Overclocking Q6600 G0 on Gigabyte GA-P45-UD3P motherboard

Ryland

2[H]4U
Joined
Nov 19, 2004
Messages
2,504
It is the system in my sig. I have had the CPU running at stock for about a year now but the motherboard and HSF (Arctic Cooling 7 Pro) were just installed last weekend. I keep seeing posts on how easy it is to hit 3 or 3.2 Ghz with this processor so was wondering what I had to do. I haven't overclocked a machine in probably 10 years so am definitely rusty. The memory is OCZ's PC8500 Platinum.

Doing a bit of research I saw that the first step should be take all voltages off Auto and change them to Normal then change the FSB to 333 but what should the multiplier be set for? I also need to go in and verify that my memory is getting 2.1 volts.
 
Leave it a 9 for now.

Set the MCH voltage +.1
Set the CPU voltage to 1.325 (very conservative but should be enough to go to 3.2GHz. )

Set your memory multiplier to 2.0 (This will likely underclock your memory but until you see where your cpu will go without major tuning it takes the memory out of the picture, later you go back and find a mulitplier, probably 2.4 or 2.5, that will run it at or near its rated speed)

With the above start at 333 for apox 3.0GHz and then start increasing FSB 10GHz until you are happy. If you like you can further raise CPU voltage to 1.375V which is still way under the Intel max of 1.55V and bump the MCH voltage up a notch too.

An ideal "sweet spot" is with the cpu mulitplier at 8 and an FSB speed of 400MHz which allows you to run PC6400/800MHz memory exactly on spec with a memory multiplier of 2.0 which is how the chipset is designed (memory runs at 2x the FSB ) by Intel. If you have 1066MHz memory a 2.5 multiplier would be perfect.


In summary, 3.2GHz is a good target and what I go for in clients machines with that heatsink. Give the CPU and MCH some extra voltage, watch the meomory speed and just start rasing the FSB. Expect the machine to recycel itself on power on starts as it resets its timing etc. that is normal. Eventually if you push it hard or have something not set right it will cycle several times and then reset back to default values. Always check the memory mulitplier when that happens and then just re-enable the cpu/fsb frequency adjustments and back off the fsb some and or add voltage.

This is just a rough guide, see the sticky at the top of the forum for more detailed info but its pretty easy and simple for a moderate 3.2ish OC and nothing fancy should need to be done.
 
3.4-3.6 is the upperlimit of alot of q6600, for example I can run mine at 3.57ghz however at 1.475 volts I dont want to risk any higher to make it prime stable. So i have no problem running this at 3.45ghz at 1.468 volts stable.

3.2 shouldnt be hard at all. Just make sure to baby your memory timings, this was the most overlooked thing when i was over clocking my q6600 and in turn gave me alot of trouble.
 
I guess I have some playing to do this weekend with the settings. I assume it would be a bad thing to use the Quick Boost overclocker in Easy Tune, right?

Im only looking to get this to 3.0-3.2Ghz which should be easily reachable with this setup, right?
 
Follow Bill's instructions to the tee. That's exactly what needs to be done in order to achieve what you want. And read the sticky at the top of this forum.
 
I will hopefully get a chance on saturday to play with this. Hopefully I can hit 3.2 without needing to up the voltage much. I didn't realize that the multiplier could be changed through the bios (I haven't had a chance to look through it much, other than to verify my memory timings).

What should I use to check for stability, Prime95 and what else?
 
Here you go. If this help. This is my setup.
SPECCCCC.jpg
 
I am assuming when you talked about raising the MCH Voltage by .1 you were talking about MCH Core and not MCH Reference or MCH/DRAM Reference, correct?

Is there a reason (like slagging my processor) as to why not just try an 8x CPU multiplier with a 400Mhz FSB with a 2.5 memory multiplier to see whether it works? That is the highest OC that I want to do but wasn't sure whether that could be considered too high as an initial jump and could cause major damage...

Edit: I see that 400FSB is probably the maximum possible so shooting for that right off the bat probably could cause damage.
 
Skyless, what are your voltages set to?

Core is 1.312 (just spotted it in the image) but I don't see the MCH voltage.
 
Would it make sense to boot off of an Ultimate Boot Cd and run the CPU Burnin program so that I don't take a chance of corrupting my windows installation incase an OC is not stable?
 
You might be able to hit 3.0 on stock auto voltage with just an FSB tweak. That is what I'm currently doing.
 
You might be able to hit 3.0 on stock auto voltage with just an FSB tweak. That is what I'm currently doing.

