cpu voltage w/ load testing question

J

JMSpeed

Weaksauce
Joined
May 28, 2007
Messages
90
System Specs:
-msi p6n platinum
-e6420 w/ stock cooling
-g.skill ddr2-800
-north bridge fan installed from mobo kit
-centurion 5 case
-adjusted ram voltage to 2.0(recommended 2.0-2.1 by mfr) up from 1.9 set as default by mobo
-all drivers up to date
-windows xp sp2

Current Issue:
I initially clocked the cpu to 2.8ghz stable in orthos on stock voltages, but wanted more due to a a few games I learned are heavy on cpu performance.

I am sitting at 2.9ghz right now with nb voltage at 1.275v(1.25v stock) and stable for an hour on orthos with no cpu voltage adjustments. The temps peaked at 56C in a few places but oscillated around 54C. I believe stock cpu voltage is 1.325v, but speedfan 4.32 reads it at 1.34v and cpu-z 1.40 at 1.336v. Under load the voltage drops to about 1.32v as reported by speedfan.

After everything was stable at 2.9ghz, I went for 3.0ghz using a cpu voltage increase of .0250 in the bios until I got to 1.360v reported by speedfan. Cpu voltage dropped down to 1.32v again under load, but it isn't stable. Temps never got passed 56C either.

So, I give voltage another notch in the bios to +.0375v. No-load cpu voltage was reporting at 1.34v again by speedfan and 1.336v again by cpu-z, even after the second increase. I am fairly new to all this, but why won't it go beyond 1.360v? ...and shouldn't this e6420 clock to 3.2ghz pretty easily (most with no voltage adjustments)? ...regardless of temps at the moment.
 
Excellent post ! All needed information is there and in a very readable and coherent form. Your post should be a sticky on how to post :D

I dont know, I should mention I dont have that board so I am only speculating.

One thing I can answer is that while it has happened before, getting to 3.2GHz on stock voltage is pretty rare.

Have you disabled all the automatic overclocking stuff and EIST and C1E ? (C1E was a hidden option, you have to hit F4 in the "CPU Feature" menu, not sure it still is. )
 
VDroop is pretty common on most motherboards, underload the power just drops a bit. However, your stability issue might be due to several things. First thing I would try is boosting up your FSB and MCH voltages (0.1V at least, 0.2V might even be better) as you're going to be at 400+FSB if you want to go to 3.2GHz and beyond. Usually what I do to test for high overclocks is I set the voltages I want to use, set everything to stock speeds, knock the memory multiplier as low as it will go with stock timings, and boot. Once in an OS, start up orthos to load your CPU and monitor temps with CoreTEmp. Then I use setFSB to boost up the FSB on the fly while running Orthos. I would use 5MHz increments (40MHz on the CPU) at a time, and wait a minute between each jump to see when orthos fails. When orthos fails, knock your speed down ~5% (150-200MHz) and that's usually 24/7 stable. Try that method and see how it works for you. Also use a finger test on your northbridge to make sure it stays cool. Your CPU temps are good for now, just keep them below 60C load.
 
Thanks Bill :) I did check all that from several different sources before I began overclocking. Also, I won't be so optimisitic on the overclock now ;)
________________________________________________________________________________________________________________________

Mr. K6, from my knowledge, the voltages I have control over are CPU, RAM, NB, and SB. What do FSB and MCH refer to?

You appear to be much more advanced in this topic than I am, so I wouldn't know a comfortable voltage to shoot for in your method. I usually used temps as my variable to adjust clock speed and increased voltages for stability.

Lastly, after you determine the final stable point, do you increase the multiplier back up? I don't really want to use a FSB higher than 400 if I don't have to. That would be 1:1 for my setup and stress it less for my needs. Oh, and setFSB doesn't mention it is compatible with nvidia 650i sli chipsets. Will this matter?
 
dont know about the SetFSB program, I played with it once but didnt really need it.

The MCH (Memory Controller HUB - Intel speak for your NB) is your Northbridge chip and giving an extra increment or two of voltage will not hurt it an probally increase stability in overclocking a bit. On Intel boards it acts as the memory controller so you want to watch the heat generated, when OCing placing a finger on the Northbridge/MCH (largest chip, probally with a heatsink fairly near the cpu/middle of board) heatsink and if its hot, consider hot melt glueing a fan on it. Its pretty critical to keep that MCH cool for high FSB OCing. You probally have the ability to "unlink" the memory speed from the CPU speed and that would be the way to go until you determine what each subsystem (cpu and memory) will do. Later you can link them and use a memory divider to experiment with "tweaking up" a perfect OC.

The FSB voltage has to do with setting voltage levels in the data transport buss between the cpu and the northbridge. I would leave it alone, typically playing with that causes more problems than it cures. Perhaps if you hit a wall, give it one increment in voltage but that would be all I would do until you get down to the place where you have maxed everything else out and are looking for any last little bit of OC. Actually it seems from your post you are not looking to set any records and are more concerned with a nice performacne boost but with stabiliy, a wise choice. So dont worry about FSB voltage.

