[First Look] DFI 680i LT NF-680i LT SLI (56K Warning~~~)

[virtualrain]

So, here's an update on my overclocking adventures... (read back to post #879 for the background).

I started with these settings:

CPU VID: 1.45 (start at reasonable overvolt)
CPU Drive: Strong
DRAM Drive: Weak
DRAM Volts: 2.22 (just to keep RAM happy)
NB Core: 1.39 (start small)
SB Core: 1.62 (as per Tony's guide)
SB Dual: 1.52
LDT: 1.22 (no OC'ing here)
VTT: 1.40 (as per Tony/Raju's suggestion)
GTL0/1: 170
GTL2/3: 95
GTLNB: 85

Started OC'ing and running 30min. test with OCCT:

300FSB = OK
320FSB = OK
340FSB = OK
360FSB = OK
380FSB = Windows Doesn't Load

Increase CPU VID: 1.475; NB: 1.42
380FSB = Windows Doesn't Load
Roll back to 370FSB
370FSB = No Post - I assume there's a hole here back to 380FSB
380FSB = Windows Doesn't Load
Try tweaking GTL2/3 in steps of 2-3 points from 95 to 128
Try tweaking GTL0/1 up by 5
Increase CPU VID: 1.5; NB: 1.44
Try VTT=1.44 with above GTL values
Increase CPU VID: 1.55; NB:1.51
Try VTT=1.49 with above GTL values
Give up on 380... can't get Windows to load - must be a hole

Start 385FSB with CPU VID= 1.55, NB=1.51, VTT=1.44, GTL0/1=180, GTL2/3=100, GTLNB=90

385FSB = OK
395FSB = OK
400FSB = OK

Great! I'm basically back where I was last Friday (blog link). So I decide to run PRIME overnight... failed after 45min.

I just don't think OCCT is a good stability test... 30min is not long enough to assess true stability. I think it uses some form of PRIME code anyway based on the way I could see it cycle loading the cores just like different FFT sizes in PRIME.

So I've proven that my board/proc will boot and be somewhat stable with a variety of settings at 400FSB. It works at:

VTT=1.44; GTL0/1=180; GTL2/3=100; GTLNB=90
VTT=1.52; GTL0/1=180; GTL2/3=180; GTLNB=90

So I guess my processor is very flexible with GTL settings but nothing is rock solid stable.

I really don't know what to do to get a real stable (i.e. 4+hours IMHO) overclock with 380+ FSB. It just doesn't seem to be in the cards here.