From what I found online using "Auto" for the voltages usually ends up with the voltage higher than necessary. I wonder if this is true on the UD3P board.
 
From what I found online using "Auto" for the voltages usually ends up with the voltage higher than necessary. I wonder if this is true on the UD3P board.

I think it's true for any board. It over compensates. Always best to manually set your vcore. Reduced heat and power consumption.
 
It's always true, they try to compensate for user inability by putting the next to highest voltage they think needed for any chip to achieve that clock. Always do it yourself.
 
One thing the board seems to have done is gotten the memory voltage dead on for the 1066 timings.
 
I am assuming when you talked about raising the MCH Voltage by .1 you were talking about MCH Core and not MCH Reference or MCH/DRAM Reference, correct?

Correct, you should not need to mess with those.

Would it make sense to boot off of an Ultimate Boot Cd and run the CPU Burnin program so that I don't take a chance of corrupting my windows installation incase an OC is not stable?

Very wise, I use a memtest86+ boot CD myself but same difference.

I didn't realize that the multiplier could be changed through the bios

Yep Yep, but you can only do down from the default 9x but not up. But very handy in finding out your boards FSB limits by reducing the cpu mulitplier to 7 or even 6 so the cpu runs slow and raising FSB to high levels to find out what the board and memory limitations are without the cpu getting in the way. And with memory mulitplier set to 2.0 as well, you kinda isolate the board by itself so you can test what the board/MCH will do for FSB before you start really fine tuning. This is a little process I go thru with my personal boards to get a feel for where the "weak link" in OCing is.


Take notes of changes and results and change only one thing at a time.

add MCH additional cooling by silicone glueing a fan to the MCH heatsink.

lowest cpu mulit, lowest memory multi, starting ast some reasonable FSB raise FSB in steps and find where the board "craps out" on memtest or boot up. Then back off FSB some and raise MCH core voltage a couple of bumps, one "notch" at a time to see how increasing MCH voltage helps, if any, with a higher FSB. So find out where increasing MCH voltage does little or no good and back off "one bump". I never go above 1.4 total volts on the MCH but thats just me.

(here once max FSB at increased MCH voltage is found you can try a bump of FSB reference voltage and see if you can get any more FSB. If so fine, make a note, but put it back to normal cause we are not ready to play in that ballpark yet. )

Then.

For a Q6600 then put mulit at 8 leave all else alone and do same as above and find out where it craps out.
For a Q6600 then put mulit at 9 leave all else alone and do same as above and find out where it craps out.

You now should know about what the FSB limits are for cpu multiplier of 8 and 9 with your CPU and what MCH voltage is needed to get there and if its is likely any FSB reference voltage will help later.

Now look at what memory mulitpliers are available for the FSB values found above. On a gigabyte its easy to figure the memory actual speed, its the FSB x memory muliplier. So for 1066 memory and a 400MHz buss a 2.5 memory mulitplier will have your memory right on spec. For a 333 FSB a 3.0 mulitplier would be slightly underclocking your memory. etc. etc. the math is easy. It is likely you will not be able to make 400MHz - but maybe. Anyway now you have to look at given an cpu muliplier and an aproximate max FSB found from above, and determine what available memory mulitplier will run your memory close to its rated spec. Typically you end up underclocking the memory a tad and thats good to start. Later you can go back to the memory and tweak/play but first getting the cpu running stable as fast as possible is the biggest bang for the Hz. If a memory mulit works out so that your memory is only slightly OCed then by all means try that but avoid tying to get a big memory OC now. Find the cpu limits, get stable as a rock, run the machien a couple of weeks to ensure rock stability in YOUR applications/games, then come back and see what you can get extra out of the memory.

In general you want to pick the cpu mulitplier that allows the higher FSB to have the data running between the cpu and the MCH as fast a possible for highest memory bandwidth but its not super critical unless your applications are heaviliy memory dependent (not likely for most of us). Once the data hits the MCH it will be reclocked by the memory mulitplier to make the trip to memory and back to the MCH at the higher speed determiend by the memory mulitplier. Having a 2.0 memory mulitplier means the data rate on both sides of the MCH is matched and even if it greatly underclocks your memory (you might be able to tighten the timings to compensate) it is sometimes the most stable setting of all as the MCH does not have to buffer or manulipulate the data stream with fancy timing tricks. But thats getting complicated and your 1066 memory would likely need to be way underclocked in that situation. As I mentioned concentrate on getting the cpu to 3.2 with some decent memory settign and ensure stability. Later tackle the memory as it can be much more complex as to options and possiblities.

