Gigabyte EP45-UD3P Owners Guide - Page 47 - Overclock.net

error10 · 12-15-08

Quote:

Originally Posted by error10

OK, I need to go tweak my NB, I think. Currently have it at 1.42V to drive 475 FSB and 8GB of RAM at 950 5-5-5-15. Almost everything runs fine, except the compiler keeps crashing. (And Firefox on occasion.) So I think the NB has a problem.

What should the GTL reference be for this chip? I haven't changed it from Auto (which comes out to 0.76V I think). That seems incredibly low. Should I increase this?

All right I've tuned my NB GTL reference voltage. Though it's at something like 72%. But Firefox doesn't seem to be crashing anymore, and it's 100% stable compiling the Linux kernel now (this has been known for over a decade as THE most stressful motherboard/northbridge test).

This was my methodology:

I started at 0.99V (70% of 1.42V). I then went DOWN to the LOWEST reference voltage that would POST (0.93V). Then I went UP to the HIGHEST reference voltage that would POST (1.15V). I set it to the one closest to the center of that range, rounding down (1.03V). And she seems to be rock solid.

arjo_reich · 12-15-08

Hey, randomly, anyone else notice that OCCT seems to report about 10C higher on the die temps than Everest? (BTW, the 1590 Beta corrects the misreporting of the 12v line) I'm more inclined to believe Everest by "name brand" but there's a little part of me that thinks... naah, my lap-job couldn't be _that good_ lol...

See what I mean?

BTW, for a 3.0GHz overclock that's still at stock voltages for the CPU and RAM is so much fun. Should I be concerned that I'm closing in on 95W consumption (bottom of image)?

Ghostleader · 12-15-08

Quote:

Originally Posted by holtzboy

I looked through about 4 different Q6600 threads on there and tried various settings and none of them worked. I even loosened and rasied most of the values above just to try it out and nothing would post. I am thinking it might be because of the 8GB of ram.. Here's the settings I tried unsuccesfully:

Have you tried to set CPU Clock skew to 300 - 500ps and NB Clock skew to 100 - 300ps?

What seems to happen is the clock advances as you increase the FSB speed, it “drifts” so you have to dial it back, so use a delay ~300 - 500 for CPU Clock Skew and ~100 - 300 for the NB Clock Skew seems to be the best.

If it drift to much you get instability.

For the CPU, higher FSB, more deley.

For the NB, higher FSB and higher load (example: 4x1 or 4x2 DIMM), more delay.

stn0092 · 12-15-08

Quote:

Originally Posted by error10

This was my methodology:

I started at 0.99V (70% of 1.42V). I then went DOWN to the LOWEST reference voltage that would POST (0.93V). Then I went UP to the HIGHEST reference voltage that would POST (1.15V). I set it to the one closest to the center of that range, rounding down (1.03V). And she seems to be rock solid.

There's a max where it stops POSTing? Hm. I might try that later to see if a stronger ref will let me lower the voltages on the northbridge.

holtzboy · 12-15-08

Quote:

Originally Posted by Ghostleader

Have you tried to set CPU Clock skew to 300 - 500ps and NB Clock skew to 100 - 300ps?

What seems to happen is the clock advances as you increase the FSB speed, it “drifts” so you have to dial it back, so use a delay ~300 - 500 for CPU Clock Skew and ~100 - 300 for the NB Clock Skew seems to be the best.

If it drift to much you get instability.

For the CPU, higher FSB, more deley.

For the NB, higher FSB and higher load (example: 4x1 or 4x2 DIMM), more delay.

I have not tried setting them that high yet so that might help out a bunch I hope! Right now I have my system at 420*9 aka 3.78Ghz which is pretty good for a quad on air. If I set the FSB any higher than that Prime95 would error within 10 minutes of testing so maybe with those higher skews it will be more stable.

stn0092 · 12-15-08

Quote:

Originally Posted by Ghostleader

Have you tried to set CPU Clock skew to 300 - 500ps and NB Clock skew to 100 - 300ps?

What seems to happen is the clock advances as you increase the FSB speed, it “drifts” so you have to dial it back, so use a delay ~300 - 500 for CPU Clock Skew and ~100 - 300 for the NB Clock Skew seems to be the best.

If it drift to much you get instability.

For the CPU, higher FSB, more deley.

For the NB, higher FSB and higher load (example: 4x1 or 4x2 DIMM), more delay.

Any chance setting skews will allow me to run lower voltages through the NB? I have it set on an uncomfortably high 1.58v. I'm running 470x9; BIOS F7a.

error10 · 12-15-08

Quote:

Originally Posted by stn0092

There's a max where it stops POSTing? Hm. I might try that later to see if a stronger ref will let me lower the voltages on the northbridge.

Yep. I basically made this technique up on the fly, based on the NB GTL reference tuning method for nvidia boards that's floating around. I think ericeod has it. The concept is the same, so why not.

holtzboy · 12-15-08

It seems my setup does not like anything above 420 FSB so I am settling at 3.78Ghz on my Q6600. When I set the CPU skew at 150ps or above the computer would not post or even get back to the bios, I had to take out the battery to reset it. Same thing went for the MCH skew, anything above 200 and it didn't like it.

error10 · 12-15-08

I've never touched CPU clock skew or MCH clock skew, nor felt the need to. Not yet anyway. Nevertheless a Q6600 at 3.78GHz is a complete beast. Mine will only run up to 3.6 before the voltages get too high.

Ledge68 · 12-15-08

Well I made it to 580FSB so far.

		Overclock.net - Overclocking.net > Intel > Intel Motherboards
Gigabyte EP45-UD3P Owners Guide

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)