>>Just like todays platforms will run on 1, 2, 3 or 4 so will>>Nehalem however just as todays systems require 2 sticks for>>dual-channel operation, Nehalem requires 3 for tri-channel>>operation and that function has more impact than todays dual>>design. >>>But why DDR3? Shouldn't DDR2 work as well?NoDDR3 onlyDDR2 is going bye-bye

So how does tri channel work with 4 dimm sockets?Bruceb

Here's an Asus P6T Deluxe X58 with 6 DIMM slots: http://www.hardocp.com/news.html?news=MzUx...HVzaWFzdCwsLDE=These could be very interesting ;)Regards,Jim Karn

I'm a bit reserved about supporting an enthusiastic response to this new platform. Remember we started out with single channel memory > FSB interfaces way back when. Then, a new dual channel interface was engineered and we had to start buying ram in 2 stick increments. As it stands now, we are NoT memory interface limited in any way. The "bandwidth" (i.e., data transfer rate potential) provided by this dual channel interface is way beyond both what the system can utilize or the memory can provide.Remember, the Nehalem's FSB is slowing to 133Mhz. Like the P4, and now clearly for FSB/memory controllers, brute speed is not the key. The same principle (or mechanism) of data transfer between the memory and the FSB will still be employed. The result of any strategy to transfer data between the memory and the FSB is called Latency . . . or more clearly stated "How long it takes to transfer each data bit between the front side buss and the memory buss." The (potentially) board-breaking, tweaked-out memory setups we have seen can transfer a data bit every ~ 45ns. We plebeians with more out-of-the box setups (DDR2-800) can transfer a data every ~70ns. Clearly, transferring data bits at a 45ns interval will transfer more data (per second) than transferring data at 70ns intervals. Actually a 45ns "rate" will transfer 55% more data per second (70/45 = 1.55). 55%! You'd think that would have a Significant effect on system performance . . . . but it Has Not. It has only produced subjective improvements in FS performance. If this 55% was real, we'd all see it.Nehalem promises to reduce latency to 32ns, but So What?! Our current systems have virtually no use for the 50% increases that we can provide now . . . even in "just" dual channel. Can we expect that Nehalem's platform will provide such a massive system improvement that it will then be able to use its projected (70/32= 2.2) 120% improvement in data transfer capability. That's yet to be seen, but I expect it may (finally) need the bandwidth we currently have available. We see this story playing out all the time. For instance, the "Gotta have" hard-sell for SATAII's 375MB/s data transfer capability for hard drives that can only transfer at 100MB/s. The real result will be this: We will be required to buy 3 sticks at-a-time of DDR3. New stuff, and three of 'em. The DDR2 sticks will not even fit the slots! This triple channel technology will be necessary someday, but for now it a just a way to fluff this current upgrade revenue event. (Here we go!)

My statement was actually a bit backwards now that I look at what I posted.. it should have been typed, some with 4 most with 3 and the perf boards 6To be quite honest I have not looked at how the 4 slot boards I saw would work in detail. What I did see some time ago was all production design and concept with a few prototype boards displayed so what they end up releasing may be different. The 3 slots being populated are required in order for Quickpath to do its job at its designed efficiency.

I guess we'll see larger-sized motherboards for a while now. That seems a little regressive to me.I once had a video card with 32k of memory. It was about as long as an 8800GTX card.RhettFS box: E8500 (@ 3.80 ghz), AC Freezer 7 Pro, ASUS P5E3 Premium, BFG 8800GTX 756 (nVidia 169 WHQL), 4gb DDR3 1600 Patriot Cas7 7-7-7-20 (2T), PC Power 750, WD 150gb 10000rpm Raptor, Seagate 500gb, Silverstone TJ09 case, Vista Ultimate 64ASX Client: AMD 3700+ (@ 2.6 ghz), 7800GT