Q6600 overclocking progress

TIMUH!

TIMUH!

TIMUH!

Associate
Joined
19 Nov 2003
Posts
371
So, I start frying my nice new chip on my nice new rig today. Will use this thread for my triumphs, failures, tears and tantrums throughout the day but mainly to post progress and ask for help!

So.
Q6600 G0, TRUE heatsink (lapped)
4 x 1gb Corsair DDR2 6400 (5-5-5-15)
Abit 35pro BIOS 14

Using Orthos, Core temp and CPU-z

So I ran small FFT's on orthos to just get a starting picture and after 30mins under load Core #0 & #1 are at 42c and core #2 & #3 are at 28c. quite a difference. Could this be the heatsink although I've checked it and it seems to be sat fine. Or the CPU itself not being flat?

Next will be the first step to hopefully 3.2 - 3.4 clock, back soon
 
Thats ridiculously low temps for prime, even at stock.

Some of those chips are dodgy yeh. There should be a small difference, usually no more than 5C between cores, but like mine you can have 10-12C.

However, the fact that it's evenly spread is wierd, hot on one side, cold on the other, at first glance id say the heatink/CPU was concave or something.

My temps idle would read roughly (OCed):

Core 0: 33
Core 1: 30
Core 2: 21
Core 3: 29

Something like that, bloody annoying i can tell you.
 
Well the thermalright 120 ue is lapped so that is nice and flat. I may have to lap the chip as well. see how we go.

Currently small fft stress testing 300 x 9 on stock volts (2.7) nice and slow, will be upping the FSB in steps of 10 to start with.

(Also I sacrificed a chicken (Kentucky fried, couldn't find a real one) this morning to the OC gods, everyone does that right?)
 
Hmm, let you in on a little known secret, haven't seen it documented much, stubled upon it by accident:

The Thermalrite 120 is made convex, as in formed outwards slightly. This is a manufacturing design as many Intel chips are produced concave, therefore increasing the contact. Several other major heatsinks do this also.

So in summary, if you lap one, you gotta lap the other, as i bet your chip is as flat as a...kentucky fried chicken :)
 
Ok, so at the early stages I'm not going to stress test for hours and hours, I will more as I approach the clock I'm after.

300 x 9 = 2.7
Cores #0 43c, #1 38c, #2 37c, #3 42c
30 minutes of Small FFT's stress cpu
ram currently on 1:1.20 (720) so well within limits

Onwards......
 
ok, progress and questions

320 x 9 (2880.5mhz) 1:1.20 (DDR2 768) on the RAM
Still using stock volts on everything.
30 mins stable on Small FFT's test

Core #0 46
Core #1 46
Core #2 32
Core #3 32

Something not right there at all I'm thinking, 14c difference is not good. May have to reseat the TRUE, certainly not lapping the chip yet. Also the two hot core lose 10c+ instantly when I stop the test.

Also CPUID Hardware Monitor is reporting these same temps as Core Temp is.

CPU-z is showing the CPU core voltage to be 1.290/1.3 although this is set at 1.3250 in the BIOS?

Is it normal to not need to up any volts at this stage?

Do I need to sacrifice another chicken?
 
If the temps are not like that idle, then i'd say it's definately the seating on the heatsink. Mine stay with a 9-12C difference at all times, it's just poor contact in the chip.

Vcore depends on chip, but that's still a low overclock, so probably normal.

As for the difference in BIOS and CPU-Z, that's normal, it's Vdroop. It's kind of a safety mechanism so you don't overvolt too much, or along those lines. Some boards liek mine can reduce that a little, but usually 1 nothc is normal (Mine is 1.45v in BIOS, 1.43v in system)
 
Thanks V3g3tto

One more clock and I think I'll reseat the heatsink. At idle Cores #0 & 1 sit at 33c and cores #2 & 3 sit at 28c. Looks like that issue is just exacerbated when load hits.
 
I've heard before that it's load temps you wanna watch for in regards to heatsinks, if they become irratic then it's usually the source of the problem.

The gap between my cores increases a couple of C under load. Your cores seem to be hot on just one side, so that sticks out for me, mine its a chip fault as it's just 1 core not being used properly or the other's not seated as well.
 
ok, 333 x 9 (2997mhz) for a nice 3ghz milestone. DDR2 @ 1:1.20 792
Still stock volts on everything and the cores are the same temps as the last clock
30 mins SMall FFT's test

Have to say its been a long time since I clocked a PC my old build being the trusty Athlon XP mobile 2500 chip on the NF7-s board which is still going strong but getting a little tired these days.

So far so good I think, 600mhz clock with no real drama at all seems pretty good to me.

Next job is to reseat the heatsink and hopefully fix the cores issues and then press on to hopefully 3.4ghz clock.

Back soon.
 
every one does it different I suppose, I would start at a speed that seems to be an average for these chips which is 3.5ghz with some nice volts, and work down from there if it crash's, if its stable then lower the voltage, till it locks up. then increase volts until you find the sweet spot.
Surely quicker than working up little by little.
 
Banny said:
every one does it different I suppose, I would start at a speed that seems to be an average for these chips which is 3.5ghz with some nice volts, and work down from there if it crash's, if its stable then lower the voltage, till it locks up. then increase volts until you find the sweet spot.
Surely quicker than working up little by little.
Click to expand...

Your probably correct my friend but I haven't OC'd in years and I'm happy to take it slow and easy! Besides which 3.4ghz on air will do me fine!
 
ok, progress time.

Reseated the heatsink and also did the funky 'coin mod' on the TRUE that I got from another forum we all use. You place a coin under the TRUE bracket to maximise the contact pressure of the 4 screws.

Didn't help that much to be honest, still have the gap between the pairs of cores but it is less, I put that down to the better contact. Looks like I'll be lapping the IHS at some point.

(As a quick aside I got an Iiyama E2201-W prolite 22" LCD with my new rig and oh boy is it yummy to look at!)

On topic:
Hit my first wall at 350 x 9, ram still at 1:1.20 (840)

would post and get to the windows login screen and then no further.

After some volt fiddling
CPU core 1.3650 in BIOS showing 1.320 in CPU-z
DDR2 up to 2.0v
CPU VTT 1.23
MCH 1.29

Small FFT's test running now, load temps are:
Core #0 47c
Core #1 43C
Core #2 40c
Core #3 46c

Maybe the reseating helped more than I thought, 6-7 degree difference between the cores now. Interesting that the hot cores haven't actually got any hotter!
 
You need to up the vcore. If i had your chips i'd be at 3.8ghz now no sweat with the specs you got...3.4ghz with 1.3v?! Mine needs more than that STOCK!

Bump vcore to 1.4v for a laugh and up FSB to 400x9, if it don't work your doing it wrong lol.
 
V3g3tto said:
You need to up the vcore. If i had your chips i'd be at 3.8ghz now no sweat with the specs you got...3.4ghz with 1.3v?! Mine needs more than that STOCK!

Bump vcore to 1.4v for a laugh and up FSB to 400x9, if it don't work your doing it wrong lol.
Click to expand...

lol, now I feel like a such a pussy.
 
TIMUH! said:
So I ran small FFT's on orthos to just get a starting picture and after 30mins under load Core #0 & #1 are at 42c and core #2 & #3 are at 28c. quite a difference. Could this be the heatsink although I've checked it and it seems to be sat fine. Or the CPU itself not being flat?
Click to expand...

Sounds like you are only running 1 instance of orthos? orthos only stresses 2 cores, and the other 2 will be idle. Bring up your task manager to see that all 4 cores are under 100% load.

Fix this by running 2 lots of orthos, or downloading the latest prime 95 (which stresses all 4 cores)
 
V|per said:
Sounds like you are only running 1 instance of orthos? orthos only stresses 2 cores, and the other 2 will be idle. Bring up your task manager to see that all 4 cores are under 100% load.

Fix this by running 2 lots of orthos, or downloading the latest prime 95 (which stresses all 4 cores)
Click to expand...

Hi there, didn't know about that but I can't find a link that will download the new version.
 
You must log in or register to reply here.
Back
Top Bottom