With the AF7 Pro heatsink around 3.2 should be your goal. 3.3 with good case airflow. Your call - but you really should go with a big TRUE heatsink/fan etc. to shoot for the 3.3+ OC. Use coretemp or Realtemp to monitor you temps and set the option to "show delta T to Tjmax and have that number at least 15 which tells you that you have 15 deg C "headroom" until the processor sends out a "I am getting hot" signal to the motherboard instructing it to put all cooling fans at 100% speed.

Frankly at least with all my office apps and games I find little or "playability" or "snappyness" increase after about 3.2ish GHz and get more by OCing my vid card a bit and fine tuning the memory but thats another post.

Have fun.
 
Thanks Bill. The max OC I was looking at was 3.2Ghz anyway so the AF7Pro should be good enough. If I can get the FSB up to 355 then a 3x multiplier would be almost perfectly in spec for my memory and would give a 3.2Ghz machine.

I burned a UBcd for running stress tests. How long do I need to run them for to "know" that everything is good?
 
People will give all kinds of numbers, I have edited my post a couple of times LOL HELP ME I cant stop yapping.

Bottom line, I get bored playing Prime and Memtest+ after an hour or so - I just lose interest. Run whatever for an hour and check temps. but most of the math based stress tests do not stress the MCH at the same time by running 3D video thru it on the way to the video card. Me personally, in my opinion, the best stress test is to turn the room thermostat up to 80 - 85 deg, take a wizz and grab some snacks and soda and play the living crap out of your most demanding game until you fall out of the chair with exhaustion or just pass out. If the machine has not reboot, reset, froze or otherwise give you a problem you are good to go.
 
People will give all kinds of numbers, I have edited my post a couple of times LOL HELP ME I cant stop yapping.

Bottom line, I get bored playing Prime and Memtest+ after an hour or so - I just lose interest. Run whatever for an hour and check temps. but most of the math based stress tests do not stress the MCH at the same time by running 3D video thru it on the way to the video card. Me personally, in my opinion, the best stress test is to turn the room thermostat up to 80 - 85 deg, take a wizz and grab some snacks and soda and play the living crap out of your most demanding game until you fall out of the chair with exhaustion or just pass out. If the machine has not reboot, reset, froze or otherwise give you a problem you are good to go.

oh man...put your flame suit on. LOL ... nothing bothers me more than flamers preaching that if you don't run P95 for 24 hours, OCCT for 12 hours, 42million passes of IBT, all concurrently while running 3dmark06 and furmark while the moon is full on the second tuesday of the third month there is a chance that a "0" will be a "1" causing a "T" to be a "Y" in text file 30 years from now. :rolleyes:


Personally I'll start P95 before I leave for work. If I come home and it's still priming, mission accomplished. *thumbsup*. If I'm at a windows logon, well, time to tweak some more. Mostly all I care about is if my apps run without crashing. Real world stressing is more important.
 
Ok so it sounds like I can probably run an hour using the UBCD just to get a baseline test (obviously if I get an error in that hour then I need to change something) but after that they try booting windows and give it a go. That makes a heck of a lot of sense (did I forget to mention I'm a software engineer who ends up testing/beta testing way too much stuff?)
 
oh man...put your flame suit on. LOL ... nothing bothers me more than flamers preaching that if you don't run P95 for 24 hours, OCCT for 12 hours, 42million passes of IBT, all concurrently while running 3dmark06 and furmark while the moon is full on the second tuesday of the third month there is a chance that a "0" will be a "1" causing a "T" to be a "Y" in text file 30 years from now. :rolleyes:


Personally I'll start P95 before I leave for work. If I come home and it's still priming, mission accomplished. *thumbsup*. If I'm at a windows logon, well, time to tweak some more. Mostly all I care about is if my apps run without crashing. Real world stressing is more important.

The problem I have is that I have a 9 year old and wife who both use that computer AND its my main recording rig for TV AND my media extender rig. Thus I can't have it down for long running Prime95 (and why Im looking at doing this on Sat since I have nothing scheduled to record until Sunday night).
 
I'm a software engineer who ends up testing/beta testing way too much stuff?)

Ahh a Pro, well not to worry the best test is the real life test as you well know you can run the code on the test enviroment/machine forever, let everyone and his brother test it and 39 seconds after it is release into production some user will find a bug never heard of before.

