Will going from 16X to 8X PCIE have a noticeable impact on the performance of the GTX 980?

Galacnor

Reputable
Sep 25, 2014
1
0
4,510
Hello.
So I'm just getting back into PC gaming (and building) after a couple years of taking it casual.
I'm currently revamping my rig and the only componates I really wanted to keep from my last build are my Processor (Intel Core i7-3770K) and my motherboard (AsRock z77 Extreme 4). I'm only keeping the motherboard because I want to keep the processor, it has bee working excellent for me and I would prefer not to have to buy a brand new higher end CPU.
I purchased a new EVGA GTX 980 a few days back and am thinking that next time payday rolls around I may get myself a second one to SLI. (I'm well aware its an unneccery move).
My question is as followed:
Since my motherboard only supports dual GPU's at 8X, will I see a noticable drop in preformance if I try to use two 980's in SLI in it? Will it give me any effects at all? Or will the drop in PCIE speed cancel out any positive effects that using a second 980 would bring?
I would not be opposed to purchasing a new motherboard if I could find one that doesn't break the bank and is in the same z77 chipset family.

Please let me know your thoughts.
Alex
 
Solution


That's a cool project. I like the info. However, I'd really be curious about the impact of going down to x4 on both 2.0 and 3.0. Since 3.0 x4 is pretty much the same effective bandwidth as 2.0 x8, I'd figure it would be the same there. The 2.0 x4 would probably be the most curious thing. If there was no difference on that, Nvidia might want to rethink their SLi limitations.

The last time I was any PCIe bandwidth problems for a video card, it was going down to PCIe 1.1 x4 before it was even noticeable.

Edit: Ha! I found something on that. http://www.behardware.com/art/imprimer/836/ It seems the step down to x4 has a...

Eximo

Titan
Ambassador


Haven't seen any PCIe benchmarks for these things yet. PCIe 3.0 at 8x is still a lot of bandwidth. I don't imagine it would be much of a noticeable difference if you had dual 16x.

Regardless it would still be faster to have the two cards as compared to one. As long as you have multiple monitors or a high res/high refresh setup.
 

dgingeri

Distinguished


That's a cool project. I like the info. However, I'd really be curious about the impact of going down to x4 on both 2.0 and 3.0. Since 3.0 x4 is pretty much the same effective bandwidth as 2.0 x8, I'd figure it would be the same there. The 2.0 x4 would probably be the most curious thing. If there was no difference on that, Nvidia might want to rethink their SLi limitations.

The last time I was any PCIe bandwidth problems for a video card, it was going down to PCIe 1.1 x4 before it was even noticeable.

Edit: Ha! I found something on that. http://www.behardware.com/art/imprimer/836/ It seems the step down to x4 has a pretty serious impact.
 
Solution

ancient-geek

Commendable
May 10, 2016
4
0
1,510

PCIe is based on a point to point packet switched bus. In other words like Ethernet without the contention. When an adapter wants to use the bus it collects as many lanes as it can. In other words a 16x adapter may not get all 16 lanes, maybe it only gets 1 if that is what is available. FYI: What passes for IP addresses are assigned dynamically at power up time. So the performance question translates into how many lanes does your motherboard provides. Also note that lane performance is measured in transactions per second, not bits per second because the number of bits is largely irrelevant to throughput.