[Prism54-users] Comparison testing finds Broadcom better

Jeff Bobzin jeff@boccorp.com
Sat, 29 Nov 2003 18:09:16 -0500


First of all let me say the Prisme54 developers have done a superb job.  =


I don't believe their Linux driver is at fault for these issues.  I =
believe my problems are in the Intersil-supplied firmware or perhaps the =
chips themselves.

Also let me say I only tested Intersil-based products marketed by SMC. =
Other brands might be better. And perhaps Globespanvirata will start =
doing firmware upgrades.

I have previously reported my observations about and the limited =
distance even with high-gain indoor antennas. On Friday I returned some =
SMC equipment to the store and picked up 2 Belkin 54G PCI cards.  All =
Belkin 'g' products are based on the Broadcom chip. As are Linksys.

I installed the trial version of the Linuxant driverloader on my two =
test machines and downloaded the latest Belkin windows driver.  This =
process through somewhat awkward was not difficult. Note that master =
mode is not supported and I saw a failure with 'iwconfig eth1 channel n' =
command. Otherwise it worked OK. They want $20 for a permanent license.

I tested ad-hoc from machine to machine.  I saw higher maximum bandwidth =
at short range. And in the room-to-room (in-home) test I was able to =
receive 15Mbits/sec at the same locations where the Intersil card was =
not able to reliably connect. These tests were same machine, same =
antenna and same test program. Both Hawking Tech antennas worked well, =
boosting bandwidth compared to the 4-inch Belkin-supplied antenna.

It is clearly a shame that Broadcom does not have a better attitude =
about making their Linux drivers available (it is known they have them). =
(see http://www.petitiononline.com/BCM4301/)

But if you also get disappointing range with Intersil and the Linuxant =
driver solution meets your needs, then Broadcom is an alternative you =
might consider.

I will not be burdening this list with additional observations. Good =
luck to you all.

---Jeff
=20