[tulip] Macronix MX98715AEC tests

Andrew D. Balsa andrewbalsa@mac.com
Wed, 27 Sep 2000 18:32:57 +0800


Dear Donald,

First, thanks for your continuing effort to bring quality NIC drivers to the
Linux community, and congratulations for the Scyld Web site.

About the MX98715AEC based cards and the Tulip driver:

I have tested version 0.92 of the Tulip driver and it works fine with the
card I tested (Buffalo LGY-PCI-TXL). I measured the performance of the card
using two Duron 600MHz linux boxes and the ttcp benchmark utility. The
throughput is approximately 11460 KBytes per second, just about what one can
expect from a 100baseTx link.

A few issues however arise with the diagnostics programs, tulip-diag and
mii-diag. The problem is, the various MX98715 variants don't have an MII
interface. These cards use the internal transceiver available on the chip,
which does not have the standard MII registers, found for example in the
Digital 2114x chips.

So the mii-diag program exits immediately with an error message, and it
seems the tulip-diag utility cannot set the media type (but it does show the
content of the CSRs and EEPROM correctly).

N-Way negotiation works well and the cards can connect 100baseTx full duplex
among themselves using crossover cables, and also 10baseT half duplex with a
cheap hub I used for testing.

The LEDs may or may not reflect the link status correctly depending on how
the small EEPROM is programmed at the factory. The Buffalo cards mentioned
above were not correctly programmed, so a single line of additional code in
the driver fixed this cosmetic issue.

If I can be of any help to other MX98715 users, please write to me directly,
as I don't subscribe to this list.

Again Donald, thanks.

Regards,
Andrew