13

[copied from superuser.com]

I'm duplicating an existing compact PC; I'd love to replace it with modern components, but for various reasons I can't.

I'm having a tough time finding the correct 32-bit PCI right-angle riser card. I bought a new one, and the dimensions are the same, but the wiring is different, and the system doesn't recognize the inserted card (an audio card).

The old, working riser (green in the pictures below) has five capacitors, and the wiring doesn't look straight-through. When I toned it out, it seemed to be straight-through, although several pairs of adjacent pins were connected. It's at least a three-layer board with an internal ground plane.

The new, not-working-for-me riser (red) is pretty clearly straight-through, with no components other than the connector. I toned it, and no adjacent pins are shorted. It's just two layers; no ground plane.

Strange thing: plugging the audio card straight into the motherboard works fine, which suggests that a straight-through riser should work.

So, the question is: are there different types of PCI riser cards? How can I find one that's electrically equivalent to the old one? (And, why won't the new straight-through riser work, when the audio card works if plugged directly into the motherboard?)

Front of 32-bit PCI cards

Rear of 32-bit PCI cards

8
  • Yeah those things are delicate. I once saw Frankenstein's own PC in a Microsoft operating system testing lab for testing very early multi-monitor capability - a single PC with 3 "PCI expanders" each with 4 video cards ... well, the software did work, the hardware was a bit ... temperamental ...
    – davidbak
    May 12, 2021 at 15:45
  • 1
    It has been too long to remember the details, but PCI can drive a limited number of electrical loads. As I recall (possibly incorrectly), a PCI device counts as one load and each plugable connection also counts as one load. The use of riser cards thus increases the number electrical loads.
    – njuffa
    May 12, 2021 at 17:45
  • It looks like the green board has internal ground and +5V planes. Take a look at, e.g., the connections on C5. May 12, 2021 at 18:28
  • There's definitely a ground plane on the green board, but there doesn't seem to be a second plane. Those last two pins on either side of the end nearest the notch are +5V; they're all tied together, with a VIA from one side to the other, but there's no apparent internal connection. May 12, 2021 at 18:41
  • C5 has two vias coming off it. There are no visible connections them besides the cap (one end of which is connected to the 5V pins on the edge connector). Additionally, neither the GND nor the 5V pins on the female connector appear connected on either side of the board. This makes me think that GND and 5V are on internal planes. The board could be 3-layer with the middle layer being a ground plane pour with additional tracks (e.g. 5V) across it. Three-layer boards are pretty rare though (4-layer tends to be cheaper and easier to manufacture). Thus I'm pretty sure that's a 4-layer board. May 12, 2021 at 20:14

2 Answers 2

22

I have used these passive PCI risers quite a lot and some of them need their "fingers" to be cleaned with something like isopropanol before they work properly. Especially on the top photo the fingers on the red riser don't all look clean enough.

5
  • 2
    It looks to me like the fingers on the red board are tinned, not gold plated. Gold is a lot softer than solder. In addition to removing any dirt or tarnish from the new board, a bit of wiggling might be needed to make a solid connection. For cleaning you might also try a pencil eraser (followed by some alcohol) since it's mildly abrasive. That might help smooth out the finish. Use that technique with care on gold contacts though since it's pretty easy to rub the gold plating off. May 12, 2021 at 18:22
  • @AlexHajnal it's probably old stock too, so slightly oxidised even though solder doesn't oxidise fast or much. A little gentle abrasion is probably necessary as common solvents won't remove oxides.
    – Chris H
    May 13, 2021 at 7:32
  • 1
    For future readers: @Edders' fix worked, and the new riser is happily functioning. May 13, 2021 at 13:50
  • @DanielGriscom Was it the male (riser's edge) or female (socket on riser) connector that was the problem? May 13, 2021 at 19:14
  • It was the edge connector (male) on the riser that I cleaned. May 13, 2021 at 19:39
7

Unfortunately I can't give you specific search terms, but I can say what stands out to me about those risers from an EE perspective.

The goal of a good quality riser is to minimise impact on reliability and signal integrity and thus minimise the chance of the riser causing problems.

The Green riser appears to be trying it's best to do that. The connector fingers are gold plated as they should be. The PCB is multilayer (probablly 4 layer) with the traces run over a ground plane connected to the ground pins on the connector. The power lines are linked to the ground lines via decoupling capacitors.

The Red riser on the other hand has been made by someone who either doesn't know or doesn't care. Someone has just slapped down some connector footprints, wired them together and sent the PCB for manufacture on the cheapest process. It has then been assembled in a way that melted the mechanical reinforcement pins on the edge connector.

1
  • 1
    Yep. All PCI risers are out-of-spec, and whether one works will depend on the host PC, the riser, the PCI card, and the phase of the moon. But one of those risers was designed by someone who realized the difficulties facing them, and the other one was knocked off in twenty minutes by a summer intern.
    – Sneftel
    Mar 20 at 15:02

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .