For those that are not quite aware, with DFI Nforce4 mobos such as the Ultra-D, SLI-D, SLI-DR, they have 2 x PCI-e slots. Lets be clear that for this discussion we ARE NOT TALKING ABOUT SLI.
Such owners of these MOBOs, have the ability to run a single GFX card in the normal x16 PCI-e slot, of they can run it in the lower slot (PCIe slot 4) which by default will have a bandwidth PCI-e lane of x2.
Now, one can set the jumpers on the MOBO for each of these slots from "normal" to "sli" (regardless of if you plan to or even can run SLi) so that the normal x16 higher slot goes from x16 to 8x, and the lower slot goes from x2 to x8 as well.
This is really for the use of SLi. However, as explained above, a lot of people use this method to set the lower slot at x8 so that they can run their card in this slot with no performance hit from x16, in order to free up the chipset area for passive sinks.
Why would there not be a performance hit I hear some of you possibly ask? Well, the idea is, that a lot of todays generation of GFX cards as they currently stand, are not powerful enough to warrant the full use of the PCI-e x16 bandwidth and hence would not see a hit running at 8x. In fact, Toms Hardware did a test about a year ago, and they found that cards then only really required a PCI-e slot at 4x.
I would like to hear of others findings to see if you experienced a performance hit. I have read on other forums misleading info. Some say there is NO performance hit. Some say there is. Some say that it is very very minimal at like 1% max.
Personally, I had to test this for myself. On my system currently I set it to a conservative clock (in that the CPU can do much more at 11x) of:
10x285 HTT 2.85ghz Opteron with ram @ 285mhz @ 1:1
The GFX card is an x800 GTO2 unlocked to 16pipes clocked to 520/600.
I get:
EDIT!!!
See my post further down as I have rebenched everything now PROPERLY and the results are all below.
Such owners of these MOBOs, have the ability to run a single GFX card in the normal x16 PCI-e slot, of they can run it in the lower slot (PCIe slot 4) which by default will have a bandwidth PCI-e lane of x2.
Now, one can set the jumpers on the MOBO for each of these slots from "normal" to "sli" (regardless of if you plan to or even can run SLi) so that the normal x16 higher slot goes from x16 to 8x, and the lower slot goes from x2 to x8 as well.
This is really for the use of SLi. However, as explained above, a lot of people use this method to set the lower slot at x8 so that they can run their card in this slot with no performance hit from x16, in order to free up the chipset area for passive sinks.
Why would there not be a performance hit I hear some of you possibly ask? Well, the idea is, that a lot of todays generation of GFX cards as they currently stand, are not powerful enough to warrant the full use of the PCI-e x16 bandwidth and hence would not see a hit running at 8x. In fact, Toms Hardware did a test about a year ago, and they found that cards then only really required a PCI-e slot at 4x.
I would like to hear of others findings to see if you experienced a performance hit. I have read on other forums misleading info. Some say there is NO performance hit. Some say there is. Some say that it is very very minimal at like 1% max.
Personally, I had to test this for myself. On my system currently I set it to a conservative clock (in that the CPU can do much more at 11x) of:
10x285 HTT 2.85ghz Opteron with ram @ 285mhz @ 1:1
The GFX card is an x800 GTO2 unlocked to 16pipes clocked to 520/600.
I get:
EDIT!!!
See my post further down as I have rebenched everything now PROPERLY and the results are all below.
Last edited: