Riser Cards

[jmac698]

Riser Cards

[jmac698]

Background

A slim case (Height<100mm) with full height PCI cards, must use a riser to lay the cards horizontal.  Usually there is 1-3 cards supported.  A mATX size mainboard has 4 card positions.  They are usually agp/pcix16, pcix1 (slot3), pci, pci (the last one is at the edge).
However, the riser card is meant to plug into slot3 (with additional cable to slot2 in order to properly handle two pci on the riser).
So in most cases (especially with integrated mainboards), it's not possible to use the PCI slots with the riser, because the riser cannot plug into slot3, which is not a PCI slot.
For example, silverstone lc11 case uses a riser card.
Example, Asus A8N-VM CSM has good layout (x1 pci pci x16),
Gigabyte GA-K8N51GMF-9 does not (x16 x1 pci pci).

|++------ pcie video card
|   |------ pci card
|   |------ pci card
4  3  2  1
card positions
"++" is spacer on riser card to push the pci-e card to the right.
"|" is a riser card; there are two shown, one for pci-e in slot4, one for two pci in slot3.

The Question

For such a slim case, if I use the wrong slot #2 (one to the right, as looking from back), now my cards run into case.  But what if I use low profile cards?  This gains me about 21mm more space.  Does anyone know if these will fit?
The other solution, is to find a riser which is meant for slot2, or a riser using flexible cables.  Do you know where to buy this?  
Thanks.

[jmac698]

It's explained better here http://www.sysopt.com/features...le.php/12024_3567526_3  .  In short, I have one of those incompatible mainboards.  My proposed solution, is use shorter cards (low profile), to make up for the riser being placed in the wrong slot (closer to the edge, less room for cards).  I just need someone who has one of these cases to make the measurement or guess if it can work.  Thanks.

[mikevalla]

Hmmm.... if I were to make a guess, I'd have to say that it'd work; however, there is the possibility of IRQ issues, and PCI brackets runnin into each other, so you would have to mod them to 'daisy chain' them together (you know, the metal on the card that attaches to the case).

[jmac698]

I checked the standards, short bracket of a low profile card is 79.2 long, full bracket 100.33mm, diff 21.13.
PCI slots are about 20.4 apart (measured).  If you move the riser one slot to the right, the bracket is now .8mm too short of where it should be.  You could plug the card in loosely, or bend the metal, and it should work.  However, the standard says the screw hole on short brackets is 1.35mm closer to the fold (and case).  This is minor as well, the bracket just doesn't fit snug against the screw.  The only disadvantage is a gap on the left side of the bracket, but that's just more room for airflow.  I still need to check theory to practice!  About about IRQ: good point, but the riser in the article above has a cable that plugs into a 2nd pci, on this cable are connected the INT lines, in other words the only signals which are different between two PCI slots.  The other pins are all connected together in common.  Thus the riser completely emulates two separate slots.  There are some miniITX boards however specially designed to share IRQ (which is not impossible).

[mikevalla]

Okay, I was concerend that you were going to use one of those PCI risers that split one PCI slot into two.  You can slap me for not checking that link you posted.

Out of curiousity, it may help to 'bolt' the similar low rise PCI brackets together, in order to ensure they are straight.  DOn't know if it's neccesary, and it would certainly make for a bigger pain, but... it's out there.

[jmac698]

Well, I've now done all the research I can.  The Riser card and ATX standards are at: http://www.formfactors.org/FFDetail.asp?FFID=1&CatID=3
The riser is meant to go in slot6(atx)/3(matx).
There is a proper way to split one PCI into two, if the mainboard has an extra 22pin connector inline with the PCI slot, but in reality no mainboard has it (unless some mITX).
The defacto solution runs a few wires from a 2nd PCI slot to the riser, then the riser can properly support two cards.
If the riser splits the PCI, there are no wires from another PCI (or mainboard has only one PCI), and the 22pin connector is not present, then there is not a proper split PCI.  However some mainboards have bios accomodations for sharing IRQ, used in some mITX boards.

So, the case and riser manufacturers did the right thing, but half of the mainboards have the wrong layout.
All mainboards by ECS have proper layout
My suggestion to manufacturers: x1/pci pci pci x16

[mikevalla]

So, does that then mean that you have to try something like this: (sorry, but I stole your little image there)

|++------ pcie video card
|
|  |------ pci card
|  |
|  |------ slim pci card
|  |  
|  |<|  |------slim pci card
|  |   |  |
4 3  2  1

with the slim PCI butted right next to each other then?  With (obviously) no split PCI risers?

