[KLUG Hardware] PCMag, consider myself as a replacement for Bill Howard ...

[Bryan J. Smith]

The key to being a good technologist is not only keeping up-to-date with
the latest technology, but keeping up-to-date with what the mainstream
magazines are [often incorrectly] telling people.  Yes, this includes
subscribing to Ziff-Davis' PCMag, the publication that pays close
attention to itself.

Some of the commentators aren't completely clueless, and to each its own
when it comes to opinions.  But those who present technology with
completely false assumptions and, worse yet, with poor analysis of data
that states completely to the contrary, some are really in need of a
good rebuking.  One such individual is Bill Howard who is featured in
PCMag's "On Technology" column.

This issue (2002 June 11) it's USB 2.0 (page 71, not on-line yet, but
you can find his columns posted here: 
http://www.pcmag.com/category/0,2999,s=1497,00.asp).

- The Non-Technologist, What's New?

In typical, "I don't the know the technology fashion," he talks about
how 480Mbps USB 2.0 is rated higher than 400Mbps IEEE1394 Firewire and
leaves it at that.  He does finally mention the 800Mbps Firewire speed
option (which isn't very common, I'll admit) as higher Firewire2 speeds
at the end.  But all that's really non-issue fluff because he doesn't
talk about how Firewire was build for direct-access, high-speed block
devices with a full command-set without bothering the host CPU (although
he does mention direct device transfer capability of Firewire late in
the article -- he should have expanded on why) and is far more efficient
than USB, which is an extension of a programmed I/O (PIO) bus that lacks
solid command sets (there are endless, and often UNsupported "industry
standard protocols" for many devices) really designed for slow,
character devices.

But I didn't expect such.  He does say some things about the speed being
"theoretical" and makes a comparison to 100Mbps Ethernet -- but actually
that would be more true of FireWire that probably gets 70-80% of
theoretical (this can be as low as 50% if you use a bridge like ATA33 to
FireWire where the bus is slowed down at the device), not USB 2.0 where
Intel has found that 50% is not even a reality, and under 20% is more
typical (about 5-8x faster than USB 1.1 due to the design of how USB
works -- remember, it's NOT built for block transfers).  I'm not going
to fault him for not knowing the technology.  Many people just see
numbers and make assumptions on past experiences with other
technologies, which is common.  God help us if he hears of iSCSI.  ;->

- Ignoring His Own Data, Now That's Bad!

But then he starts ignoring his own testing.  "I've achieved speeds up
to 50Mbps with USB 2.0" ... "this should give new life to external hard
drives because they're essentially as fast as internal drives."  NOT! 
Billy, 50Mbps (6.25MBps) is still _dog_slow_ for hard drives.  Even your
own PCMag found that USB 2.0 is great for high-speed optical, various
external devices, but it is far _too_slow_ for hard drives with
64-80Mbps (8-10MBps) being more of the "limit" right now.  Most hard
drives can burst upto 5x that, and even sustained transfer rates are
easily 2-3x that.  Even external FireWire drives with cheap
ATA33-to-FireWire converters (ext. case+converters are now under $50
mail-order, $69 at CompUSA nowdays) can achieve 160-200Mbps (20-25MBps)
and are far closer to the speeds of internal drives.  Since FireWire
hasn't become commodity thanx to Intel's _reniging_ on adoption of the
technology in its chipsets (although AMD, SiS and ViA are changing
that), SerialATA will be the solution that "equalizes" internal and
external storage performance.

- From Bad to Worse:  Ignoring EHCI, No 2.0 in Win/DOS

But some people are bad at math, or don't stop to catch their own DTR
conversions (or lack thereof).  Well, Bill doesn't stop there.  He talks
about 4, 5 and 6-port USB 2.0 controllers.  Hey Bill, ever stop to
realize that most USB 2.0 controllers _only_ support 2.0 speeds on _one_
or _two_ ports?!?!?!  Nope.  He could have thrown in a paragraph or two
about the older 12Mbps USB 1.1 OHCI versus 480Mbps USB 2.0 EHCI -- EHCI
*IS* USB 2.0!  In fact, he goes out of his way to "recommend" 6 port USB
cards instead of 4 port, but doesn't realize they _both_ usually have
(of the 3 major USB 2.0 chipsets out now) but *1* single 480Mbps EHCI
controller (possibly 2 ports connected to it if you're lucky).  Sorry
Bill, _same_difference_!

Worse yetAnd then he moves on to talking about how the five devices he
used installed "glitch-free using Microsoft Windows 98 and XP."  But he
doesn't realize that 480Mbps USB 2.0 devices are backward compatible to
12Mbps USB 1.1 _because_ that's all Windows 98 supports, 12Mbps!  All
Win/DOS versions do NOT support 480Mbps on EHCI ports, and the only
Win/NT version is Windows XP (with a patch).  And although Windows 2000
USB 2.0 is planned, it's not available yet (i.e. possibly never?).

Bill, bill, bill, bill ... man!  Know thy hardware and OS dude!  I know
you have to write these articles weeks in advance, but much of this info
has been available since _last_year_!  Freely available on the 'Net!

It's times like these I hope Ziff-Davis seriously considers hiring
someone like me to be their technology guru.

-- Bryan

-- 
The US government could be 100x more effective, and 1/100th the
Constitutional worry, if it dictated its policy to Microsoft as
THE MAJOR CUSTOMER it is, and not THE REGULATOR it fails to be.
---------------------------------------------------------------
Bryan J. Smith, SmithConcepts, Inc.   mailto:b.j.smith@ieee.org
Engineers and IT Professionals     http://www.SmithConcepts.com