A performance gap separates the leading 56-kbps modem technologies, and
despite the recent adoption of a 56-kbps standard, elements of these
proprietary technologies are likely to creep into future modems, according to a new report.
In the course of comparing call completion rates for Internet service
providers (ISPs), Boardwatch
Magazine made a surprising discovery: Calls made using 3Com's x2 technology are much more likely
to be completed and to connect at higher speeds than calls made using Rockwell's K56flex technology.
The disparity is likely to persist, according to Boardwatch
publisher Jack Rickard, because the recently approved International Telecommunications Union
56-kbps standard was crafted to allow vendors to incrementally upgrade
their products. That is, the proprietary differences that result in the performance disparity will not necessarily be solved by the new high-speed modem standard.
Rockwell vehemently denounced the findings, calling it a flawed report.
Prior to the adoption of the new standard, dubbed v.90, consumers had to
choose between 56-kbps modems using either 3Com's x2 or Rockwell's K56flex technology. Users then had to find an ISP that supported the modem
technology they had purchased.
56-kbps modems promise connectivity at a significantly faster rate than
commonly used 28.8-kbps modems, but despite industry consensus on the v.90
standard, users will continue to see some performance problems because of
differences in telephone line quality around the country, analysts say. In
the past, those differences were obscured because the modem connection was
already excruciatingly slow.
In its attempt to assess the reliability of reaching ISPs, Boardwatch conducted a monthlong test, making over 140,000 calls to 445 Point of Presence (POP) phone numbers serviced by 89 ISPs. The calls were made at peak volume periods, between 5 p.m. to 9 p.m., from the magazine's Denver, Colorado,
office.
The survey employed x2 modems from 3Com subsidiary U.S. Robotics, and K56flex modems from Hayes, Diamond Multimedia, and Motorola, and instructed the dialing
machine to determine whether an ISP was using K56flex or x2 technology and
to use the appropriate client modem technology.
"By the end of the first week, we noticed something kind of alarming," said
Rickard. "Both call completion rates and connection speeds were much better
on the x2 side."
The study found that out of the 140,000 total calls attempted, slightly over 86 percent connected to an ISP. Out of those calls, the average connection speed was about 30,241 bps, or 30 kbps.
Boardwatch completed 90.4 percent of the calls it attempted from a US
Robotics modem to an x2 port. The average x2 connection speed was 45,192
bps (45 kbps), with 90 percent of the connections reaching speeds over 40
kbps.
In comparison, 79 percent of calls from a modem using a Rockwell chipset to
a K56flex port were completed, at an average speed of 30,849 bps. Importantly, only 6.52 percent of completed K56flex calls were connected at speeds of 40 kbps or better.
Rickard predicted the performance disparity will carry over to the new v.90
standard because of the way the standard is written. "This is going to
persist in the v.90," he said. "V.90 is the standard for how the modems
talk to each other. But how the actual process is implemented is up to the
actual modem developers."
Understandably delighted by the report, John Powell, x2 field trial manager
for 3Com, agreed with Rickard's assessment that the disparity would
continue. "Everybody's client implementation is going to be different," he
said. "It gives you a highway, but you choose the car."
Rockwell disputed the findings of the study, saying that because all of the
calls in the study were made from the same location, the results cannot be
extrapolated to indicate worldwide performance rates.
"We told them what was wrong with their methodology and they published a
flawed report," said Dean Grumlose, product line manager of central site
modems for Rockwell. "It flies in the face of common sense that if we had
the kind of performance problems indicated in this study, that we would
have had the kind of success that we have had," he said.
Grumlose also said that some ISPs had found the lines used in the
Boardwatch study to be "atypical."
3Com's Powell said that he examined the lines used in the study and found
them to be "extremely average."
Without commenting directly on the report, International Data Corporation modem analyst
Abner Germanow gave Boardwatch a vote of confidence. "The analysis
and the coverage that I've seen Boardwatch write is typically very
well done," he said. "I've never had any reason to doubt them."
"The Boardwatch study points out what could be a very troublesome
performance problem for the K56flex camp," he continued. "However, with the
v.90 upgrade they have the opportunity to potentially avert that performance problem. Over the next month, when testing on the v.90-based product can be done, we'll see where the performance issues lie, and separate the men from the boys."
The study will be published in Boardwatch's Winter 98 Directory of
Internet Service Providers. Rockwell has already sent Boardwatch a letter detailing its reasons for disputing the study.