Posted on 12/07/2009 11:16:55 AM PST by ShadowAce
USB 2.0 is so last millennium. No, seriously. We've had the specification since the year 2000, which is technically the previous millennium. Given how quickly the PC industry moves, though, it might as well have been a full millennium ago. Back in 2000, Intel was pushing Pentium III CPUs, 3dfx was still selling graphics processors, and Windows XP was a year away. Things have certainly changed since the so-called Hi-Speed USB spec was released, and its 480Mbps peak data rate has been grossly inadequate for quite some time now.
In the real world, you're lucky to push more than about 37MB/s through a USB 2.0 port. To give you an idea of just how slow that is, the latest crop of 500GB, 5,400-RPM notebook drives (which just happen to be popping up in all sorts of USB-attached external enclosures), can sustain transfer rates in the 67MB/s range. The 3.5" desktop drives that typically populate bulkier USB storage devices are even faster still. Heck, just about every hard drive we've ever reviewed is capable of saturating a second-gen USB link.
Making the most of the copious bandwidth available in USB 3.0 is going to be a little more challenging, though. The gen-three USB spec boasts a peak data rate of 5Gbpsroundabout 600MB/s, without taking overhead into account. Interestingly, this latest leap in USB bandwidth is proportionally smaller than the previous one. The original USB spec topped out at 12Mbps, making the jump to 480Mbps for USB 2.0 a forty-fold increase in available bandwidth. USB 3.0 only amounts to about a 10X increase over the prior standard.
Nevertheless, this new "SuperSpeed" spec should easily handle next-gen storage devices, even when you factor in overhead. According to the final specification, taking into account SuperSpeed USB's 8b/10b encoding, flow control, packet framing, and protocol overhead reduces effective throughput to a "realistic" 400MB/s for actual applications. Mechanical hard drives still have a long way to go before they can sustain transfer rates that saturate an old-school 150MB/s Serial ATA link, and even the fastest solid-state drives on the market aren't pushing data at much more than 200MB/s.
Despite the fact that the SuperSpeed standard was made available in late 2008, industry heavyweights haven't flocked to USB 3.0. The new spec isn't supported by any core-logic chipsets on the marketnot even Intel's P55 Express, which debuted just a few months ago. Rumor has it that Intel chipsets may not add USB 3.0 support until some time in 2011. AMD and Nvidia haven't announced plans for SuperSpeed-compliant core-logic products, either.
You don't have to wait on chipset providers to get in on SuperSpeed USB, though. NEC has crafted a USB 3.0 controller of its own: the D720200. The NEC chip offers a pair of third-generation USB ports, and it interfaces with the host system via a single gen-two PCI Express lane. Second-gen PCI Express lanes only offer 500MB/s of bandwidth in each direction, however, so the NEC controller won't be able to take full advantage of SuperSpeed's fat data pipe. The limited host interface bandwidth will have to be shared between both of the D720200's USB ports, as well.
Already, Gigabyte has revamped much of its enthusiast-oriented motherboard lineup to incorporate the NEC controller. Asus has added the chip to a number of its premium motherboards, too, and the firm has also whipped up a U3S6 add-in card that features the D720200 alongside Marvell's 88SE9123 6Gbps Serial ATA controller.
The Asus U3S6 situates is next-gen I/O chips behind a PLX bridge chip. The bridge splits the card's four-lane PCI Express interface evenly between the two I/O chips. Asus' decision to use a physical x4 interface means the card won't work in an x1 slot that isn't notched to accept longer cards. However, it does ensure the NEC and Marvell controllers will each get 500MB/s of bandwidth (the maximum each can use), even in a system that's saddled with first-gen PCIe or the P55 Express chipset's half-speed gen-two lanes.
Two USB 3.0 ports grace the U3S6's expansion slot cover, and a pair of 6Gbps SATA ports can be found on the card. We probed the Marvell controller's performance in great detail when we recently tested Seagate's SATA 6Gbps Barracuda XT hard drive. I encourage you to read that review for the skinny on this new chip. Today, our focus is on the U3S6's SuperSpeed USB performance.
Performance
So, just how fast is this USB 3.0 implementation? To find out, we needed an external drive with a SuperSpeed interface, which Asus conveniently sent along with the U3S6.
This Vantec-branded enclosure houses a Seagate Barracuda 7200.12 500GB hard drive whose peak transfer rates aren't anywhere close to 600MB/s. However, the 'cuda is still one of the fastest mechanical hard drives you can buy. You're going to have to get into multi-drive RAID enclosures to come close to saturating USB 3.0's available bandwidth with today's hard drives.
