P45 x8/x8 Crossfire bottleneck impossible?

Chang Noi

Distinguished
Nov 7, 2008
5
0
18,510
I realize this has come up for discussion before, but the idea that the Crossfire PCI 2.0 x8 on an P45 MB would limit, for example an HD4850 just doesn't make sense.

PCIe 2.0 x8 is just as fast as PCI-e 1.1 x16, yet no-one is talking about that 1.1 limiting a 4850 or 4870. In fact, the consensus seems to be that there's no way the card could max out the slot. So why would it happen in a situation with exactly the same bandwidth?

And if it could, wouldn't that mean that the 4870x2 would max out it's full PCIe 2.0 x16 slot? After all, it would be the number of GPUs and the same bandwidth as with x8/x8 Crossfire.

Could there be some way to account for tweaktown (and others) benchmarks other than bandwidth? Do the x38/x48 use a different Crossfire implementation?

What do you think about this?

 

Zecow

Distinguished
Oct 29, 2008
402
0
18,780
Coz no one is using pci-e in crossfire anymore, if they are - its the old technology card.

PCI-e 1.1 era, the max resolution is not more than 1600x1200. Now we go beyond 1900x1200. There isn't enough data bandwidth for the data at that resolution. It will take a hit.

4870x2 on a 16 data path vs a 4870 crossfire on a 8x8 data path.. i think u can do the math

if the 4870 requires 10x on the data path - but the lanes allow only 8.... ermmm you do the math.

You don't get performance hits on the P45 unless you are playing on res beyond 1900x1200. There's difference of only a few frames at 1900x1200. The higher you go, the more data bandwidth is required.
 

tankster

Distinguished
Jun 9, 2009
16
0
18,510
but ppl still use pci-e 1.0/1.1 for single card solutions...

so what the question is... is that b/c pci-e 2.0 x8 is the same as pci-e 1.0 x16 ... when a p45 is in crossfireX it must not be bottlenecking..?