drdeathxThese are all extreme overclocks. I would like to see the same person use a board with the PLX chip. Marketing at its best I guess.

KonceptzI only buy EVGA or MSI boards.

EarthDog:confused:

If it does touch a PLX chip for some reason on single GPU, the added latency means it could be a bit slower anyway. I dont understand the point of that statement...why would you want to see that?

EDIT:I mean, that is like asking that runner Usain Bolt (w/e his name is) to put on some Timberland boots to run the 100m, ya know? Whats the point in handicapping something as they are going for speed here.

EarthDogI'll try to find some benchmarks, but I recall seeing that Native 16x/16x (X79) vs PLX 16x/16x the PLX lowered benchmarks a bit due to the added latency of the PLX chips in some results...

With all this data transference (and that should data be going the other way to memory then the PLX chip will have to have a buffer in order to prevent data loss) the PEX introduces a latency to the process. This is a combination of the extra routing and the action of the PEX to adjust ‘on-the-fly’ as required. According to the PLX documentation, this is in the region of 100 nanoseconds and is combined with large packet memory.

Back in the days of the NF200, we experienced a 1-3% overhead in like-for-like comparisons in many of our game testing.....

(a paragraph or so down........)

The downside of this comes with a single GPU setup, whereby the added latency and routing caused by the PLX chip can reduce single card performance. The PEX 8747 chip attempts to promise a reduction in this overhead,

(a couple more below that......)

While the chip can handle multiple inputs and outputs, that eight lane restriction to the CPU could become an issue.

EarthDogwww.anandtech.com/show/6170/four-multigpu-z77-boards-from-280350-plx-pex-8747-featuring-gigabyte-asrock-ecs-and-evga



Positive? It introduces latency. Native lanes are the way to go... now, the difference is barely above negligible, but... its the way to go. Perhaps my sprinter analogy was a bit dramatic though, LOL! Please read the entire page as I know I took those snips out of context. :)

MSI Mpower is a Z77 based board and cannot support 16x/16x natively so that testing wont help as I am talking about SLI/Crossfire for the most part... but it still does add latency to single GPU setups...as negligible as that may be.

There seems to be a constancy, though, when we're under high VGA load. Not hitting a CPU brick wall or anything like that, the x8 / x8 via the native P67 chip is a faster setup. When we see lower resolutions and the CPU play a larger picture, the x16 / x16 via the NF200 is faster.


So what's faster? You know, there's probably not a clear winner when it comes to overall speed. The better question would be; so what's better? Well, the x8 / x8 setup that ASUS choose to implement seems to be. Yes, it's not always faster, but when we're all about the video card power, it is the faster setup. The times we see the NF200 setup come out ahead is when we're looking at benchmarks with really high FPS.

We can see under intensive situations like Aliens vs. Predator and Unigine Heaven, the x8 / x8 via the native P67 chip is the better option. When it all comes down to it, there's little difference between the two setups. The decision for ASUS to go down the x8 / x8 path via the Intel chip instead of the better looking x16 / x16 NF200 path seems to be the right decision.

EarthDogI appreciate that... but since we were both really talking SLI/Crossfire and native 16x/16x lane performance vs PEX 16x/16x, not sure how you can manage that with Z77 boards. You need an X79(native 16x/16x) and Z77 with PEX. Also, less than 1% is really within the margin of error so Im not sure if this would be conclusive in the first place if that is the difference.

I'll catch the results when I wake up.

EDIT: I suppose this will show single GPU latency from it like that article states. Make sure both rigs have the same hardware and same clocks all around, memory speeds and timings, chipset drivers... etc.

EDIT2: Found another link...www.tweaktown.com/articles/4147/nvidia_nf200_x16_x16_vs_intel_x8_x8_p67_performance_analysis/index10.html

drdeathxI have no idea where you get X79. This thread is Z77

drdeathxHere is MSI MPower

EarthDogYou dont seem to understand how things work to cpmare what I am talking about, or I am somehow completely missing the point here...:laugh:. I have reread the your post, then mine, and am betting on the former (or Im just THAT tired!!). :p

The only way to test the performance difference in SLI/Crossfire with and without a PEX is by using 2 different(Chipset) boards.

X79 = 16x/16x native
Z77 = 8x/8x native
Z77 = 16x/16x PEX

Out of those, the 16x/16x native on the X79 should be the fastest by a small margin over the rest as its native. The only reason it could be different is that the X79 based boards are PCIe2.0 with SB-E chips in them...vs Z77 with IB in it is PCIe3.

Ok..........the other SS........................................?

MxPhenom 216Im sorry, but EVGA boards are trash.

EarthDogAs I understand things, that isn't possible. IB doesn'thave 32 pcie lanes in the first place. So how can it bypass that chip to run 16x/16x? Without those chips, z77 goes to 8x/8x.

Though my understanding could be wrong, lol!