Note in the picture above that USB 3.0 uses a slightly different B connector type than USB 2.0. The plug is backward compatible with USB 2.0 cables, but a SuperSpeed B connector won't fit into a USB 2.0 B plug.
USB 2.0's A connector type, which is commonly found on motherboard port clusters and in notebooks, is the same in the new generation. You can plug USB 2.0 devices into USB 3.0 ports and vice versa. The SuperSpeed spec also calls for a dual-bus architecture that allows USB 3.0 and 2.0 devices to operate simultaneously at their optimal speeds.
To gauge performance, we plugged the U3S6 into an Asus P7P55D Premium motherboard's secondary PCIe x16 slot. That slot has an eight-lane PCIe gen-two connection to the system's Core i7-870 CPU. We connected the drive enclosure to the U3S6's USB 3.0 port and tested with HDTach. We then ran the same test with the drive connected to the motherboard's USB 2.0 port.
We also yanked the Barracuda hard drive from the enclosure and hooked it up to one of the board's Serial ATA ports to get an idea of the kind of performance an eSATA setup might yield.
USB 3.0 easily outpaces its predecessor, offering up to 3.5 times the throughput of the old Hi-Speed standard in HD Tach's sustained transfer rate tests. The Barracuda 7200.12 isn't capable of transfer rates much higher than 120MB/s, though. Our Serial ATA scores suggest that an eSATA implementation of this drive would be just as quick.
The SATA setup is even faster in the burst speed test, pushing 46MB/s more than SuperSpeed USB. I'm not sure what to make of this result, but it suggests that the NEC controller is hitting a bottleneck, at least when it comes to short burst transfers. Of course, you're still getting much higher burst rates than USB 2.0.
SuperSpeed's higher transfer rates don't cost much in the way of CPU time. HD Tach's margin of error for this test is +/- 2%, and we're well within that range.
Conclusions
USB 3.0 has been a long time coming, and it looks like widespread adoption will take longer still. That's a shame, because USB 2.0 is painfully slow considering the capabilities of today's mainstream storage devices. Even with our enclosure's run-of-the-mill hard drive, we saw more than a three-fold increase in transfer rates by jumping to SuperSpeed USB. That's huge.
Of course, we also saw a similar jump in performance when moving the hard drive over to its native Serial ATA interface. eSATA hasn't really caught on, I suspect because initial implementations required an auxiliary power cable. However, hybrid eSATA/USB ports are slowly populating motherboards and notebooks, and they may offer a better interim solution until USB 3.0 sees widespread adoption and more robust implementations. The market is hardly teeming with hybrid eSATA/USB storage devices, though.
I suppose I'd be more enthusiastic about SuperSpeed USB if this first implementation didn't feel a little half-baked. The NEC controller doesn't have the host interface bandwidth to properly take advantage of USB 3.0's full potential, and the burst transfer rates we observed suggest other bottlenecks may exist. The U3S6's Marvell SATA controller has a similarly inadequate PCIe interface and some troubling performance issues of its own, too.
But hey, Asus says the U3S6 will sell for only $30 when it hits North America, which should be soon. So it won't cost much to add USB 3.0 and 6Gbps SATA to an existing system, even if these first implementations of the new standards may not be as good as more mature solutions that surely lie over the horizon. At least with the U3S6, you won't have to buy a new motherboard in order to make the upgrade.
Technically, this is a protocol/technology review, rather than a single product that is based on the technology. I don't like to post product reviews, but I do like the technology advancements.
Perhaps I misunderstand what you are saying. If so I apologize. You do realize that Gigabit ethernet is much much slower than what we are talking here?
>Taking into account SuperSpeed USB’s 8b/10b encoding, flow control, packet framing, and protocol overhead reduces effective throughput to a “realistic” 400MB/s for actual applications.
What is interesting is that FireWire’s original spec was generally faster than USB 2.0’s when you compare the two on the actual throughput; now with USB 3.0, perhaps it would be best to compare it to FireWire’s third revision (3200) as the following article tries: http://www.techiesouls.com/2008/12/03/usb-30-vs-firewire-3200/
I am in full agreement with the commentator there who said:
“Ive been a fan of firewire for many years. It has never let me down.”
“I only have on problem with USB and that is they have always claimed speeds of 480mpbs while only delivering about 65% of that. Firewire on the other hand runs at about 97% of its claims. With that said Im guessing Firewire 3200 will blow USB 3.0 out of the water. Just my $.02”
Just a few days ago, someone posted pictures of a truly filthy interior of a desktop PC, and it was pointed out that there are two advances in technology that have lagged since the very first IBM PC.
