Portables deliver noteworthy flexibility

After
checking them for shipping damage, the staff categorized them by processor and speed,
weighed them and examined components. The lab staff powered up each system with AC
current, installed an operating system if necessary and looked for conflicts.

The GCN Lab then loaded test applications and files and ran them in the same sequence:
Symantec Norton Utilities 3.0 for Microsoft Windows 9x with appropriate updates,
GCNdex32TM benchmark suite, year 2000-readiness tests, multimedia files and an OpenGL 3-D
video benchmark.

During testing, the lab disabled components of the hardware, OS or test applications if
they adversely affected benchmark performance.

To simulate typical patterns for battery life, we carried the notebooks out of the lab
for a period of time. After each unit was returned to the lab, it was turned it off and
attached to an AC outlet to charge overnight before running the maximum-drainage battery
test.

Next the lab examined and recorded each notebook’s features in the testing
database, attached each unit to the GCN Lab network, and downloaded and printed the
results of all tests. When testing was complete, we wiped each unit’s hard drive
before returning it to the manufacturer.

For further details about the testing methodology, visit the lab’s Web site at http://www.gcn.com/gcnlab and click on “More about the GCN
Lab.” 

—Michael Cheek

GCN Lab assistant Donovan Campbell contributed to this report.

Sidebars

Too many jolts make OmniBook Sojourn
fall short in workmanship.

Thirty portable computers made their way to the GCN Lab for the 1998 round of notebook
reviews. Some have the mighty Pentium II processor and other sophisticated components. But
how much has changed since the lab staff examined 27 road warriors two years ago?

In this issue, GCN Lab manager Michael Cheek looks at portables powered by Pentium MMX
and AMD K6 processors, as well as a mininotebook and a subnotebook. In the Oct. 19 issue,
Cheek reviews 17 Pentium II units, several of which boast Intel Corp.’s newly
released 300-MHz mobile processor.

Now that notebook computers have high-end Pentium II processors, is there any point
buying a Pentium MMX notebook? You bet.

The 17 Pentium II portables in next week’s comparison have one overwhelming
drawback: short battery life. Their speedy processors and large displays rapidly drain a
charge. Pentium MMX notebooks are more affordable and run a lot longer.

When the GCN Lab staff ran this week’s dozen units—10 Pentium MMX models and
two with K6 chips from Advanced Micro Devices Inc. of Sunnyvale, Calif.—through the
lab’s maximum-drainage battery test, a few lasted barely an hour. Some, however,
passed muster by surviving longer than two hours.

Dominating the dozen was a surprising newcomer, MetroBook Computer Corp.’s
MetroBook LT. It surpassed even units from elite vendors Hewlett-Packard Co., Compaq
Computer Corp. and IBM Corp.

The MetroBook LT’s good performance, features and weight earned it the only
Reviewer’s Choice designation in the field. Unlike other third-tier
manufacturers’ bland designs, the LT looked as sleek as Dell Computer Corp.’s
Latitude CPi, which will be reviewed in next week’s issue.

The LT demonstrates why there are good reasons for buying Pentium MMX notebooks—at
least until Intel and LCD manufacturers slim down their components’ appetites for
power.

Because buyers demand that notebooks function at the same level as desktop systems,
makers struggle to balance performance and features against weight and battery life.
Sometimes they go to extremes in units such as Compaq’s Armada 1598DMT, which has
almost every extra imaginable but weighs nearly 8 pounds, or Toshiba America Information
Systems 100CT, which weighs 21'2 pounds but lacks some needed components.

Every notebook should give the user simultaneous access to the hard drive and CD-ROM
drive, or what manufacturers call dual-spindle access. A spindle is the shaft around which
a disk platter rotates.

Two notebooks, the Toshiba Libretto 100CT and the IBM ThinkPad 560X, lacked this dual
access, which is essential for software installation. Both had a plug-in floppy drive, but
who wants to load a suite of applications from 20 or more diskettes? The notebooks did not
offer integrated Ethernet connectivity to load software from the network.

The remaining 10 notebooks had at least dual-spindle access. Seven boasted
triple-spindle, all-in-one access to hard, floppy and CD-ROM drives.

The CD and hard drive get the most action, and that’s where the GCN Lab
concentrated its maximum-drainage battery life test.

The torture test keeps CD-ROM and hard drives spinning while the processor crunches at
top speed and the display is lit. No power-saving software utilities are permitted to
engage. The maximum-drainage test reveals the absolute minimum time users can expect from
a notebook battery charge.

When notebook manufacturers claim up to 10 hours of operation, they generally mean that
some or all components can rest in an idle state part of the time. There’s no nap
time in our maximum-drainage test.

The GCN Lab staff considers two hours the minimum acceptable operating time under the
maximum-drainage test; any time beyond 21'2 hours deserves an exceptional score.

The lab also records how long a portable survives under ordinary use—ideally,
about three hours. Any operating time beyond four hours is extraordinary.

Would the lab staff judge an 8-pound notebook with a 11'2-pound battery better than a
6-pound portable with a 3'4-pound cell, merely because the first unit survives 30 minutes
to an hour longer? The question prompted the lab staff to include a new benchmark gauge:
minutes of life per ounce of battery weight.

