i7 O/C and 6 Memory Sticks...

Arthalen

Arthalen

Arthalen

Associate
Joined
20 Apr 2006
Posts
2,033
Location
Leeds, UK
Am I right in assuming that if I fully populate a X58 board with 6 sticks for 12GB of ram, I'm going to limit my OC'ing results on a 950?

I'm looking at starting with a 950/P6X58D-E with 6GB of triple channel memory, but will be looking to got 12GB in the future.

From my (limited) understanding of i7 overclocking, I'd be looking at a 174mhz 'FSB' (or whatever its called these days) at a 23x multiplier to hit 4Ghz. Witha 1:5 Memory multiplier, thats a memory speed of 1740mhz DDR.

Am I realistically going to hit this with 6 sticks, or would I be better off going with a 1:4 memory multiplier (1392mhz DDR) and thus the cheaper 1600mhz RAM?
 
I'm running a X58 board with 6 sticks at 191Mhz BCLCK no problem. Board is a P6T, memory is OCZ1600, 920 not 950 so lower multi.
 
I found that it made chasing instabilities rather more time consuming and troublesome. 4.2ghz with three sticks was trivial, but 4ghz with six took a while. Some notes my progress here.

I don't have a 1:5 multiplier available, only 1:6 and 1:8. 1600mhz ram with all six slots occupied just isn't going to happen for me, though PaulyD has managed it. I think its limited by the integrated memory controller. If buying again I'd go with 1333MHz CAS 9 stuff, and not the 1600MHz cas8 I'm using. The ram is capable of far more than the processor is when all slots are in use.

Stability was an absolute nightmare. Here's a screenshot of it passing silly numbers of loops of IBT, but it hung consistently if I moved a couple of gb of data into the ram (starting a virtual machine). I think Biffa ran into similar stability problems, but I could be remembering incorrectly.

2va1kpy.jpg


Whether that was IBT failing to test it properly, windows itself or the overclock I never worked out. At 4.2ghz it was stable under virtualisation, but passing ibt and failing under normal use really hurt my faith in stability testing.
 
For memory and old-style Northbridge testing, I couldn't recommend using Prime95's Blend mode any higher.

All too often its easy to get a Small FFT (Purely core testing) or Large FFT (Core & L2 testing) loop of testing done and passed over 8 hours, to then see the blend test fail after 10 minutes or so due to lack of stability in accessing memory at any given overclock. On the 775 platform, a mixture of bootstrap, northbridge clock/divider, voltage and a pinch of magic pixie dust was required to get a Blend test passed and working.

So, I think I'm right in thinking that the additional stress on the built in CPU memory controller is going cause a few issues with overclocking it seems. I think having a higher multiplier CPU is the trick here, requiring less BCLK to get the overclock required. And as w3bbo mentioned, QPI voltage will need tweaking.

But, as you've shown Jon, it is possible :D And that's what I think is going to swing me in a i7 vs X6 debate.

Hmm, roll on the end of the month when I can purchase my hardware :D
 
Just to clarify I've not had issues running vmware images or large memory overheads. I'm running a p6t all I have had are normal overclocking issues. I'm currently finding on all my mobos: p6t, ds3r, p6t7 regardless of memory size need big (1.8V+) QPI voltages and less vcore to be stable. And I'm not running special low vcore, high bclck, hand picked uberchips. :)
 
My apologies Biffa, memory failing me in my old age.

1.8V qpi seems quite high, I ran into considerable problems keeping the chip cool enough over 1.45V. Fair play to you.
 
Really? That's the opposite of what I have found with QPI voltage not really having much affect in fact often letting me lower vcore and that having a positive affect. :confused:
 
I wonder if we're talking about a different voltage. Table 2.6, page 22 of this datasheet lists some maximum voltages for the i7 900 range.

By QPI voltage I mean the voltage applied to the integrated memory controller and QPI link, max 1.35V according to that table. PLL behaves roughly as you've described, and is around 1.8V, so perhaps our boards are using different terminology.
 
You know Jon you are right. My apologies, can only blame old age or lack of sleep :)

On my Gigabyte X58-UD3R

CPU PLL @ 1.84
QPI/Vtt @ 1.375
CPU Vcore@ 1.2875

On my P6T7

CPU PLL @ 1.84
QPI/Vtt @ 1.375
CPU Vcore @ 1.325

On my P6T
CPU PLL @ 1.80
QPI/Vtt @ 1.325
CPU Vcore @ 1.300
 
You must log in or register to reply here.
Back
Top Bottom