The two are connected, because they both exist because of the same problem: cabinet heat. Because of cabinet heat, desktops use “gawky” and primitive electronic boards, instead of sealed cases, and use fans that expel hot air away from hot components. Which bring in copious amounts of dust.
If you can overcome the problem of cabinet heat, the entire computer could be modular and tidy and assembled in just a few minutes.
It would seem that the way to do this would be for the computer cabinet to be sealed, so no new external air and dust is introduced, and the ambient air temperature inside the box is low, even cold. In turn, cabinets would be vacuum insulated, to keep the inside cold.
Fans would still be used, but smaller fans on each module that generated heat.
Initially, when the computer was turned on, the heat exchanger would take a minute to cool down the ambient air. In turn, an external heat sink would get hot for a bit. But during normal use, the external heat sink would be a little warm, just exchanging heat as it was generated inside the box, warming the air that was then cooled.
>xternal USB disks have been clunky and really only good for occasional archiving. FireWire was going to revolutionize that segment, but it never took off.
I use Firewire for archiving, I like it a lot. It’s also nice that it doesn’t bog my systems down when I am doing so too, but that’s the synergistic effect of having SCSI hard-drives for my working-machine.
(If the CPU doesn’t NEED to be involved, why MAKE it be involved?)
Ooops. I saw the rates for USB 2.0
(I now have a total of 11 external USB 2.0 ports.)
HEY! Who's been messing with my pc?
I'm guilty of baaaaaaad maintenance (slapping myself).
Because they didn't really want you to buy the cheapest one. VGA connectors are cheap and don't require license fees like displayport and HDMI do.
Well, the cool think about the Ethernet ones is that they’re their own node on your network, so they end up working kind of like a server. The data on the drive is accessible to any machine with access to the network. There’s no one host machine that has to be left on for others to get at the data.
Well, I’d rather use FireWire, but the peripherals all come with USB. Drive enclosures, printers, digital cameras, cell phones, etc. To a lesser extent you have the same problem with laptops, which always have USB but rarely FireWire. That can be fixed with a PC Card, but what are you going to do about the peripherals?
Everything you suggested is available now for industrial system on a chip PCs, sealed cases with peltier exchanges, the cases themselves milled aluminum with 5 sides having radiator fins, external power supply, SSD, etc.
The system costs about $3000 for a single core pentium though.
I’m all for speed but USB 2.0 transfer rates are hardly objectionable in all but the most demanding or extreme circumstances.
With multiple-terabyte drives USB 3.0 will definitely be of use but like all things adoption will be mostly dictated by OEMs building it into their products.
I move a lot of multi-gigabyte video files around all the time. The increase in speed is substantial for me.
But I’d still prefer the next FireWire. The protocols were designed for high-speed from the ground-up, while USB was designed for low-speed, but had high-speed tacked on. It’s also not host-based, so it doesn’t require a computer and thus no CPU or main memory. It also carries more power (you can run a desktop hard drive without a power brick).
Micro USB (2.0) is already vexing.
I have a product from Netzero Netgear(?) that I plug a 500 gb USB drive into and then into my wireless router. All my computers in the house can access it.
When I first set it up I could access it from an IP address outside the network, but something isn’t working and I haven’t gotten around to getting it right to access it outside the network. Great idea, but kind of a stupid interface with the software.
Another drawback is the drive needs to be formatted in a proprietary formant for it to work. You can’t just plug a USB drive into it.
HDMI requires a fee, but DisplayPort doesn't. Apple saw the light and realized that going open standards with no fee was the best way to achieve wide DisplayPort adoption.
>Well, Id rather use FireWire, but the peripherals all come with USB.
I agree that you DO have to keep your eyes open.
>Drive enclosures,
I’ve seen plenty for FireWire.
>printers, digital cameras, cell phones, etc.
Printers and cell phones, nope.. I haven’t seen any of those that were FireWire enabled; digital cameras if they are video and somewhat higher quality are typically FireWire enabled (I believe they call it DV and often use the 4-pin Sony i.Link connector).
>To a lesser extent you have the same problem with laptops, which always have USB but rarely FireWire.
Really? I’ve seen a lot with the i.Link FireWire connection, my one and only laptop has an i.Link port.
>That can be fixed with a PC Card, but what are you going to do about the peripherals?
Good point. I guess I’ll just have to start up a business specializing in FireWire enabled devices... Sadly I don’t have the capital to do that.
http://www.superperolas.com/qual-foi-a-ultima-vez-que-voce-limpou-seu-computador/
Really dirty computer pics, though refreshingly SFW.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.