News & Trends: USB 2.0's Real Deal
PC users love their machines' dependable Universal Serial Bus ports, but speedy peripherals such as external hard drives can turn USB's 1.5-megabytes-per-second (12-megabits-per-second) transfer rate into a performance bottleneck. Enter USB 2.0, or Hi-Speed USB, which promises the versatility of its predecessor, compatibility with today's USB products, and a transfer rate of up to 60 MBps (480 mbps) with USB 2.0 peripherals--that's 40 times faster than USB 1.1.
Can USB 2.0 live up to the hype? The first products that we tested (see October 2001's "
The good news: Although some vendors of peripherals recommend specific USB 2.0 cards, tests conducted by the PC World Test Center showed solid compatibility among Hi-Speed USB products. However, performance didn't match the widely advertised claims. The best performance by a USB 2.0 peripheral, achieved while moving data between a PC and an external hard drive, accelerated the transfer rate by 12.6 times. We saw substantially lower jumps with other USB 2.0 devices--an external CD-RW drive and a scanner--because of their lower top-end performance capabilities.
Still, Hi-Speed USB is far from being a failure. You'd be hard-pressed to obtain the enhancements that it delivers through any other PC upgrade, aside from adding an IEEE 1394 interface--an older, competing standard that promises 50 MBps (400 mbps) transfers. The Hi-Speed USB boost is even more impressive given that a USB 2.0 PCI card and cable should cost less than $100 together.
We looked at five Hi-Speed USB PCI cards: Adaptec's USB2connect 3100LP ($49), Belkin's USB 2.0 F5U220 ($69), Keyspan's USB 2.0 U2PCI-5 ($59), Orange Micro's OrangeUSB 2.0 Hi-Speed PCI ($69), and SIIG's USB 2.0 5-Port PCI ($40); all of the prices identified here are list. We used the vendor's drivers for each card.
We tested all five USB 2.0 PCI cards using Maxtor's $200 Personal Storage 3000LE, a 40GB, 5400-rpm external hard drive; Epson's $400 Perfection 2450, a flatbed photo scanner; and TDK's $200 241040UE VeloCD, a 24X/10X/40X external CD-RW drive.
We ran our tests on an IBM NetVista PC configured with a 1.4-GHz Intel Pentium 4 processor, 256MB of memory, a 60GB internal hard drive, and Windows XP Professional. We installed and tested each card in the computer separately, running each of the three peripherals before moving on to the next card. To establish comparable scores for USB 1.1, we ran tests via the same PC's integrated USB 1.1 ports (see chart, page 18).
First things first: Every card and peripheral combination worked. Of course, that's what should happen with products using any standard, but some vendors of early Hi-Speed USB-compliant peripherals recommended specific PCI cards for use with their products, which raised questions about interoperability.
For example, when Sony shipped its slender CD-RW/DVD-ROM combination drive (model CRX85U/A2) last fall, it recommended Adaptec's card after "extensive compatibility testing." Today Sony still recommends the Adaptec card, but a company spokesperson observed that the drive should work with any card that carries the Hi-Speed USB logo.
The packaging for all the PCI cards we tested carried that logo, meaning that the product had passed a compliance test established by the USB Implementers Forum, the organization behind the standard.
Not only did all five cards work with all peripherals, but their performance was strikingly similar: In most tests, the variance was a percentage point or less. We attribute this result largely to each card's using the same NEC host controller chip.
Shortly after we completed testing for this story, Microsoft released final USB 2.0 Windows XP drivers for download. The company says it will also soon release drivers for Windows 2000, but it has no such plans for its Windows Me and Windows 98 OSs.
The maxtor external hard drive realized the largest increase in transfer rate of the three peripherals we tested with USB 2.0 cards. The average time all five cards took to complete our file-copying task using the drive was 58 seconds, versus 12 minutes, 13 seconds with USB 1.1--12.6 times faster. The five-card average on our Photoshop test was 4 minutes, 24 seconds--8.5 times faster than the 37 minutes, 19 seconds that USB 1.1 took. Even so, our anecdotal analysis indicates that a 5400-rpm internal hard drive with a standard UDMA/100 bus still performs noticeably faster than our USB 2.0 drive.
Don't blame Maxtor's external drive for the slower performance--it's capable of sustained transfer rates of up to 46.7 MBps (about 374 mbps). That's slower than Hi-Speed USB's theoretical maximum, but much faster than the effective transfer rate of 11.2 MBps (90 mbps) we achieved in our file-copying task.
For answers, we turned to an executive at the USB Implementers Forum. First, according to Jason Ziller, chairman of the USB-IF and an Intel technology initiatives manager, at least 10 to 15 percent of the stated 60 MBps (480 mbps) of Hi-Speed USB goes to overhead--the communication protocol between the card and the peripheral. Overhead is a component of all connectivity standards.
In addition, Ziller said, our lower-than-expected scores may be the result of an operating system and/or a host controller chip that is not yet fully optimized for maximum performance. Transfer speeds should improve as chip vendors and driver writers fine-tune their products.
Our tests using TDK's CD-RW drive also revealed noticeable performance improvements with Hi-Speed USB--though the limitations of even a very fast CD-RW drive keep it from speeding up as much as a hard drive. In the digital audio extraction test, our five Hi-Speed USB boards completed the task in an average of 98 seconds, versus 6 minutes, 32 seconds for USB 1.1--a fourfold improvement. In our write-on-the-fly test, Hi-Speed USB offered a fivefold performance boost, which amounts to about 2.7 MBps (21 mbps). The drive's 24X write speed is theoretically capable of achieving a maximum transfer rate of 3.6 MBps (28.8 mbps); its 40X read speed equals a maximum of 6 MBps (48 mbps). Our test drive's scores are on a par with comparable internal CD-RW drives we've tested in the past. In short, with USB 2.0 you no longer sacrifice performance for the convenience of an external drive.
The transfer-rate limitations of a scanner are even more noticeable. The Epson scanner ran 1.7 times faster with USB 2.0 than with USB 1.1 on our 300-dpi image test, and it showed a twofold improvement on the 1600-dpi image. That's in line with what Epson engineers say they expected, since the memory buffer of most scanners is simply too small to take full advantage of Hi-Speed USB. Even so, the higher-resolution image scan required only 6 minutes, 44 seconds to complete using USB 2.0--about 7 minutes faster than using USB 1.1. That difference can really add up if you do a lot of scanning.
Chances are your next PC will come equipped with Hi-Speed USB. In January, Gateway became the first major PC vendor to offer systems with Hi-Speed USB included on the motherboard. The USB-IF's Ziller says that he expects other system vendors to follow suit in the coming months.
But the big push will come later this year, when chip set manufacturers begin adding USB 2.0 hardware to their products. Ziller says that Intel will launch a new chip set with built-in Hi-Speed USB by this summer. Chip set vendors Silicon Integrated Systems and Via Technologies say they plan to add Hi-Speed USB to their product lines in the next few months.
In the meantime, if you're in the market for a new scanner, an external CD-RW drive, an external hard drive, or any other peripheral that can at least partly use USB 2.0's faster transfer rates, we'd recommend upgrading. After all, any upgrade that painlessly improves your PC's performance by anywhere from two to five times or even more--and costs less than $100--is well worth considering.