Hacker News new | past | comments | ask | show | jobs | submit login
Manawyrm/ISA8019 – RTL8019-based ISA network card, NE2000-compatible (github.com/manawyrm)
50 points by todsacerdoti 49 days ago | hide | past | favorite | 27 comments

I'm getting a real kick out of the "new hardware for older computers" scene. Presumably cheap contract manufacturing has been the enabling factor.

Along w/ the Snark Barker[1] that's made the rounds on Hacker News as-of-late, I recently learned about the Orpheus[2] 16-bit ISA sound card.

I've gotten some stuff from BMoW[3] for the Apple ][ line, and Individual Computers[4] for my Amiga gear. It's been a lot of fun.

Emulation is alright, but it's not a substitute for experiencing the real hardware.

[1] https://news.ycombinator.com/item?id=19006336

[2] http://orpheus-soundcard.eu/

[3] https://www.bigmessowires.com/shop

[4] http://wiki.icomp.de/

Being able to get a multiplayer PCB made for just a couple bucks certainly makes it easier.

Ne2000 was my first NIC and I remember it dearly. It allowed my crummy old computer to connect to the family computer and by proxy its modem and the internet. Having access to the internet from my own bedroom was a game-changer for me as a kid!

This board, along with the SoundBlaster-compatible MCA board I saw on Hackaday recently makes me think there's a sizable community for retro PCs from the 8088-486 era, like there is for the Amiga.

A smaller form-factor would be more appealing. It's been a long time since I've seen one, but the card looks a bit larger than the original. More recent PCI network cards had gotten down to one main chip, plus just enough board to plug it in and fit the riser slot.

I do wonder a bit about the utility of cloning something made in the millions or at hundreds of thousands. Is there more to this project than a drop-in replacement?

And use a $150 USB to ISA adaptor to plug it into you phone: http://arstech.com/install/ecom-prodshow/usb2isar.html

Also, don't forget to make sure to flash each eeprom with different MAC address otherwise you will have some fun to figure out why you have 50% packet loss.

It does make me sad that we're having to build new "clones" of pieces of hardware that were made, very literally, in their millions because we're so hopeless at recycling and salvage as an industry.

Not at all. This is fun stuff.

There is a booming business in recycled gear.

An old boss retired and started getting calls from other geezers who were getting calls from old clients (he was a CE for DEC and SGI way back).

He ended up starting a little business with maintenance contracts or subcontracts with other providers. A couple of years ago he was making a good buck and had 7-8 guys in the field.

Usually customers were really big or really small and had some ancient system that was too expensive to figure out. One utility depended on a cluster of 33Mhz PowerPC AS/400 boxes to do their billing — they paid a lot of money for best effort break fix. Another used 286 PCs for controlling elevators.

ISA network cards are actually readily available on eBay for about $20.

I've been selling a lot of old ISA stuff to retro computing enthusiasts here in Norway, pretty much the only thing that doesn't sell seems to be network cards!

Although not all, a great number of retro computing enthusiasts seem focused on home system configurations. For the most part home networking didn't become a thing until the PCI era.

I have a Zenith/Heathkit Z-160 luggable, an 8-bit ISA bus system, which I installed an 8-bit ISA network card in. 8-bit ISA network cards are much more difficult to find than 16-bit ISA network cards but still it was only $20 on ebay.

The RTL8019 appears to still be in production; the oldest datasheet I could find has a date of Jan 1996 on it, making it close to 25 years old.

Was a long time since I last heard about NE2000. Getting flashbacks of me creating NetWare 2 images with Novell linker, disk-jockeying 5,25" floppies while making notes about jumpers for IRQ and I/O address, trying to bind AppleTalk cards for IPX and TCP/IP, etc.

Is there even a new x86 ISA system that you could buy today?

There are quite a few, mostly meant for industry applications where there is a controller card for a specialized device, e.g.: https://www.nixsys.com/motherboard-with-isa-slot/

Though you'll find that a lot of them are single-board computers meant to interface with a bus backplane, either PC/104 (as the sibling comment notes) or... more advanced, like this thing: https://www.advantech.com/products/1-2jkn6l/pca-6028/mod_e93...

which will accept aCore i7s and 16GB of RAM, and has native ISA.

There are still 486-class single board computers (SBC) available w/ PC/104 interfaces (which is electrically the ISA bus but on a pin header). I believe there are boards that let you interface card-edge connector slots to a PC/104 SBC.