[jmac698]

You have to paste this into notepad, sorry.

Normal Arrangement,
eg Asus A8N-VM CSM mainboard + Coolermaster CM Media 260
|+++======  PCIe video card
|  |======  full PCI card
|  |======  full PCI card
|  |--|     splitting cable
4  3  2  1
x  p  p  x  (slots: pcie-x16 pci pci pcie-x1)

Notations:
"+++" spacer to bring pcie card to the right
"======" full length pci card
"|" a riser card
"--|" a cable from the riser card to a pci slot; for splitting

Proposed Arrangement,
eg MSI K8NGM2-FID + Coolermaster CM Media 260
|+++====== PCIe video card
|     |=== LP PCI card
|     |=== LP PCI card
|     |--| splitting cable
4  3  2  1
x  x  p  p (slots: pcie-x1 pcie-x16 pci pci)

Notations:
"===" low profile pci card

Notice the empty space to the left of the low profile cards now, which used to be covered by a full bracket.  This is good for airflow, bad for EMI, and may provide the hole needed to connect a parallel VFD out to the back of the case.  But at least the cards are secured.
Not using a riser is not an option because there is no cutouts to put the back of the cards into.  The cutouts are horizontal, which is why you are using riser cards.

[mikevalla]

Okay, I see now.  That should be fine, but still, some modification may be necessary, it seems, to the LP PCI card bracket if things don't match up well.  I don't think it'll affect EMI much.

[jmac698]

Ok.. well, the EMI thing is mentioned in the ATX spec.  Anyhow, when a wave goes through a slit, it starts coming out the other side in a circle.  The tiniest hole can let radiation escape, especially with the high frequencies in a computer.  There even used to be programs you could run on your computer to play music on a nearby AM radio.
http://www.erikyyy.de/tempest/ , though I can't find exactly what I was looking for (there's an even easier way).

[jmac698]

http://www.risercardshop.com/  I'm in heaven!   Why didn't I find this before?

[mikevalla]

I've been looking for that kind of store myself... you're the hero now!

To get back on subject- yeah, there is the risk of EMI, but as the cards will be further away from the motherboard than in their usual orientation, I don't expect that it shall become much of an issue.  Are there any toroid coils right near the lowest PCI slot?  MOSFETs or capacitors?  If not, then the risk of EMI leakage is very much reduced.  However, if you are intent on plugging those holes, then all you would need is some scrap steel or aluminum, and some electrical tape.    It's not hard, but it sure ain't pretty.

[jmac698]

Yes I agree with you, I even have some self adhesive copper foil I wasn't sure which EMI you were referring to at first, I was thinking http://en.wikipedia.org/wiki/Faraday_cage Faraday Cage, where if something is completely enclosed, there's no way for EMI to get in or out.  As for flexible risers, one store did mention there may be a signal integrity issue with high bandwidth cards.  Maybe this is due to the impedance of ribbon cable at high frequencies? As for toroid coils, they are very low EMI due to their shape.
http://www.electronics-tutorials.com/basics/toroids.htm "Some of the advantages of toroids are: ...No interaction or coupling with adjacent components (unlike air wound and other inductors)."
As for EMI with the cables, they just have to make sure every other wire is GND to ensure a good conductor for the RF.
As for gaps in the case, my understanding was that there's a relation between size of the opening and frequency that can escape.
http://www.elmac.co.uk/desfremc.htm
"Shielding effectiveness is determined almost entirely by the ratio of the wavelength of the highest frequency of interest to the dimension of apertures in the shield, so these apertures or seams should be minimised."  A 3cm hole would seem to make a difference.
Beyond that, I don't know anymore, if I'm right at all

[mikevalla]

I'd have to say you'd be all right.  There are bigger holes in the caes, where the fans go, after all, right?  But, in all honesty, it would be best to plug those holes, if only for airflow's sake.

And again, thanks for that site.  I'm going to research them a bit, but I may end up buying my risers from them.

[jmac698]

Why can you see through a microwave door?  Because light is small enough to get through but microwaves aren't.  Cases use many small holes to let air through without letting EMI out. Anyways, there's cheaper riser card stores.  Just search for it.
Later.