Quick Shot review 39:
Reviewer: Mark Cocquio
Corsair Voyager GT 4Gb USB Flash drive
Review date: 16th June 2007. Last modified 03-Dec-2011.
When can a thumb drive be faster than a hard drive?
When latency matters, that's when.
Consumer hard drives these days can move at least forty megabytes of data per second - twice that if you're accessing the outer tracks. The very fastest USB Flash memory drives struggle to make it to the lower figure, and are much slower for writes.
But Flash RAM has no moving parts, so you never have to wait for something to physically line up before an operation can start. That can make a big difference, for some jobs.
Apart from being an incredibly famous writer, my other main job at the moment is as a hapless Java programmer. I spend a couple of days a week maintaining a system worthy of mention on a certain Web site (no, I have not submitted any code to that site... yet).
The work is done on site, and they provide me with a laptop to use. Said laptop is a fairly high-end Dell; it has a fast CPU (for a laptop), and a good amount of RAM, but its chief downfall is that the hard drive in it is super slow.
In particular, when you do things that multitask heavily while relying on the filesystem (like when you compile a suite of J2EE apps and then restart a Weblogic server, which I do countless times a day), the laptop grinds away at its drive interminably.
Pretty soon my teeth started grinding as well. Even though I get paid by the hour, there's only so much inefficiency a geek can take, and besides, the hard drive grinding slowed down the other important apps I was using while waiting.
There had to be a better way. I didn't need particularly high access speeds, just low seek times.
This is pretty much the domain of solid state storage, and it was time for an experiment.
Enter today's most exciting object, the Corsair Voyager GT flash drive.
Corsair have been making RAM for ages, and they've been in the flash market for a while too. The Voyager line consists of tough, rubberised USB drives, which can supposedly stand a fair bit of punishment.
The Voyagers used to be the toughest in Corsair's range, but now there's the waterproof, aluminium "Survivor" range (cue "Eye of the Tiger").
By the way, that rubber covering makes the Voyagers pick up dust like there's no tomorrow. I took the photos when the drive was new, but it was already slightly grimy, and I am insufficiently dedicated to you readers to Photoshop it back to cleanliness.
The Voyager drive itself is quite elegant, with a small blue activity LED down the bottom. The LED has the drive capacity written on it, and the metal USB connector also has the capacity laser-etched onto it, along with the model number.
Even though they're not made of metal, the Voyager drives are still pretty tough - not that there's much to break in any Flash drive. The Voyager GT range looks exactly the same as the rest of the Voyager range apart from the colour of the rubber bits (red on the GTs, blue on the standard Voyagers), but the GTs also have higher read/write speed specifications. The GTs also only come in sizes of up to 8Gb; the standard Voyager range includes a 16Gb whopper.
All of the Voyagers come with a little lanyard and the usual short-ish USB extension lead, so that people with irritatingly located USB ports can use them without fumbling behind their desk.
The model I ended up getting is the 4Gb, high speed GT drive. All of the ridiculously bloated Java code and support files I need for work fit on it nicely, with plenty of room left over for some completely legal music.
So, what can one say about a flash drive? Well, appearance-wise, it's solid and looks as if it can take some punishment. Apparently you can run them over with cars and so on, and they'll keep on trucking (boom boom).
I'm not dedicated enough to test the drive's durability, seeing as I actually (God forbid) PAID for it. But I've been chucking it around for some months now, and it's never missed a beat.
That said, you wouldn't expect it to. All solid state memory should have excellent reliability, because it has no delicate moving parts. Furthermore, modern Flash drives have very high capacities, and therefore many more writes can happen before individual cells start to fail, because the drive firmware tries to distribute usage evenly.
My only slight concern about the design of the Voyager is that, as with most other Flash drives, you could easily lose the little end cap. But only when it's already removed. The cap is rubber, and fits well, so it doesn't slip off the drive in transit.
Performance-wise, Aus PC Market list the continuous read and write speeds of the GT as 34 and 21 Mb/second respectively (non-GT Voyagers are in the 19/13Mb/sec range - still nothing to be sneezed at).
As a result, the GT is compatible with Vista's ReadyBoost technology. I've had a play with it, and while I haven't noticed any earth-shattering improvements in performance (it seems that ReadyBoost doesn't do much of anything if you've got more than 512Mb of memory, which any computer running Windows Vista really ought to have), it hasn't hurt anything either.
You can use Flash memory to speed up other versions of Windows too; Dan comments on this, and flash memory in general, here.
So, how fast is the GT?
I did some benchmarking, and discovered something interesting about my home PC's USB controller - it appears to suck.
I've got an Intel 975-based motherboard which is getting a bit long in the tooth now, so I only managed to get maximum read and write speeds of 18 megabytes per second - a motherboard hardware bottleneck as it turns out, and a fair bit below Aus PC's speed claims.
The Dell laptop I use at work was much more respectable, with transfer rates of 31Mb/s and 23Mb/s for reads and writes, respectively. A second Dell desktop machine there gave similar results, vindicating (not to mention exceeding, in the case of writes) Aus PC's figures.
But does the Flash drive help with those disk-flogging Java compile times?
Well, the compile times are down, by a modest amount. Reading and writing zillions of small files is always going to be much slower than a device's potential top speed - and Flash memory has to erase blocks of cells before they can be written to, which is why its write speed is so much slower than its read - but the tiny seek time helps enormously. The big gains come when I read and write to the flash drive simultaneously; the tiny latency of flash memory does its thing nicely.
[On the subject of latency, many laptop drives have rather more of it than current desktop drives. That's because 5400RPM drives are still normal for laptops, rather than the 7200RPM units that've been standard equipment for desktop PCs for years now. The slower drives use less power and make less heat, and probably also wear out slower, which is all good news for laptop users. But the rotational latency of a 5400RPM drive, the average time needed for the disk to spin a particular location around so the heads can access it, is 1.33 times that of a 7200RPM drive. The lower rotational speed also generally moves less data per second under the heads, which reduces peak transfer rates. -Dan]
I wanted to try to quantify this effect, but my glorious employers have since moved me from the laptop to another machine, and so the original computer's no longer available for testing.
The next best thing was to create a read-write flog-fest on my home PC, and see how the Voyager performed.
First up, I filled a folder with about 800Mb of assorted files. I chucked in some big files, and a bunch of some small ones. Some were highly compressible, and some weren't.
Test number one simply involved compressing and decompressing this folder using NTFS file compression, and timing the operation.
Doing this basically triggers a simultaneous read and write process, over multiple files of various sizes. NTFS compression is not terribly CPU intensive, so most of the time involved in the process is taken up by the data transfers themselves.
The second test was much the same as the first. Using about 500Mb of data this time, I compressed the folder while at the same time creating a copy of it onto the same drive. Of course, this caused plenty of drive flogging.
I've got various hard drives to choose from, but I picked my C drive, because it's getting close to full and is reasonably fragmented at the moment. This meant that the drive head had to do a lot of seeking.
The results of the first test were quite interesting. The compression operation took 2:30 (two and a half minutes) on the fairly fragmented C drive, and 2:10 to decompress.
The Voyager took 3:20 to compress the files - typical slow Flash write performance - but decompressed them faster, at 1:53.
Running the second test (compress and copy) on the C drive took 2:05, and 4:08 on the Voyager.
As the saying goes, there are lies, damn lies, statistics, and then benchmarks, and the tests above are hardly rigorous or particularly scientific. There's not a lot you can strongly conclude here, except that indeed the lower latency of the flash drive does come in to play when lots of seeking is involved.
One also has to bear in mind that my C drive, fragmented as it is, still has a maximum read rate more than three times that of the Voyager. Despite this, the operations on the C drive were not three times faster at all. In fact, they were barely twice as fast at best, and even slower at worst.
[...and things would be even worse with the 5400RPM laptop drive. -Dan]
The old faithful HD Tach listed my C drive's latency at 13.8ms, versus the Voyager's at 1ms, and, unscientific as the tests above are, I could definitely see this in effect.
[That 1ms figure is probably approximately correct, by the way; it's not just the lowest figure HD Tach can report. Dedicated Flash RAM Solid State Drives can manage latency in the tenths of milliseconds, but cheap USB Flash storage devices all seem to float around the one-millisecond mark. This is still about a tenfold improvement over hard disk latency, but it's not the hundredfold improvement you can get from proper SSDs (including the Gigabyte i-RAM), or the roughly 5000-fold improvement you'll see from a software RAM disk. That last option isn't a very good one, since it eats system memory and will thus usually slow down the computer in general more than it speeds up your disk-flogging task in particular. It's still worth considering if you're running some stupid app that just can't be fixed in a better way, though. -Dan]
One of the nicer things about flash drives these days is that they're not very expensive.
The 4Gb Voyager GT will only set you back $AU115.50 delivered from Aus PC Market, provided you live here in Australia. The 8Gb version is $AU181.50 delivered - less than twice as much - although if you're after bulk storage, that much money will get you two hundred and fifty gigabytes in an external box. It'll just be a lot bulkier and more fragile.
[NOTE: This review went up half-way through 2007; Flash storage has gotten a lot cheaper, and bigger, since then. See the "Buy one!" section at the end of this review for current pricing.]
People elsewhere in the world can pick up a Voyager from just about any good online retailer (and no doubt many bad ones too). It's not as if they're rare.
If you need a decent amount of quick storage on the road then the Voyager GT definitely comes recommended from me, even though it didn't end up revolutionising my code-compilation experience.
Aus PC Market no longer stocks the 4Gb and 8Gb Voyager GTs.
They do still have the 16Gb Voyager, though, for only $AU110 delivered; Australian shoppers who'd like to order one can click here to do so. If you'd prefer a 32Gb Voyager for $AU198 delivered, click here!
A 16Gb Voyager GT will set Aussie shoppers back $130.90; click here to order that one.