i will be purchasing a quad core cpu in due time, and from what i understand, the pros of buying a x38 motherboard is that they will support the next 45nm cpus, along with a 1333 fsb and ddr3, and support dual 16x pcie 2.0 lanes.
looking at some 680i motherboards, i see that they also support a fsb of 1333, and dual 16x lanes, although they are pcie 1.1. and from what i understand, a simple bios flash for some boards will allow support of the next 45nm cpus.
(the topic of which card to sli is not in debate here, though i am sure some will be tempted to challenge that )
in a setup of sli'd 8800GTSs, a q6600, and ddr2 800 (2gb), under full load in a game, which of these would be a limiting factor?
if the 8800GTSs we're in pcie 2.0 slots, would they be able to take advantage of the bandwidth? if not, would a bios upgrade allow them to take advantage of the bandwidth? or is it just unable to because of hardware limitations which will not be present in the g90 gpus?
what advantages does the x38 have over the 680i, other than ddr3 support and native 45nm cpu support?
also, will the 45nm cpus use the lga775 socket? and how much longer will the lga 775 stay with us until the next intel socket?
In most benchmarks Intel chipsets are a little (about 3% or so) faster than it´s nvidia counterparts. Usually the SATA works better on Intel chips. Yet i consider the difference too marginal to be of importance. I don´t know how good or bad the x38 will perform, but i expect the usual improvement.
As for the added PCIe speed, i don´t really know what to make of it. I somewhat doubt that the GTS is in need of higher PCIe bandwidth, but i won´t rule that out. On the other hand someone could SLI them through the motherboard and throw that SLI bridge in the gutter. That would be something to see.
Another benefit of the x38 is the higher power output of the PCIe slots. It won´t affect the next generation of cards but probably the one after that. Still it would be nice to see a 8800 GTS refresh running without additonal power plugs.
edit: Regarding the 45nm move, i believe that intel will go for an integrated memory controller once the 45nm refresh is finished. While that doesn´t automatically call for a new socket, it means it will be incompatible with existing 775 boards anyway so a new socket is very likely.
I am going with almost the same exact build, with the exception of the graphics card, as I am sure from your comment you know. Anyway, from what I have read I do not believe the bandwidth to make a big deal at this point. As to the 775 well I figure it should be good for at least another year, maybe. Long enough to buy one more upgrade on the CPU before the new ones roll out. At that point I figure I will have to wait at least a year for the bugs to be worked out and the price to drop to something even remotely reasonable.
(A slight off shot to my comment, if AMD is not in a competing mode with INTEL at that time than I would say it might just take longer for the prices to drop, just something to think about.)
Just get what you want when you want it. There is always something new around the corner, something faster, something cheaper, something more efficient.
The 600-series nForce chipsets already accept 45nm Penryn chips btw. THG has proven that x16 speeds are better than x8 speeds, but not by that much. Even though transfer speeds are greater at x16 than x8 it doesn't mean that any graphic card saturates all x16 just that they can saturate more than x8. So PCI-express 2.0 is probably going to be overkill for a long time to come, considering it's supposed to run up to x32 speeds.
DDR3 will be a waste of time and money until this time next year. It's too expensive and although it's faster than DDR2 it's not that much faster.