never a fan of multi-GPU, but even so, i always think the x8 PCIe channels on P55 compared to x16 PCIe channels makes little difference. Tom's hardware has done a comparison in crossfire here. 161% vs 155%, that's less than 4% performance loss compared to X58. or about 4% performance gain in crossfire only for more than £100 spent on the platform. performance differences will be even less with weaker cards. currently, it seems i5 750 is definitely a good bargain. and best of all, you can do crossfire on P55/Lynnfield/LGA1156 with less than 4% performance loss compared to the big brother X58/Bloomfield/LGA1366. the former should be the platform of choice for everyone unless they want to get tri-crossfire or encode video more more than 80% of their computing time (as pointed out by a fellow Bittech forumer) i know many x58 users will point out you already spent that much money, why not just a little bit more for ultimate performance. the way i see it: unless you have 30inch monitors to drive, single graphics cards are more enough, unless you are spending so much, £100 doesn't mean anything to you, you might as well save it. or unless you are a benchmarking freak.
http://www.bit-tech.net/hardware/cpus/2009/09/14/lynnfield-pci-express-gaming-performance/1 Obviously my article doesn't Google quite as well Toms article uses newer drivers though I expect.
I realise this isn't quite as specific as P55 and X58 as chances are you arent going to be using 4x slots on a P55 board unless you go tri-SLI/CX but I still found this to be a really interesting comparison when showing how PCI-e lanes affect graphics card performance. http://www.techpowerup.com/reviews/AMD/HD_5870_PCI-Express_Scaling/25.html
Bindi's article is a comparison of how P55 and X58 manages two graphics cards. TechPowerUp have done some testing of PCIe 2 lanes by electrically isolating which does appear to indicate that there is no bandwidth issue with PCIe 2. The drop in framerate that Bindi observed could result from the performance of the P55 PCIe controller as opposed to bandwith limitation of the PCIe link.
How does AMD hold up running crossfireX? Ive just been looking at the core2 results in customPC and using that. Is doing that accurate enough? Wonder if next gen we'll see a big difference?
really sorry Bindi, forgot about your article, and it also confirms: unless you are a benchmarking freak, P55 is much more bang-for-bucks.
You find that all the time with High end vs. Premium. Premium, in this case X58 has a very high price tag and on a chart you see a preformance increase. But put that in real life situation and it's not even noticeable.
i saw this article that states some p55 boards are better than others for dual card set ups http://www.hardware-revolution.com/p55-motherboards-crossfire-sli-performance-problem/
It's fine I'm just teasing I like TechPowerUp articles! It could be driver related, or how the X58 handles the traffic better than P55s internal links? Or memory bandwidth or latency difference?
P55 doesn't even have a PCI-e controller, it's housed on the CPU die which, if anything, should speed things up a bit. As Bindi said, it's probably a mixture of the reduced PCI-e bandwidth and the various other connections being slower on P55 vs. X58.
The reference to P55 was intended to encompass the whole P55 system which includes the CPU and the P55 PCH. Bindi is quite right the drivers, internal interconnects or memory could contibute to the P55 not having the same performance with two graphics cards as the X58. It could even be as I suspected the additional overhead within the PCIe controller when it splits from 1 link onto 2. The one thing that it cannot be is the PCIe bandwidth. Unless you can find a flaw in the testing carried out by TechPowerUp.