner than they are wide so many more of them could fit in the same space! Effectively, this re-alignment would increase your plate's areal density. (Proven, of course, by the fact that your waiter would have to bring you more rice to fill the same space.)

That's the crux of it. Naturally there are technical details, urban legends, and nuances of conduct that really define the technology. Feel free to do the research. This isn't the time or the place for that. We're gathered here today (after polishing off a reasonable meal of shrimp with lobster sauce) to determine if there's any speed difference between the two different technologies, horizontal and perpendicular. That's why they pay me the big roll of quarters.

The test system we'll use began life with an ASUS P5WD2-E Premium motherboard stuffed with a Pentium 955 Extreme Edition processor (cranked up to 4GHz) and 1GB of memory. The motherboard supports RAID and that's the connection point for the hard drives.

Both pairs are from Seagate. The first was two 300GB ST3300622AS horizontal drives. The second pair was composed of two 320GB ST3320620AS perpendicular drives. All four are SATA 3.0GB with 16MB of cache. The operating system for the computer is Windows XP Media Center Edition. The test used is called HD Tach version 3.0.1.0 from Simpli Software. No adult beverages were consumed during the tests.

Logic would lead you to assume that moving from one small bit to another should be faster than moving from one larger bit to another. There are two problems with yhat: A hard drive is a type of hybrid device. It's both mechanical and electronic and the same rules don't apply to each of its natures. The second is that you should never really assume anything, especially when you're going through an operating system and an interface.

According to HD Tach, the horizontal RAID configuration had a sequential read speed of 99.3MB/s and a burst speed of 317MB/s. The perpendicular RAID configuration had a 118.7MB/s sequential read speed and a burst speed of 227MB/s. Now that's confusing, isn't it? How can a higher burst speed produce a lower access rate?

First let's tackle the burst speed. Traditionally, that's a measurement of the data transfer speed between the drive and the cache on the electronics installed on the drive. Ignoring the interface and any delays the bus into the computer might generate, theoretically, a high burst speed for long sequential data reads as it keeps the information flowing. You would suspect, of course, that drives such as these, with 16MB caches, would do well.

According to HD Tach, however, the slightly lower score for the perpendicular drives means they're better at random access data acquisition. This too makes some sense as, all else being equal on the mechanical side, reading a smaller segments of bits and then moving on to a new small segment should (emphasis on the possibility factor) be slightly faster.

That's fine for benchmarking and for all of you out there who make your living running benchmarks I'm sure you're pleased. However, to bring a reality check to the table, I dusted off my copy of Ulead VideoStudio 10 Plus, dissected an hour's worth of captured TV video to get rid of the commercials, and then stitched it back together into a roughly 43-minute, commercial-free, clip.

Wouldn't you know it? There was really no discernable difference between the time it took to render the completed video to the perpendicular RAID array than it did to the horizontal RAID array! Zip, zero, nada %#151; despite the benchmark results.How could that be? Surely benchmarks don't lie! Might there be some symbiotic relationship in this computer that makes the destination drives wait for output pulled from a source drive and then heavily processed before it's written to disk? Could the big secret be that, for some systems, "fast" is as fast as it gets and paying for "ludicrous speed" is really just a waste of money? Yeah.

Benchmarks don't lie. Dealing only with the drives in question and not the system as a whole, the perpendicular array is faster than the horizontal array under certain specific conditions and vice versa under others. That's life in the city. But when you get to real life, where benchmarks aren't the applications you run, neither of these pairs of SATA drives necessarily has any advantage over the other and much of whatever they might have depends on the what's going on with the rest of your computer..

"For example, I also tested the poor 250GB PATA drive I used as the source drive. It ran up an amazingly unimpressive 92.7MB/s burst rate and an equally demoralizing 49.9MB/s average read speed. That's just about half the throughput of either the perpendicular or horizontal drives and probably not enough to feed data at a speed that could keep them running at their best effort. That makes it a limiting factor and wastes any extra money you spend on the SATA drives. So if you're trying to build a fast computer, it might be wise to avoid PATA drives at this point. The world has just about passed them by.

The semi-weekly Best Bits column surveys the soft underbelly of PC hardware from the unique viewpoint of Bill O'Brien, who can be blamed for more than 2,000 articles on computers and technology topics. With his writing partner, Alice Hill, Bill co-authored "The Hard Edge," the longest-running (1992 to 2004) technology column penned by a techno duo. For more, go to www.technudge.com.