jaydub wrote:
> According to the manual for the Gigabyte GAP35DS4 motherboard:
>
> "When the second PCI Express x16 slot is in use, the three PCI Express x1
> slots become unavailable".
>
> I've gotten advice from others that that if I populate the second PCI
> Express x16 slot with a 1x card (e.g. capture card, network card, tuner card
> etc.), then the other PCI Express 1x slots will still be available. I was
> hoping to populate all four PCIE slots (other than the primary 16x video
> card slot) for 1x cards.
>
> However, I e-mailed Gigabyte support with this hypothetical, but they say
> that the manual is correct, period. Card in 2nd x16 slot, all PCI Express
> slots die, regardless of the type of card inserted.
>
> So, whose correct? Anyone have experience with this?
>
>
The answer to the puzzle, could be in the block diagram in the manual.
The Southbridge has a total of six lanes to use.
As near as I can tell, from the intent of the diagram, it looks like
1) The RTL8111B Gigabit Ethernet chip uses x1 lane.
http://www.realtek.com.tw/products/productsView.aspx?Langid=1&PFid=5&L...l=5&Con
2) The "Gigabyte SATA controller" is probably a Jmicron part, relabeled, using an x1 lane
http://www.jmicron.com/Product_JMB363.htm
That leaves four lanes. The configurations could be:
1) Plug in a x4 card, to the x4 slot, and it disables the other three.
2) Plug in a x1 card, into the x4 slot, and the "switch" chips should be
able to support 1,1,1,1. (It is not really a switch chip, because if
it was, the functionality claimed would be different than what is
stated. A proper packet switching chip would allow any config you
wanted, but with bandwidth implications.)
The fact that one x1 lane travels directly from the Northbridge to the
x4 slot, tells you that they intended to support using an x1 card,
at the same time as the other three x1 slots had cards. If all four
lanes had been routed through the "switch" chips, the answer would not
be as clear.
It is unclear to me, how this design figures out what kind of card
is plugged into the x4 slot. I see two rectangular chips, located
between the two x16 slots, and they could be Pericom PI2PCIE212.
Those are signal multiplexer chips, and something has to drive the
SEL line, to put them in one configuration (x4 slot) or the
other (x1,x1,x1,x1). They could be using two dual muxes, and wiring
three of the four available lanes (one lane unused). Implying the direct
connection of the root x1 lane, to the x4 slot, is so they can establish
contact with the x4 slot to figure out what it is ?
http://www.pericom.com/products/pciexpress/index.php
2:1 mux, x2 diff lanes total, two of these needed. A0/A1 is a lane. A2/A3 is a lane.
http://www.pericom.com/pdf/datasheets/PI2PCIE212.pdf
They could be using the PRSNT2 signal on the slot, to figure out
whether a card larger than x1 is plugged in. Perhaps PRSNT2 is what
controls the SEL signal on the mux chips. (The following figure shows
the pinout of a PCI Express slot.)
http://web.archive.org/web/20070322211528/http://images.tomshardware.c...2004/11
So my guess is, you can do what you want, which is use
four x1 cards. At least, based on a lot of guesses from
60,000 feet. The pictures I have of that motherboard, are
not good enough to verify whether those are Pericom chips
or not, or what their part number is.
We await your test results
Maybe you could run the question past the Gigabyte Tech Support
again, asking them to have an engineer reply, as to whether
x4,0,0,0 and x1,x1,x1,x1 is implied by their diagram.
In engineering circles, it is popular to "tease" other engineers
with terse diagrams. So their block diagram implies the board will
do what you want, without a simple english statement to that
effect (which is what a customer would prefer to see in the manual).
For someone to test this for you, all they need is two PCI Express
x1 cards, one to shove in the x4 slot, and one to test any of the
other three slots.
Best guess,
Paul
FAQ
Profile
Private Messages
Log in
