|
by Tom Cantrell
Start ı Nearvana
ı Waiting for Baudot ı Not
So Fast ı VolksPort ı Going
Mobile ı Sources and PDF
NOT SO FAST
Just as mistaken as the notion that 2.0
will blitz the market overnight is the idea that it will replace 1.1
gear. Look at it this way. Every 2.0-capable host that does get installed
provides a new home for 1.1 devices as well.
Even the staunchest USB 2.0 promoter
isnıt going to argue that you need a 480-Mbps mouse or keyboard. Going
further, and against conventional wisdom, Iıd argue there are a lot
of middle-of-the-road applications that can get by just fine with
1.1. Why? There are two reasons.
First is a smart feature of 2.0, speed
matching. In 1.1, a low-speed connection propagates upstream through
intervening hubs clogging up the final connection at the PC itself.
By contrast, a 2.0 hub is designed to accommodate whatever data rate
the device prefers while communicating upstream towards the PC at
high speed.
Second, remember that from the point
of view of a USB peripheral the connection is point-to-point. It doesnıt
matter how efficiently the peripheral uses cable bandwidth, because
it doesnıt have to share it with anyone. Thus, thereıs little to be
gained from boosting peripheral USB bandwidth much beyond the bandwidth
of the device itself.
Consider the typical low-cost inkjet
printer. At a few pages-per-minute throughput, letıs say bandwidth
into the paper hopper is about 1 MBps (e.g., a 20-MB color picture
takes 20 s to print). Of course, the CPU inside the PC always wants
to do I/O much faster than poky mechanical devices. Thatıs why buffering
(i.e., speed matching) is so important. In the case of the inkjet
printer, the file is actually spooled to disk at a higher speed.
Yes, thereıs a lot of protocol overhead
for USB, and the connection gets more crowded as you head upstream.
But, remember that the entire bandwidth is available on the particular
cable that connects each peripheral. A 12-Mbps USB connection fully
dedicated to efficient (bulk in USB speak) transfers probably delivers
close to the 1 MBps the printer can ultimately digest.
What would having a 480-Mbps 2.0 port
in the printer buy you? If you donıt turn off the spooling, the user
experience (the time from clicking Print until the system comes back)
is still the same (the time to write the spool file to disk). None
of this makes the printer heads fly or ink spew any faster, so the
paper wonıt appear in the hopper any sooner.
If you do turn off spooling, it means
youıll need a huge buffer somewhere down the line. Yes, RAM is cheap,
but 20 to 30 MB in a sub-$100 inkjet is pushing it. And even if the
printer can buffer a complete page, what happens when you try to print
two?
PREVIOUS
NEXT
Circuit Cellar
provides up-to-date information for engineers. Visit www.circuitcellar.com
for more information and additional articles.
For subscription information, call (860) 875-2199, subscribe@circuitcellar.com
or subscribe
online. ıCircuit Cellar, the Magazine for Computer Applications.
Posted with permission.
|