For example, say Vendor A’s notebook runs 21'2 hours on the maximum-drainage test
and lasts 33'4 hours under normal use with a battery that weighs 13 ounces. Vendor
B’s unit lives for 31'2 hours under test and 43'4 hours normally with its 1-pound,
5-ounce battery.

At first glance, Vendor B’s portable might seem the better of the two. But if you
calculate by battery weight, Vendor B’s battery fuels about 10 minutes of operation
per ounce, whereas Vendor A’s cell gives more than 11 minutes per ounce. Vendor
A’s system is slightly more efficient at conserving power.

The GCN Lab’s tests on units with a minimum of two spindles produced a wide
spectrum of performance, ranging from 4 minutes to 15 minutes per battery ounce.

The lab rated as inefficient any notebook that gave less than 8 minutes of operation
per battery ounce. Portables with a rating greater than 10 minutes per ounce were judged
power-efficient.

When it came to power conservation, the MetroBook LT earned distinction under maximum
drainage, in normal use and in time per ounce.

The 61'4-pound LT weighed the least among units with two or more spindles. To access a
third spindle, a lightweight cable plugged into the side and an external modular bay
accepted either CD-ROM or floppy drives.

On the GCNdex32TM benchmarks, the LT scored among the highest of all 233-MHz Pentium
MMX notebooks the lab staff tested and it had the best 2-D video marks.

The 2-year-old MetroBook company may be new to the field, but if the LT is any
indicator of future designs, other notebook makers should watch out.

Compaq’s Armada 1598DMT also performed well on the benchmarks, earning the best
math scores. Its all-in-one design conveniently integrated an AC adapter and modem. But
the Armada’s bulk kept it from winning a Reviewer’s Choice designation. At 73'4
pounds, the Armada notebook alone weighed more than the MetroBook’s luggable weight.

Luggable weight includes all required cables, AC adapters, spindles and, in most cases,
a PC Card modem. The lab considers 8 pounds the cutoff for notebooks. The Armada came
close to that mark, and even its price was heavy at $2,524—the most expensive of the
dozen. Compaq discontinued the 1598DMT recently, although a few may still be available.

Toshiba’s Libretto 100CT distinguished itself by a low, 21'2-pound weight. The
mininotebook had certain inconveniences, however. Its keyboard was just too small, making
the simplest tasks hard.

The QWERTY keys measured 14 mm by 11 mm, or 154 square mm. Function and other keys were
even smaller.

Most notebooks have QWERTY keys measuring 18 mm by 18 mm—almost double the area.

The Libretto’s button pointing device, just left of the 7-inch active-matrix
display, lacked sufficient accuracy for cursor actions. I disliked the mouse buttons,
located on the back side of the display.

The Libretto had a PC Card floppy drive but not a CD-ROM drive. It needed bundled
CD-ROM connectivity. But its low-powered, 166-MHz Pentium MMX processor and the absence of
a second spindle kept the little unit running for quite some time. A port replicator slice
offered Universal Serial Bus, parallel and serial ports, but not a CD-ROM port.

At the other end of the weight scale, the MetroBook LT’s big brother, the
MetroBook DT, stood out as the oddest portable of the lot. At more than 10 pounds, it was
too hefty for even a diehard road warrior. But it had desktop replacement features,
including a 15.1-inch active-matrix display and a 102-key keyboard with a real numeric
keypad.

Although the display had two vents, the DT ran too hot to be comfortable as a laptop. A
warning sticker on the bottom recommended avoiding poorly ventilated locations.

The DT, like the LT, claimed to be a voice notebook. It bundled NaturallySpeaking
Personal Edition 1.0 software from Dragon Systems Inc. of Newton, Mass., along with an
inexpensive headset with microphone.

But as I trained NaturallySpeaking to understand my voice, I could not adjust the sound
to anything better than below average, as the application itself termed it. The GCN Lab
has tested NaturallySpeaking before, and I was familiar with the setup and training
requirements [GCN, Aug. 25, 1997, Page 1].

The DT arrived with only 32M of RAM, which explained some of its difficulty in
crunching the voice training information—when I could get the unit to work. I do not
advise buying it with anything less than 64M, and I recommend the maximum 128M RAM.

The Fujitsu PC Corp. LifeBook 280Dx, like the Armada, had an integrated modem. Its
ErgoTrac pointing device, a big round button that propelled the cursor, was frustrating to
use while trying to tap the mouse buttons or dragging.

The LifeBook’s passive-matrix display extended the battery life but was not bright
enough.

Micro Express Inc.’s NP3233 had a 233-MHz K6 processor and an impressive
price—just $1,499. The K6 unit performed as well as or better than some 233-MHz
Pentium MMX desktop PCs the lab has examined.

One big problem plagued the NP3233, however. Under Microsoft Windows 98, the battery
indicator always showed a 50 percent charge. After an overnight recharge, it still showed
50 percent. Even just before failure, it showed 50 percent.

EPS Technologies Inc.’s Apex IL-233 had a serious configuration error: Only 2G of
its 3G hard drive was formatted and mounted. An entire gigabyte of hard drive space
remained unused. Otherwise, the Apex performed adequately although it was a little heavy
and less energy-efficient than expected.

The OmniBook 2100 from Hewlett-Packard had only a 200-MHz Pentium MMX processor, but
benchmark performance exceeded that of most 233- and 266-MHz portables in the comparison.
The OmniBook earned high marks for its 3-D and hard drive access performance.

Battery life on the maximum-drainage test was an impressive two hours with a
nickel-metal hydride battery. Moreover, the price was an attractive $1,499.

The OmniBook’s 12.1-inch passive-matrix display lacked brightness in ordinary work
environments. Its double-layer supertwist nematic (DSTN) technology weaves light through
two layers to achieve high contrast.

DSTN provides satisfactory display quality for short-term use, but the OmniBook 2100
didn’t push enough light through its screen, which needed a better anti-glare coating
to keep ambient light from overpowering the images.

A year or so ago, the Panasonic Personal Computer Co. ToughBook 45’s 200-MHz
Pentium MMX might have been an impressive performer in a comparison such as this. At
present, it cannot quite measure up to the other units because of its weight, awkward
trackball and fairly slow processor. The ToughBook 45 performed well enough, although the
hard drive was sluggish.

The ToughBook’s claim to fame is sturdiness, although other Panasonic models seem
more rugged than the 45. For example, two foam strips on one side comprised all the
shock-proofing for the 45’s hard drive. The ToughBook 35 had a series of gel strips
on both sides of the drive.

Weighing little more than 4 pounds, IBM’s ThinkPad 560X was an easy carry but had
no CD-ROM connectivity. IBM did, however, load the notebook with client management and
antivirus tools plus Lotus Development Corp.’s SmartSuite 97 office software.

The 560X’s battery stayed alive only an hour on the lab’s maximum-drainage
test with only one spindle—likely the reason no CD-ROM drive was included.

When I first started up the 560X under full charge, the Windows 95 battery meter and
IBM’s own fuel indicator predicted two hours of use.

Both utilities showed about 50 percent charge with an hour left to run, yet the cells
had already drained. The 560X shut down without warning.

At next boot, both utilities predicted only an hour of potential life. That shows the
indicators were not polling the battery for any remaining charge; they were merely
estimating from previous patterns. Sometimes that works, but it is often inaccurate, as it
was in this case.

In addition, the function keystroke combination that brings up IBM’s fuel
indicator did not work, leading me to suspect the 560X had not been properly configured.

Arm Computer Inc.’s ArmNote TS175 had one of the newest AMD processors, the
300-MHz K6-2.

It performed well enough on benchmarks and scored the highest of the bunch at 2-D
video. But inefficient battery use, dim passive-matrix display and hefty weight counted
against this notebook. Its sibling, the ArmNote TS1200L, ran down even sooner because of
an active-matrix display.

On the maximum-drainage test, the TS1200L’s nickel-metal hydride cells lasted an
hour. Without USB and infrared ports, the unit seemed a little behind current
technology, and the PC Card bus did not work. 

You can’t help admiring a notebook computer that weighs only 3 ' pounds, measures
no more than three-quarters of an inch thick and manages to incorporate a 12.1-inch
active-matrix display.

Yet what makes Hewlett-Packard Co.’s $3,699 OmniBook Sojourn so compelling also
makes it fragile, even flimsy.

When the GCN Lab reviews products, most companies send us brand-new units fresh off the
production line. Hewlett-Packard generally sends review units from a demo pool, which
means other folks have handled and tested the units.

I suspect my test Sojourn unit had passed through several sets of hands and had been
dropped a few times in shipping across the country. The lightweight case just didn’t
seem strong enough for road warriors or less-than-careful users.

At first, the Sojourn worked haltingly, going blank or locking up. Over several weeks
of sitting in the lab and not going anywhere, it grew progressively worse. Apparently
connections in the flat-panel display’s graphics hardware had come loose, but images
looked fine as long as I connected the Sojourn to an external display.

The unit crashed periodically. Also, the power button failed to turn it off completely.

Whenever it crashed, I had to resort to sticking a bent paper clip into a tiny reset
hole. And when the Sojourn suspended, it would crash from any stray motion against the
touchpad during the resume process.

The membrane keyboard took a little getting used to for touch typing. As for the other
parts of this comparative review, the Sojourn could not complete most of them, although I
was able to put it through the GCNdex32™ benchmark suite. Math and 2-D video
scores were respectable, but the 2.1G hard drive’s access scores were low for a
233-MHz Pentium MMX notebook.

An extra battery and a multimedia slice with CD-ROM and floppy drives raised the
Sojourn’s weight to more than 8 pounds. Hewlett-Packard claimed up to 1' hours of
operation for the base unit and another three or four hours with the battery slice.

Despite the Sojourn’s failure at most of the test runs, I remain impressed by its
engineering. Its flaws probably were not inherent but stemmed from its demo pool handling.