**Official DFI LanParty UT P35-T2R Review/Overclock/Guide Thread**

[Raja@ASUS]

Just had a look at that link again Raja. Makes some interesting reading. Especially with regards your gtl refs, which don't stick to the default x 0.67 rule.

My chip as it happens seems to be doing quite nicely using x .67. However I think these values are certainly worth a bit of experimentation. Certainly for those trying to get the vcore down.

A question that comes to mind is with regards the NB and VTT settings. Is running these settings at 1.6+ suitable for a 24/7.

I know it's a bit of a contentious issue right now, but like EVA does with i4memory, it would be nice to see a copy of what you've posted there over here. It would also be nice to see some more testing and conclusions, much as you've already done. Call me impatient, but I look forward to that

Edit: Just an Edit. Obviously it would be nice to see tests done with some more run of mill processors. retail q6600's and the like.

May well spurn some others into jumping on board and doing the same. In addition it may well attract a few more potential customers and get the ball rolling here again.

Thanks for the link and response, appreciated

RLM

I have a tuned clock of 8 hours prime at lower vdimm than you see in the OCCT run, NBgtl tweaking brings the ability (if done right) of lower vdimm not to mention lower vcore.

In terms of 67% of VTT, remember onboard resistors are genrally of the 1% tolerance class. Take into account thermal drift (resistor value drift with ambient or onboard temp rises etc and while passing current), you will understand how things are not always A + B = C (old rgonian proverb, we need to get a book together of his stuff), we have terminations inside the prcessor and NB that are subject to this 1% of variance at the very least. Couple this with Intel speed binning, fsb walls and the like, GTL tweaking becomes something that cannot be taught, it is a game of absolute perseverance and realistic acceptance. It took me 5 attempts to get the NB GTL right, further than the OCCT stable shot with high Vdmm, this nailed down me to only having only core 4 (aka core 3) failures during prime , a further 1 digit off GTL 2/3 (ignore bios numbering, should read 0/1 - 2/3).

I have a retail B3 stepping Qx6700 here which I will test too...

regards
Raja

[RLM]

I have a tuned clock of 8 hours prime at lower vdimm than you see in the OCCT run, NBgtl tweaking brings the ability (if done right) of lower vdimm not to mention lower vcore.

In terms of 67% of VTT, remember onboard resistors are genrally of the 1% tolerance class. Take into account thermal drift (resistor value drift with ambient or onboard temp rises etc and while passing current), you will understand how things are not always A + B = C (old rgonian proverb, we need to get a book together of his stuff), we have terminations inside the prcessor and NB that are subject to this 1% of variance at the very least. Couple this with Intel speed binning, fsb walls and the like, GTL tweaking becomes something that cannot be taught, it is a game of absolute perseverance and realistic acceptance. It took me 5 attempts to get the NB GTL right, further than the OCCT stable shot with high Vdmm, this nailed down me to only having only core 4 (aka core 3) failures during prime , a further 1 digit off GTL 2/3 (ignore bios numbering, should read 0/1 - 2/3).

I have a retail B3 stepping Qx6700 here which I will test too...

regards
Raja

I'm no engineer, but I do get the gist of what you are saying. Does make me wonder, given your comments, if that sweet spot can change from day to day due to ambient temps.

Any methodology in your approach to working with the GTL refs, or it solely intuition?

My FSB on this quad seems to be limited to 470 at the moment, maybe some adjustment to nb core etc and GTL refs may give it a boost. That and the chosen multiplier.

Cheers

RLM