Lynnfield - how fast can 8GB (4x 2GB DIMMs) be pushed?

Hi, can anyone tell me how far typically RAM speeds can be pushed on P55/H55/H57 systems, specifically when all 4 DIMM slots are being used? I've been playing with settings for a couple of weeks and even got a faster CPU to try the higher memory multiplier. A couple of people of other forums think it's necessary to run VTT at exceptionally high levels i.e. 1.36V-1.4V (I think it's a very bad idea to go that high) to reach 1800Mhz-2000Mhz with 8GB. However I suspect the problem I have may be more of a timings issue. System consists of:

• i5 750 and i7 860
• P7H55D-M
• Kingston HyperX 2000Mhz (8 8 8 24)

Why I think there is a timings problem: interestingly at various settings it will appear stable for dozens of runs of IntelBurnTest, often the case after manually adjusting the clock skew. Then after switching the computer off for a while, or reloading (having tried other tweaks) the previously stable settings it will instantly fail IntelBurnTest, or BSOD, or reset with a black screen and reboot. It's like the reference clock isn't quite consistent between system restarts, or some other hidden setting is automatically chosen but sometimes it's correct and other times it's awry.


Possible clue about timings or red herring?
I did notice with the DRAM timings some things don't remain exactly consistent when reloading saved BIOS profile e.g. REF Cycle Time left to 'auto' returns odd numbers, usually 105, 107 or 109. The BIOS will only allow manual settings in even intervals. Now if I set everything outside of DRAM timings manually, then set the key DRAM timings to 8 8 8 24 6 (as per XMP profile) but leave other DRAM timings on auto, REF Cycle Time doesn't always return the same value even though BCLK, multipliers, voltages etc are restored to fixed settings. It's something I don't understand, but I have tried setting manually to 110 and fixing all other DRAM timings to the values the XMP profile sets. It doesn't appear to make any difference to the overall pattern of stability then instability.


VTT not the problem?
The XMP profile chooses 1.3V for the IMC/VTT. I'm not convinced it needs all that, using the i5 750 I have observed long test runs of IntelBurnTest using all 8GB at 190 x 10 and also had some success with VTT set lower at 1.24V (though it failed when I tried to replicate the settings later) Right now I'm trying the i7 860. VCORE is highish at 1.275V, VTT 1.3V or 1.32V

• 161 x 12 for memory at 1936MHz
• 161 x 22 for CPU at 3550MHz

Completed 41 loops of IBT at maximum, then got an erroneous result (didn't hang completely) Yesterday I got 15-20 loops completed with the same BCLK but other settings were different. But if I try other less stressful BCLK or higher BCLK with lower dividers it's hopeless, even with the CPU clocked much lower. Yesterday I tried variations of BCLK between 150-160 and manual skew adjustments but couldn't get stable results.



Should it be possible to run 4x 2GB at these sorts of speeds on 1156 systems? are the long runs on IntelBurnTest just a fluke?

I am having the same issue. I have an i5 750 on an Asus p7p55d board with Kingston DDR3 2000 ram. I just purchased 4GB more and am having trouble running over 1600mhz with voltages I am comfortable with. Were you able to figure anything out?

Give these settings a try...

i am using a EVGA P55 classified 200 with DDR3-2000 Sector5 8-8-8-26 from Patriot

2 DIMMs run perfect 1000MHz and more - 4 DIMMs i only get at 800MHz stable

i was posting over at patriot forum, they told me that 4 DIMM with 4,2GHz clock on my i7 860 puts very high load on the memorycontroller of that Lynnfield chip - so it will be very tough to get that running stable..

i tired almost everything.. raising voltage significantly (vcore, vtt, vdimm) that did not help at all...

i relaxed the latencies - even 11-11-11-30 .. and p.ex. trfc 100+ did not help.. relaxed round trip delay to CHA 73 CHB 76 - no effect either

the only thing that helped to get them running all 4 stable -> speed settle back to DDR3-1600 / 800MHz - i guess we are hitting a limit of the Lynnfield here

perhapps some people claim to have DDR3-2000 stable beyond 4GHz CPU clock - but i believe that only when i see screenshots of LinX 6GB Problem Size 100 repeats without error ;-)

i demand my system to be 100,00% stable - otherwise oc is absolutely worthless