As far as CPU voltage goes 1.45V is a 10% increase (little less actually) and where I recommend to draw the line because I do not want to risk reducing the life of the processor. Intel absolute Vcore max per the datasheet is 1.55V so 1.45 should not be considered extream, high but not extream. On my personal machine I do not go over 1.40V but I did try up to 1.45 and it made no difference so no reason to keep it that high.

Nvidia has an OC guide for the 680i, you will not have some of the settings but the basic procedure etc. should be very close, it is here:
http://www.nvidia.com/docs/CP/45121/nforce_680i_sli_overclocking.pdf
 
Thank you for clearing a lot of that up :)

Lucky for me, the MSI P6N Platinum came with a northbridge fan, which I have installed - loud though :(

I will adjust the northbridge and cpu voltage a little more dramatically, and if it isn't stable at 1.40v @ cpu and 1.40v nb, then I could say that something else is wrong?

EDIT - Just remembered the last time I bumped cpu voltage +.0375v (beyond 1.36v) in the bios, both the voltage reporters showed voltage at 1.34v. Checked again, and still does this even when bumped +.075v in bios.
EDIT2 - Bumping the cpu voltage via MSI's provided software works though. I have it set to +.0625v and under load the reporter shows it oscillating fractionally around 1.38v. So far, 12 minutes into orthos blend test and everything is fine at 3.0ghz with temps reported by coretemp and speedfan around 54C.
EDIT3 - 25 minutes in and cpu voltage steady at 1.375, but temps were peaking occasionally to 62C, so I canceled.
 
Well thats not an alarming temp, and you cannot (and I mean absolutely no freakin way) damage the cpu due to heat. It will just shutdown somewhere near 80C internal core temp. It was designed so we cant fry em. My personal machine is set to alarm at 65C so a long run of Orthos I would not get too upset at even seeing65C for short times but it does sound like you are going to need a better HS to get to 3.0GHz with more acceptable full load temps.

One thing you can do that will probally get you where you want to be is to replace the thermal material on the stock heatsink with good thermal paste like AS5 (Artic Silver 5) and lapp the heatsink (Google "HS lapping" ) which is easy and cheap, couple bucks for wet n dry sandpaper and a pane of glass. I will be willing to bet those two items will drop your full load temps 5C or more. Later you can think about a aftermarket performance heatsink if you wish to go for 3.2GHz, my favorite "sweet" spot and will have had lapping experience and AS5 on hand.

Take it easy on the NB Voltage, 1.4V would be my limit but I am not an expert on nvidia chipsets, just a couple of increments over stock should do. The CPU however can take it.
 
Ok, cool. It was just so short into the run and figured if it hit 62 so soon, 65 and higher would eventually occur.

However, I just completed a 3.1ghz run with cpu voltage increased to 1.40v in the bios, yet cpu-z and speedfan reading 1.34v, and it was stable for 3.5 hours with temps not exceeding 57C (spikes). I backed ram voltage down to 1.9v from 2.0v, and left northbridge voltage at 1.40v (1.25v stock / 1.50v limit) before performing the test. How do I make sense of that? lol. Do I believe what the bios monitor tells me, or what the software is telling me in terms of voltage?

I will also try upping the processor another 50-100mhz to 3.2ghz depending on temps, and then work down the voltage. Does that sound like a reasonable plan?

Also, I want to thank you for sticking with me through this. Once I get 3.2ghz stable, I will look into the thermal paste and lapping. I used to work on cars a lot, cleaning metals and paint. So, I suspect I should have a nice go at the intel heat sink.
 
Sorry if I gave too much too fast, Bill cleared it up well though :cool:.

Sounds like the NB voltage did the trick. Always check all your variables to see if anything could be holding you back. I usually set the voltage I want to work at and then go for the maximum speed possible at that voltage, the method works more easily for me. However, there's nothing wrong with doing it the other way around, have a go at it.
 
Mr. K6 said:
Sorry if I gave too much too fast, Bill cleared it up well though :cool:.

Sounds like the NB voltage did the trick. Always check all your variables to see if anything could be holding you back. I usually set the voltage I want to work at and then go for the maximum speed possible at that voltage, the method works more easily for me. However, there's nothing wrong with doing it the other way around, have a go at it.
Click to expand...

It's quite alright. You prompted a lot of my questions that I would not have thought of :) Thanks for checking back though!

EDIT - Tried to boot up at 3.2ghz, and it never got to OS - close though. This was with the settings that ran 3.1ghz fine for 3.5 hours. So, even if I wanted to use software to make 3.2ghz stable, it would never make it :( I wanted to try for higher when I got better cooling, but I guess 3.1ghz is fine for everyday use :)
 
You must log in or register to reply here.
Back
Top