PCI-E lanes (x4 vs x8) any major difference?

[rex000]

Hello,

I just needed some clarification on regarding the performance of 'x4' vs 'x8' pci-e slots.

I am using an i3 2120, GTX560Ti on a Gigabyte Z68 motherboard:

I am placing a GTX560Ti in the first PCI-E x16 slot, allowing that to run full tilt. I have an old overclocked 9600 GSO that I would like to use as a dedicated PhysX card, so I am wondering which pci-e slot would be the best to place it in.

In addition to the x16 slot, it has an x8 and an x4 as well. My concern with placing the 9600GSO in the x8 slot is that if it is populated, the first slot will run at only x8 speed. (The other minor concern is that the x8 slot is right at the bottom, so a dual slot cooler with interfere with all the ports at the base of the mobo)

However, if I populated the x4 slot (closer to the middle)with the 9600GSO, the first x16 slot should stay as an x16. Correct?

I know there isn't that much of a performance between x16 and x8 for most graphics cards. I am wondering if there is a difference between x4 and x8 for a low-end card like the 9600GSO?

Thanks!


Gigabyte GA-Z68MA-D2H-B3
http://www.gigabyte.com/products/product-page.aspx?pid=3975

XFX 9600 GSO
http://xfxforce.com/en-us/Products/Graphics-Cards/NVIDIA/NVIDIA-GeForce-9-Series/NVIDIA-GeForce-9600-GSO.aspx

 

[dudewitbow]

the 4x for a physx card i think should be enough? there is a difference between 4x and 8x but it shows up more as the card gets stronger.

 

[Maxx_Power]

The 9600GSO isn't a quick card by today's standards, and it will only be doing PhysX duties, so you should be fine. The differences between 8x to 4x for only PhysX should be small.

 

[rex000]

Thanks for the replies.

Yeah, that was my theory as well. I didn't think the 9600GSO would even saturate x4 lanes but I wanted to be sure. Can anyone else comment on this? Any proofs/charts? Thanks

My next question is would I see better performance by using just one GTX560Ti to do all the work? Or would I see better performance by adding in the 9600GSO as a PhysX card.. I know it's slow and there would be no point if it bottle-necked graphic performance.

 

[Maxx_Power]

Thanks for the replies.

Yeah, that was my theory as well. I didn't think the 9600GSO would even saturate x4 lanes but I wanted to be sure. Can anyone else comment on this? Any proofs/charts? Thanks

My next question is would I see better performance by using just one GTX560Ti to do all the work? Or would I see better performance by adding in the 9600GSO as a PhysX card.. I know it's slow and there would be no point if it bottle-necked graphic performance.

Well, there aren't any PCI-E scaling tests with respect to only PhysX, due to the paucity of titles and benchmarks with PhysX support. So here is a link to Tom's PCIe scaling with graphics cards under demanding situations:

http://www.tomshardware.com/reviews/pcie-geforce-gtx-480-x16-x8-x4,2696-3.html

In the 3DMark Vantage tests, with a GTX480, at Extreme settings and 1920x1200, there is a 2% difference in favour of 16x lanes vs. 4x lanes.

Based on that data, I would say the suspicion we hold on the 4x being sufficient for PhysX is largely correct.

 

[rex000]

Best answer selected by rex000.

 

[jaquith]

I realize this Post is 'answered' but this is good info and the most through testing I've seen -> http://www.techpowerup.com/reviews/Intel/Ivy_Bridge_PCI-Express_Scaling/