Tech Flashback: Fast Ethernet Network Interface Card Collection (Part 2)

This post is part of a two post series looking at the Fast Ethernet NICs in my collection, with some reminiscence of the legacy Fast Ethernet has left behind. Here are some more interesting and varied NICs in my collection.


First, lets start off with this image of a three pin to three pin “white” cable. This particular cable is the wake-on-LAN cable for early PCI slot implementations which could not signal WOL over the PCI pins. Thus, to have functional WOL, this three pin cable had to be connected from the card to the motherboard. This cable wasn’t strictly necessary, and indeed, with many later generic adapters, it wasn’t even included. It did confuse some end users though …


Another interesting artifact was finding a printed, stapled, thick manual for a network card. This would have come with an EzStart floppy disk with drivers for most operating systems. The manual itself is very well specified, with lots of details about the networks it interfaces with.

smcman-spec1 smcman-spec2

The registration card also gives you an idea of what sort of network operating systems were around at the time – Novell Netware, or Banyan Vines anyone?

smcregcard-rear smcregcard-front

Anyway, lets get on with the show!

Netgear FA310TX


Netgear products of the era were generally consumer products held with high repute and sometimes a premium price tag. Their blue colouration was everywhere, and the blue and yellow colour scheme of their products was a refreshing delight compared to the rest of the products. This would have been a relatively early design product, at the time when Bay Networks (later acquired by Nortel, itself acquiring Synoptics and Wellfleet Comms) was trying to spin off its consumer products as Netgear.

This particular design has a Netgear branded discrete MAC and a Broadcom BCM5202 PHY. The MAC appears to be supported by the Tulip driver, which suggests it’s an Intel/Digital 21x4x based device. Despite it looking to be a premium device, the Option ROM socket and Wake on LAN sockets are not present, although the magnetics date code suggests a Week 52 of 1998 production date.


The Netgear man makes a pleasant appearance on the rear.


The rear plate features printed text and the most comprehensive LED set-up I’ve seen in the whole collection. Interestingly, an LED to indicate collisions is provided, which should be so rare as to be never seen (maybe with the exception of talking to a slow 10Mbit/s device through a switch which forces collisions as a form of “back pressure” to slow down the send rate).

Netgear FA331

20151028-2108-0141 20151028-2108-0143

A much later Netgear adapter, well into the generic adapter days. It features a very much cut-down PCB with no options at all for wake on LAN cable or option ROM. It is based on a single National Semiconductor DP83815 MacPhyter integrated chipset. It is dated Week 18 of 2002, and it features the OASIS branding on the PCB (seen on another NIC in the previous part).


Two LEDs are provided on the rear, with one of them multi-coloured to provide the link rate indication. The Netgear man also makes an appearance, stamped onto the rear plate and silkscreened on the rear of the card.

Intel EtherExpress PRO/100+ 689661-004


Any discussion of NICs would be incomplete without some nice Intel branded cards. Generally Intel cards performed well, and were well supported, and were hence the card to use if you were running home servers. This particular card features an all-integrated SB82558B chipset with Pulse branded magnetics and a different type of Option-ROM socket (as was common on Intel cards).


This card was made in Week 46 of 1998 but was not fitted until March 1999 by a local PC builder.


Three LEDs were mounted on the board with holes in the backplate for viewing. The card branding is stamped onto the backplate – and the internally shielded socket is a specialty peculiar to Intel NICs.

Intel/Compaq NC3121 323557-001

20151028-2113-0157 20151028-2113-0158

A very similar card utilizing the same chipset was also found under a different number, which is a Compaq OEM product. This one features a WOL header, pre-soldered option ROM and an additional U8 chip for some reason.


The rear backplate lacks the Intel branding, and has a no-telephone-line symbol instead. It is dated Week 52 of 1998.

Intel/Dell Ethercard 100TX 721383-007


This is an example of a later card featuring an improved Intel GD82559 chipset. This chipset was one of the chipsets to bring hardware accelerated networking to the mainstream, with TCP/UDP checksum off-load capabilities, transmit chaining/queue and back-to-back transmission support. This card appears to have been OEMed for Dell machines, and features a WOL header and soldered down option ROM.


The card is dated Week 31 of 1999. The rear features two LEDs instead of the three above, and is branding-free due to the OEM nature.


Intel PRO/100S Desktop Adapter 751767-004

20151028-2111-0151 20151028-2112-0152

Almost visually indistinguishable from the above adapter, this is actually a secure adapter, featuring onboard Triple-DES encryption offloading for IPSec acceleration. It uses the 82550EY chipset, which is pretty much pin-for-pin compatible with the other chipset above, hence using a similar PCB base. A definitive date is hard to come by, but the magnetics seem to claim Week 15 of 2001.


Intel/IBM 19K5544 Pro/100 PCI-X Adapter


In what I will call the strangest shaped NIC in my collection, this is very much just a wedge. It is intended for 64-bit PCI operation, but utilizes very few pins in the PCI-X connector, and so is cut short to save PCB space. The unit is backwards compatible with regular PCI and features the smallest chipset (DA82562ET Platform LAN Connect) I’ve seen so far. The PCB is from Topsearch and the magnetics claim a date of Week 43 of 2000.


OEMed for IBM, it carries an FRU number of 19K4885011.


No LEDs are provided separately, instead, they have been integrated with the LAN jack as more modern devices also have.

Sun 270-5406-02 Quad-Port Adapter


This particular adapter is a massive adapter intended for Sun Microsystems servers. This adapter was intended for use with PCI-X 64-bit PCI, although is backwards compatible if plugged into a regular PCI slot. It features 4x100Mbit/s ports and its construction is straightforward, if not, clunky. An Intel bus controller is used to connect four Sun “BigMAC” controllers through buffers onto the PCI Bus. Each of these MACs drives a Lucent PHY to drive the twisted pair interface. Despite its clunkiness, the PCB is elegantly symmetrical.


Judging from the non-integrated nature of the construction, it’s likely that this card is an older device – from the manual, probably about 1997 to 1998. Despite being intended for Sun servers, it does actually operate properly under most Linux distributions.


The rear has the four ports, all separate independent interfaces, with combination link/activity LEDs for each port. This makes this card useful for network monitoring, switching, or running separate networks into a single node.

However, other quad-port consumer cards also existed, with some being a switch on a card, or router on a card, which did not have independent ports. These were rare and I never actually used one myself.



20151028-2048-009220151028-2049-0093This is the only USB fast Ethernet adapter I still have in my possession. It is a more recent, USB 2.0 model, which is capable of delivering more complete performance than a former USB 1.1 model I had (which barely eclipsed 10Mbit/s despite being sold as 100Mbit/s due to USB interface limitations).

Being made into a simple bus-powered puck design, with a single LED, it was actually quite handy as there was support in Windows XP for the device.

20151028-2050-0096 20151028-2050-0095

It was very compact, and uses an integrated ADMtek Pegasus II ADM8511 chipset. ADMtek was acquired by Infineon in 2004. This unit was made in Week 14 of 2002.

3Com Megahertz 10/100 LAN PC Card 3CCFE574BT

20151028-2105-0133 20151028-2106-0135

This card is intended for use with laptops that had no integrated LAN connectivity. It is an older 5V 16-bit PC Card with no DMA capabilities, and is estimated to date from 1998 (based on copyright date). Due to the limited expansion capabilities of laptops, these were the only real option to add connectivity to them.

20151028-2106-0136 20151028-2106-0134

The end features a slim slot through which a break-out cable is used to connect to the CAT-5 cable. These connectors were hateworthy, as lateral force on them easily led to damage and cracking of the plastic (such as this sample). Sadly, this was the only option, and other vendors (e.g. Xircom, Intel) all had similar issues.

20151028-2106-0139 20151028-2106-0138

The breakouts themselves were not generally interchangeable across vendors due to the use of different connectors. LEDs are provided on the connector itself to indicate link speed and activity.


I hope you’ve enjoyed looking at this collection of Fast Ethernet era gear. Sadly, Fast Ethernet is almost completely eclipsed by Gigabit Ethernet now, so this gear isn’t really that useful anymore, but it does remind us of how important industry standardized technologies are. It’s surprising that for the past 20 years, such a technology saw us through the days of Serial, USB 1.1, USB 2.0 and now into USB 3.0 and 3.1. Where other technologies evolved, Ethernet was a dependable source of connectivity. It seems that wireless technology has also eroded some of the market formerly served with wired Ethernet, and slimmer form factors have led to the loss of Ethernet ports on some computers, however, where network performance matters, cabled Ethernet is still the prime choice for most consumers. Thanks IEEE!

Note: There is some Fast Ethernet gear in other postings as well (e.g. the Intel EtherExpress PRO/100 Mobile Adapter post) so do have a look around from time to time.

About lui_gough

I'm a bit of a nut for electronics, computing, photography, radio, satellite and other technical hobbies. Click for more about me!
This entry was posted in Computing, Tech Flashback, Telecommunications and tagged , , . Bookmark the permalink.

4 Responses to Tech Flashback: Fast Ethernet Network Interface Card Collection (Part 2)

  1. sparcie says:

    What kind of architecture machine did you have that sun card working under linux? I’m interested as I have one or two of cards like this, but they are from sun servers with the ultra SPARC architecture. Of course Linux works on that architecture (as well as *BSD and Solaris), but I wasn’t aware they also worked on x86 or any other arch.

    I also have a few single port ethernet cards for the sun machines that have a similar number of indicator lights to that netgear card.

    The interesting thing about most sun cards intended for their SPARC based servers is that they don’t have a MAC address assigned to each card, rather the machines had MAC addresses stored in their configuration ROM, or on a configuration smart card.


    • lui_gough says:

      It was x86 funnily enough, nothing special, just a regular install of an Ubuntu from memory. The Quad Port cards were favourites of networking students who needed it to simulate real-life 10+ client connections to Cisco switches, amongst other things, so it seems like someone wrote a driver for it and it got incorporated into the kernel. Of course, no harm plugging and trying if you’re unsure.

      Ah yes, the onboard RTC of the Ultra 10’s of mine in storage had pooped it several times before because it had basically expired its internal battery and I do remember the Ethernet address being something that had to be re-written into it. That being said, if you look at many of the cards, you can see a tiny serial EEPROM with a sticker on it that has the MAC address (most obvious on the Intel ISA 16-bit 10Mbit/s card), and that would be pre-flashed by the manufacturer with the intended MAC before leaving the factory. Later cards were probably just programmed via a debug interface, so they had a hardware MAC but can also spoof any other MAC by driver configuration.

      EDIT: And I forgot to say that more modern machines have their onboard LAN MACs programmed in the BIOS, so when BIOS flashing, certain blocks are not reprogrammed as they contain these values. When replacing with a hot-flashed blank chip, or doing some flash recovery by flashing a raw image, certain users found their machines not to be able to communicate on the network, at which then they discovered all their machines claimed to have a MAC of 00-00-00-00-00-00 or FF-FF-FF-FF-FF-FF!

      – Gough

  2. numeric says:

    Gough, I love you. About ‘Intel/IBM 19K5544 Pro/100 PCI-X Adapter’ you are so wrong, that it hurts. How can you look at that thing, and call it a PCI-X card? How do you explain a 64-bit data bus, using thirty-nine electrical contacts? That is not a PCI-X card, nor conventional PCI card, although I guess it appears as a logical PCI device when connected to compatible platform. I never saw that one before. It looks like an Intel Platform technology, an add-in which connects to chipset logical port.

    It looks something like CNR, ACR, AMR for full formfactor; and MDC for mobile. Intel chipset Integrated Controller Hub provides many logical PCI devices in a single integrated chip, rather than a set of many PCI chips. Two logical PCI parts were usually separate chips on their own (physical) bus/port: ISA bridge (LPC ‘low pin count’), and network “chipset” (single chip, as we see).

    An example of what I describe is in IBM ThinkPad 2001-2002. (ThinkPad 2000 used Ethernet mini PCI cards, such as 3Com; ThinkPad late 2002 integrated Ethernet in system board, no longer using daughter card). When EDC is installed, it appears to software as PCI device on same bus as rest of ICH.

    A tip about IBM part numbers: in the IBM universe around 2000, parts often have three numbers. Usually (but not always), each number is seven digits, usually two decimals, one letter, four decimals. Part number is often shortened to PN or P/N. ‘Marketing PN’ is same thing as ‘Option PN’ (as in ‘Options by IBM’™: branded peripherals, parts, and accessories), it can be thought of as “retail box PN”. ‘FRU PN’ is for replaceable part assemblies, such as a drive, cable, or NIC. (a FRU may contain smaller pieces without numbers, such as a remote control has removable battery cover, and a new NIC comes with sticker labels) ‘ASM PN’ seems to refer to per-manufacturer, or manufacture revision, parts.

    Imaginary example: a IBM Portable CD drive. It has a marketing/Option P/N, inside box is many parts: drive unit, power supply unit, travel case, computer data cable, basic bundled headphone. Each of those pieces has a FRU. IBM might source the drive from multiple makers; a Sony-produced unit has one ASM PN, TEAC another ASM p/n, and the part from HLDS a different ASM pn. Same goes for the black wall-wart AC adapter: made by one of several switching power supply manufacturers, one ASM PN assigned per-source, but the wall warts are all functionally identical, all same FRU number. Or, a Option might contain different FRUs in different world regions.

    Regarding your specimen: I do not know what is text in copper 19K5544. ASM PN is 19K5543. FRU PN is 19K4885.

    I do not mean to hurt your feelings. I make false statements frequently.

    • lui_gough says:

      No, by all means, I appreciate these sorts of comments because I too had doubts about that particular card when I received it and when I catalogued it. It does physically look like a shortened PCI-X, and other second hand parts vendors claimed it to be, hence my confusion. I did link the datasheet which does say it’s a Platform LAN Connect chip, so you are right, it’s actually intended for ICH2 bus connection. I initially thought “Platform LAN Connect” was a family of chipsets, rather than an interconnection bus, hence my confusion. Thanks for the correction and for the information that will be handy in the future.



Error: Comment is Missing!