Question about Lynnfield IMC limitations (4 DIMMs)?

[WindowsRevengeII]

I recently acquired up a Xeon X3460 (essentially the same as an i7-860). It's a Lynnfield processor (I know not cutting edge lol but picking up parts on the cheap) and I was wondering about the memory controller limitations which I think I've come across on my own... (Sorry this turned out as a longish post so I bolded the important parts.)

So I got a DDR3 kit here I'm using with the above CPU on an H57 board. I got the RAM pretty cheap but the thing is it's a "Quad Channel" HyperX kit, rated for DDR3-2400. So it's 4x 2GB DIMMs giving a total of 8GB.

First off OCing Lynnfield has been "fun" to say the least, lol. Almost as much "fun" as OCing Engineering Sample P4s on DTX boards with limited power supplies back in the day. Meaning of course, it's anything but straightforward and has been a rather large headache let me tell you.

Anyway very long OC story short(er), I got the CPU running well/stable at ~3.6Ghz with HT on and very close to stock voltage (which for the chip I have the VID is 1.26250--fairly high but that's what it is). I have BCLK at 172 x 21 multi = 3612Mhz. I'm not interested in going any higher because I don't want to raise the voltage (and thus power consumption and heat) much higher than stock. CPU also will absolutely not run with even 10MHz OC on BCLK with LLC enabled (which is a whole other story)

Back to the main point though... Thing I found though was I couldn't get the RAM going very fast at first so I just used the 6x multiplier to keep it low while I figured out what the CPU could do. When I went back to the RAM clocks after, I found that there was no way I could get the RAM over about DDR3-1750 (875Mhz) and stay stable...even with 1.35V Vtt/QPI and 1.7V DDR. Trying to boot with anything higher than DDR3-2000 would simply not boot...at all. Kinda odd for "DDR3-2400" RAM...or at least I thought at first. Trying to stress test at 1750 would either crash/freeze or BSOD, so that wasn't working either.

There's two XMP profiles on the RAM:
1: DDR3-2400 1.65V, 1.35V QPI, 11-14-11-30
2. DDR3-2133 1.65V, 1.3V QPI, 11-12-11-30

When I couldn't even get the 2133 profile to boot even with the CPU at stock clock, I started thinking the RAM was defective. But then that made no sense really. Started trying a lot of different things and then it hit me...this is a Quad Channel kit, made for later/higher-end i7 CPUs. I'm trying to run the four DIMMs in dual-channel, which is a totally different story apparently!

Sure enough, if I pull two of the DIMMs and go down to 4GB (in dual channel), I can easily boot at over 2000... Even 2133 no hangs, no BIOS resets, nothing.

So I'm guessing what I'm trying to do (get the RAM running at "stock" speed) is impossible on a Lynnfield CPU? The IMC just can't do it? I thought about increasing the Vtt to some stupid voltages but then decided against that 'cause the DDR speed I'm looking for it's way too far above spec to even try, correct?

Decided to settle for DDR3-1600 type speeds, and got to the clock I'm at now with four DIMMs which gives me DDR3-1720 / 860Mhz, which is pretty much pushing the limit of the IMC at reasonable voltages I think. Motherboard manufacturer (Gigabyte) states DDR3-2200 is possible but when you look a little further into it, the support list shows only 2 DIMMs in dual-channel, at anything DDR3-1800 and above.

So anyway, this is the whole "problem" in a nutshell, right--it's not the RAM it's the fact that Lynnfield's IMC can't do those clocks with 4 DIMMs?

Really I guess I should probably just run with two DIMMs and/or sell the kit and buy an 8GB kit of 2 DIMMs? Less stress on IMC that way, yeah?

 

[drescherjm]

I say this is way too much of an overclock of the IMC. Remember these were designed to work with DDR3 1333 max.