LGR has made a video recently, installing one (it had an ISA edge connector, not PC/104) with a backplane in a desktop case: https://youtu.be/Q4qJf50YSVk

NE2000 compatible sounds like “Hayes AT” compatible: a true blast from the past.

"Adlib compatible"

How long before the same thing - reimplementation or new design - of PCIe cards is possible by hobbyists?

Depends what you mean by hobbyist. PCIe-capable FPGAs are getting cheaper and cheaper, and there are plenty of devboards in a PCIe card form factor, just bring your own bitstream and I/O. Hell, you might be able to use the open source Yosys/NextPNR toolchain to target the Lattice ECP5's PCIe SERDESes some day [1].

Nothing prevents you from emulating, I don't know, an Intel PRO/1000 or RTL8169 on one of these boards, or rolling your own board design from scratch. All the datasheets are out there.

[1] - https://github.com/ECP5-PCIe/ECP5-PCIe

I even can't remember last time i've seen ISA card.

What does it mean to be a 'a modern [...] hardware design'? Newly created? Robust in some way? Uses in-production parts? I would consider being focused on an archaic driver means it's decidedly NOT modern. Maybe "Open source NE2000 compatible NIC made with in-production parts' is a better opening statement.

The ISA bus interface and NE2000 programming are not modern--although I'm sure there's some modern industrial motherboards that still have an ISA slot.

But the design of this card is modern--everything is down to two ICs and a few other tiny surface mount parts that look more at home on a cellphone board than a PC ISA card from the 80's.

Compare with pictures of actual old ISA NE2000-compatible cards (eBay a good source) and I'm sure the BOM is at least 2x of this NIC.

Plus this is open source.

Err...the only thing "modern" about this is you can purchase/fabricate one using new material and components in active production. If a modern cell phone integrated those SMT packages, it'd be the size of a brick, and the PCB itself is vanilla 2-layer Tg130 FR4...couldn't even be bothered with a controlled impedance layout, let alone length-match bus traces. Recently, bunnie talked about one "relatively modern" PCB design element---microvias---and explains why you just don't see them much in OSHW designs[1]...an excellent read if you get the chance. JLC is a well known PCB fab house because their low-volume prices are dirt cheap, not because their process is modern.

[1] https://www.bunniestudios.com/blog/?p=6011

Why would anyone bother about controlled impedance or matched trace length on an ISA card of all things? At the speeds ISA runs at you could connect everything with wire coat hangers

You've summarily conflated a signal's fundamental frequency with the significant harmonic contributions of its edges.

Take a moment to review ISA v2.01 spec[1] § 10.4:

> To minimize crosstalk between signals and bus reflections, the driver rise/fall rates must be greater than 3ns; only LS or ALS devices are allowed. Faster parts such as F and AS cannot be used to drive the bus.

If you thought it odd that normative language like this explicitly calls out tech by name, then you might have some sense of just how concerned Intel was about interoperability at the time...no thanks to wire coat hanger designers.

Also consider the language of PC/104 v2.6 spec[2] § 4.7, which is probably one of the last remaining modern platforms with native ISA bus integration:

> Whether [ac] termination is needed and where it should be located, is dependent on the specific system configuration and must be determined by the system designer.

Now consider for a moment why this language and the paragraphs that precede it exist in this independent standard. This design is a commidity peripheral that will be integrated into arbitrary systems.

Length matching literally costs nothing...but even this is apparently a "stretch goal" given the obvious fact that this design completely dismissed the normative max 2.5" stub length requirement.

So no, I disagree. For the sake of sensible SI that's too often conveniently dismissed by hobbyists, keep your wire coat hangers in the closet.

[1] https://archive.org/details/bitsavers_intelbusSpep89_3342148...

[2] https://resources.winsystems.com/specs/PC104Spec.pdf

> Length matching literally costs nothing

Well arguably it does cost some time yet is utterly pointless on an ISA card (unless your traces are 40 meters long I guess) there's plenty of pedantic things you could waste time on designing an ISA card

Usage of surface mount and fewer but more integrated ICs are isn't what I would consider modern. Modern designs use those things, sure, but the technology and practice is not modern. It's actually pretty dated. Maybe not as-dated as a through-hole and discrete component design, but certainly not exclusive to recent (read: modern) designs.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact