USB 3.0 arriving with speed to burn
Summary
Intel demonstrated a working version of USB 3.0 at CES last week and showed why it will make eSATA and FireWire obsolete. It promises to be 10 times faster than the current USB 2.0.
Topics
Intel demonstrated a working version of USB 3.0 at the Consumer Electronics Show (CES) in Las Vegas last week. Here's why it will make eSATA and FireWire obsolete.
When USB 3.0 is expected to hit the market in early 2010, it will have been 10 years since the now ubiquitous USB 2.0 was introduced (April 2000). The current USB 2.0 specification runs at a theoretical maximum speed of 480Mbps, and can supply power.
According to the USB Implementers Forum, there were 2 billion USB 2.0 devices shipped in 2006 (one for every three people in the world), and the install base was 6 billion (almost one for every person in the world). In November 2007, the USB Implementers forum announced the USB 3.0 specifications, and Intel officially demonstrated the technology at CES 2009.
Now, the juice: USB 3.0 promises a theoretical maximum rate of 5Gbps, meaning it's 10 times faster than USB 2.0. USB 3.0 is also full duplex, meaning it can upload and download simultaneously (it's bi-directional); USB 2.0 is only half duplex.
Put side by side with eSATA and FireWire 800, USB 3.0 is far superior. eSATA, an external connection that runs at the same speed as the internal SATA 1.0 bus, has a maximum theoretical of 3Gbps. This makes USB 3.0 faster than eSATA and about six times faster than FireWire 800 (full duplex at 800Mbps).
USB 3.0 also provides another advantage; while eSATA is faster than FireWire 800, unlike FireWire it cannot supply power. USB 3.0 has the advantage of being faster than both, even while supplying power.
Finally, USB 3.0 has improved power management, meaning that devices can move into idle, suspend and sleep states. This potentially means more battery life out of laptops and other battery-based USB-supporting devices like cameras and mobile phones.
Of course, there are other factors to consider; the FireWire 3200 standard is also in the works and promises to allow 3.2GHz speeds on existing FireWire 800 hardware. USB 2.0 generally doesn't meet its theoretical maximum throughput, due to its dependence on hardware and software configuration, where FireWire gets much closer.
It's hard to say whether USB 3.0's updated architecture will still use more CPU time than FireWire does.
But in the age of powerful hardware (can anyone say "3.2GHz, quad-core CPUs"?), all of this means that FireWire is still not going to match USB 3.0's theoretical maximum of 5Gbps.
The ultimate signal that this war has already been won is Apple's recent decision to ditch FireWire from its consumer line in favor of USB. Previously, Cupertino had been one of FireWire's greatest advocates. And surely the company will be one of the first to adopt USB 3.0.
All in all, we can't wait for motherboard manufacturers like Gigabyte and Asus to start supporting the technology and mainstream PC builders like Dell to start integrating it into their products. Bring on the speed.
Talkback Most Recent of 17 Talkback(s)
-
sweet!
nt
Been_Done_Before13th Jan 2009 -
why it took so long?
USB3 has been in works for many years.
Initially the big selling point will be the backward compatibility since I expect a premium for its implementation in computers and other electronics.
I'll put my spending on hold untill usb3 becomes more mainstream and cheaper.
Linux Geek13th Jan 2009 -
Spelling error...
Eighth paragraph down, "... works and promises to allow
3.2GHz" should be 3.2Gbps.
olePigeon13th Jan 2009 -
USB 3.0 doesn't make FireWire obsolete...
USB 3.0 doesn't make Firewire obsolete, it's just pushed
Firewire into the niche pro-consumer market.
Firewire is a superior connectivity technology compared to
USB for external devices such as HD camcorders, pro audio
equipment, and external data storage. USB can not
compete with FireWire simply by virtue of it being a
host/slave paradigm and relying on the CPU.
The important feature about Firewire is its sustainable
speeds and non reliance on the host CPU. If you're
doing CPU intensive tasks, or, need guaranteed
throughput, you have to use Firewire. That is why pro
audio equipment uses Firewire, and why external scratch
storage for video and audio uses Firewire. If you want to
transfer RAW HD video from an HD camcorder, you're
going to need Firewire.
eSATA I never understood. eSATA isn't meant for hot-swapping since it has to reassign hardware addresses
every time you plug and unplug the device. You also can't
daisy-chain eSATA like you can with Firewire. It also isn't
bus powered like Firewire, forcing you to carry extra cables
around. eSATA is intended for permanent external
storage.
Also, speaking of the pro consumer market, IEEE-1394c
spec makes it so you can do Firewire over Ethernet, which
will be extremely useful.olePigeon13th Jan 2009 -
Without seeing more specs...
Without seeing more of the USB3 specs, how do you know much of the work has not been offloaded onto the USB3 hardware itself, similar to, say, ethernet interfaces? A ten times increase in throughput is substantial, and I won't be surprised if the new spec includes offloading much of the work to the interface itself.
Let me guess... You are just upset because one of the only industry accepted hardware "innovations" from Apple is fading away?
Qbt13th Jan 2009 -
No...
No, I'm just clearing up any misconceptions which have
obviously been the basis of your opinion.
Intel has already stated that multi-core CPUs will greatly
benefit USB 3.0 speed, insinuating that USB 3.0
performance is still largely based on the host CPU for data
management.
The only area where Firewire competes with USB is external
storage, which is where USB is taking over. That is largely
inconsequential to the importance of Firewire as it is used
in many areas where USB isn't (or is incapable.) For
example, Firewire is the staple of aviation communication
and is the standard systems communications cable for
nearly all new aircraft. It is also the industry standard
interconnect cable for professional audio and video
equipment.
Recently the IEEE-1394c specification was also ratified.
This allows for Firewire to be run over standard ethernet.
This is an extremely important step for not just the
professional audio and video markets, but also for aviation
and automotive. This will allow Firewire interconnects over
cheaper, more robust, longer, and (more importantly with
aviation) lighter cables.
With Power-Over-Ethernet being upgraded to 24W of
power, Firewire is being explored as a much cheaper
solution to iSCSI and Fiber for networked storage.
As I sated, USB and Firewire compete in different markets.
The only place where Firewire would be "fading" would be
consumer desktop storage and perhaps consumer digital
video, the only two markets where USB and Firewire
actually competed.olePigeon13th Jan 2009 -
Well
"USB can not compete with FireWire simply by virtue of it being a host/slave paradigm and relying on the CPU."
We haven't really seen what USB 3.0 says about that yet. That didn't stop adoption of USB 2.0, though. These days there is a lot of CPU to spare, often in multiple cores, so even though using CPU is a bad thing the drawbacks of using the CPU may not be big enough to justify using something else.
"eSATA I never understood. eSATA isn't meant for hot-swapping since it has to reassign hardware addresses every time you plug and unplug the device."
Actually eSATA is designed for hot swapping. Hardware reassignment is in fact what makes it possible, as it can be re-plugged into a new computer, which may have the address already assigned to another device. It needs to have that ability in order to work without conflicts.
Being able to reassign addresses and IRQs and all of that stuff is what makes hot swapping and plug and play possible. Nobody wants to go back to the old ways of manually setting addresses and dealing with conflicts because some device stubbornly doesn't want to change its address. It has to be dynamic to be hot swappable.
"You also can't daisy-chain eSATA like you can with Firewire."
Because it is meant for drives, not for general device handling. It doesn't make sense to daisy chain drives.
CobraA113th Jan 2009 -
Not quite...
"We haven't really seen what USB 3.0 says about that yet.
That didn't stop adoption of USB 2.0, though. These days
there is a lot of CPU to spare, often in multiple cores, so
even though using CPU is a bad thing the drawbacks of
using the CPU may not be big enough to justify using
something else."
Not quite. Intel demonstrated USB 3.0 on a Linux machine
at CES. It managed speeds of about 5x faster than USB
2.0, well short of the 10x. Intel also stated in a myth
buster document that USB 3.0 will benefit from multi core
CPUs. That indicates that USB 3.0 speed is still dependent
on the CPU.
In terms of external storage and several consumer devices,
yes, USB was adopted over Firewire. Because Firewire
requires a dedicated host controller, it makes it more
expensive and larger where space is considered a
premium.
There are many justifications for using Firewire over
USB because of USB's dependency on the CPU. Most of all
is the professional audio and video markets. Even 8-Core
CPUs are being pushed to the max in the professional
market.
For example, when you're pulling HD DV from a camera,
you can't use USB. Most of your CPU is going to
processing the incoming video stream, and the stream
needs to be uninterrupted. Likewise with professional
audio equipment in which lots of data is transferred and
processed in real-time. You can't lose or reduce the
speed.
Also, as I mentioned in my other reply, Firewire is used
extensively in aviation and now automotive industries.
Recent ratification of the 1394c spec will increase its use
(and usefulness) to many different industries, allowing
Firewire to be run over standard ethernet. In combination
with an upcoming new PoE spec to push power to 24W,
Firewire is being considered as a cheap alternative to iSCSI
and fiber network storage.
So as I stated before, they're different markets. USB and
Firewire overlapped in consumer storage and briefly with
iPods, but that was about it. Many computers are still
equipped with Firewire 800, and it still makes a superior
connection for external, mobile storage.
"Being able to reassign addresses and IRQs and all of
that stuff is what makes hot swapping and plug and play
possible."
I understand that, but it's an absolutely horrible way of
managing external devices.
"Because it is meant for drives, not for general device
handling. It doesn't make sense to daisy chain drives."
Yes. Firewire, on the other hand, can be used for both.
Since a Firewire host controller talks directly to other
Firewire hosts, it doesn't even need a computer to do the
negotiating. Unlike USB, daisy chaining Firewire drives
does not significantly impact the speed of various attached
drives (or your computer's responsiveness.)
So having Firewire makes daisy chaining sensible especially
if you're hotswapping.
So this begs to question why people use eSATA. Probably
because eSATA is "cool" like USB, and that Firewire is "teh
suck." *roll eyes*olePigeon13th Jan 2009 -
may know about USB and Firewire but don't know jack shiznit about eSATA..
as others have said eSATA *IS* hot swappable...
and i think people use eSATA for the same reason i went to it...
because in real world application it MUCH faster than Firewire 400
or 800... had a drive fail with about 500GB of movies restoring
over USB... ouch! that was painful!
i have tested Firewire 400 versus eSATA on my system (Mac G5)
and typical transfers are nearly ~3 times faster over eSATA (eSATA
only enclosure.. with multi interface enclosure eSATA can suck)...
so if you have a whole bunch of video you want to move around
you want eSATA because it much faster than any flavour of
Firewire... not really fair to compare USB 2.0 to eSATA since USB 2.0
is just a dog on macs... but were talking eSATA transfers of 11
time USB 2.0 speeds... yes 11 times faster!
doctorSpoc13th Jan 2009 -
The key question's going to still be overhead
Its understandable that there's going to be some degree of conflict and competition for USB3 -vs- eSATA -vs- FW3200.
Similarly, it is straightforward for any of who are technologists to recognize that they're simultaneously both similar and different: their sponsors have slightly divergent goals, which results in different design priorities and trade-offs.
Insofar as eSATA, we should also remember that eSATA began as an internal SATA cable that was externalized to the host and was a klunge with headaches. While these have finally been sorted out, they didn't necessarily capture key marketplace customers such as Firewire has done (Video; transporation).
In terms of I/O performance, for a single-spindle hard drive, it is the HD that's the bottleneck, so there was no real practical differences between FW800, SATA-I, eSATA and now USB3 as well. However, RAID arrays can go higher, so some of the differences can now show up there...but that footnote is relevant to only the proverbial ~5% of the ~5% portion of the market (ie, the few that actually run external RAIDs).
In any event, USB3 has all of the indications of repeating the same issues that USB2 has had: the specification talks to its peak theoretical bandwidth, but disclosure by its sponsors as to what percentage of bandwidth is lost to overhead invariably seems to never be mentioned. I've found some documentation for USB2 that says that it has a ~40% bandwidth loss due to protocol overhead. Similarly, FW3200 has stated that its I/O will be 97% of its bandwidth, indicating only a 3% overhead loss.
So while we as technologists can all run through the numbers to see what overhead % rate that USB3 has to improve by in order to match the FW3200 performance, the non-technical reality is that the sponsors of USB3 are going to force its broad adoption and its going to be another example of "Better" being killed off by "Good Enough".
While I do appreciate the likely-large gains that USB3 will provide over USB2 (and even FW400 and FW800), what I actually do *not* look forward to is USB3 in the plug-compatible design interface that they've promised. It is going to make it very painful for the lay-person to actually wire up his system to actually get USB3-level speeds, because while the plugs are technically backwards compatible, this is effectively a disservice, since a USB3 device plugged in with a USB2 cable (or hub) will only be able to achieve USB2 speeds, since there's a piece of fiber missing along the path between PC and peripheral.
And finally, since the USB2-"B" plug is an utterly horrible design from a human interface perspective, even if USB3 is a success in the marketplace, I will nevertheless loathe it for perpetuating the loatheful design instead of killing it for anything better. If nothing else, the Firewire plug interface design is less bad.
-hh
-hh15th Jan 2009 -
Well, no
USB 2 beat FireWire in consumer PCs because it was
offered essentially free, whereas FireWire cost a few cents
extra at manufacturing time. USB 2 was never as fast as
FireWire S400 and was left in the dust by S800, but guess
which interface "won"?
It is a shame Apple chose to drop FireWire, but I guess
they had no choice. Their pro laptops had one S800, one
S400 and a couple of USB2 ports, but most consumers
would prefer 4 USB ports. It's a shame that consumers
never realised the benefits of FireWire, so now they are
stuck with an inferior product.
C'est la vie.
Incidentally, eSATA can be hot swap too, it depends on the
driver.
Fred Fredrickson13th Jan 2009 -
Apple only dropped Firewire from their consumer line...
Apple only dropped Firewire from their consumer line of
products. Firewire is still very much apart of their
professional line.olePigeon13th Jan 2009 -
Nice.
Nice. No more slow external drives. I've always had a tendency to avoid external drives for that reason.
eSATA is nice as well, but it was for drives only.CobraA113th Jan 2009 -
bvwj13th Jan 2009 -
well drive speeds need to go up...
most drives can barely saturate a SATA I interface
(150MB/s) so transfer speed will not go up at all vs
the likes of eSATA (1.5Gb/s SATA I or 3Gb/s SATA II
max)... the bottleneck, at least today is the drives
not the interface...
as the guy in the article says, you are not going to
get any faster than ~150MB/s because that's the limit
of what the drive can do.doctorSpoc14th Jan 2009
Talkback - Tell Us What You Think
The best of ZDNet, delivered
ZDNet Newsletters
Get the best of ZDNet delivered straight to your inbox
Facebook Activity