3.2 is not a streach by any means and if the heatsink OK and case airflow is OK there should be little effort in getting 3.2 and if it should crash the odds are its a sofware issue. Unless you got a complete dud of a CPU all of this is not really going to be an issue. Play a little per my earlier post and then try 9 x 356 = 3.2 with a memory mulitplier of 3,0 with mch core +.2, Vcore at 1.375 if it loads windows after a brief stress test drop Vcore down a notch and retest to find lowest stable, same with MCH and your are done, use the machine and if an issue develops drop the FSB 5MHz and start using ti again.

Me <-- Hardware Engineer who has written enough code to know what a nightmare it can be.
Motto "When in doubt - Blame the software "
 
Me <-- Hardware Engineer who has written enough code to know what a nightmare it can be.
Motto "When in doubt - Blame the software "

A good QA person is one who can break the code no matter what the engineers do. Anything over a few hundred lines of code is bound to have a bug somewhere (whether its in the code, compiler, etc).

The flipside of me. "When in doubt blame the hardware" is my motto :)

Im currently working on a terminal that prints tickets that are supposed to be readable by another terminal but the customer is complaining that the other terminal is having problems reading my tickets. Now I have neither the other terminal NOR has the customer bothered to actually send me a sample of what the tickets look like that the other terminal prints (when he did I found I was actually really close).
 
I have a 120mm intake fan at the front of the case and a 120mm fan output fan at the back of the case which the fan on the AC7Pro is blowing into. My cables are mostly wire tied out of the way but I will finish that job once my replacement 260 comes in from evga.
 
ASUS P5Q3 Deluxe WiFi-Ap@n
Intel Quad Q6600 G0
ASUS Silent Knight AL CPU Cooler
Kingston Hyper X DDR3 1375MHz 2x1GB
OCZ Memory Cooler
XFX Black Edition GTX 260 666MHZ[Stock]
2x Vel. Raptor 300GB

Here this might be able to help.

AI Overclocking: [Manual]
CPU Ratio Control [Manual]
CPU Voltage: [1.25v]
FSB Frequency: [333]
PCIE Frequency always: [100]
DRAM Frequency: DDR3 1375MHZ
DRAM Command Rate: [2T]
DRAM Timing Control: [Manual]
Memory Timing: 7-7-7-20
DRAM Voltage: [1.7v]
FSB Termination Voltage: [1.20v]
North Bridge Voltage: [1.25]
South Bridge: [1.05v]
C1E Support: [Disable]
Mutiplier: [9.0]

BAM 3.0GHz

Just switch a little things from 3.0GHZ

CPU Voltage: [1.31v]
FSB Frequency: [389]
DRAM Frequency: DDR3 1333MHZ
FSB Termination Voltage: [1.40v]
North Bridge Voltage [1.5v]

NOTE: This is on my AUS P5Q3 Deluxe WiFi-Ap@n. Still for sell, MOBO, CPU, MEM $350 for combo.
 
Thanks for the values. I was running a stress test on my non-OC'd system last night with Prime95 and Linx and was watching cpu-z so that I could see what the core voltage was and it wavered between 1.20 and 1.216. Is this due to VDroop? Isn't the default voltage 1.30?
 
ASUS P5Q3 Deluxe WiFi-Ap@n
Intel Quad Q6600 G0
ASUS Silent Knight AL CPU Cooler
Kingston Hyper X DDR3 1375MHz 2x1GB
OCZ Memory Cooler
XFX Black Edition GTX 260 666MHZ[Stock]
2x Vel. Raptor 300GB

Here this might be able to help.

AI Overclocking: [Manual]
CPU Ratio Control [Manual]
CPU Voltage: [1.25v]
FSB Frequency: [333]
PCIE Frequency always: [100]
DRAM Frequency: DDR3 1375MHZ
DRAM Command Rate: [2T]
DRAM Timing Control: [Manual]
Memory Timing: 7-7-7-20
DRAM Voltage: [1.7v]
FSB Termination Voltage: [1.20v]
North Bridge Voltage: [1.25]
South Bridge: [1.05v]
C1E Support: [Disable]
Mutiplier: [9.0]

BAM 3.0GHz

Just switch a little things from 3.0GHZ

CPU Voltage: [1.31v]
FSB Frequency: [389]
DRAM Frequency: DDR3 1333MHZ
FSB Termination Voltage: [1.40v]
North Bridge Voltage [1.5v]

NOTE: This is on my AUS P5Q3 Deluxe WiFi-Ap@n. Still for sell, MOBO, CPU, MEM $350 for combo.

Just switch a little things from 3.0GHZ

CPU Voltage: [1.31v]
FSB Frequency: [389]
DRAM Frequency: DDR3 1333MHZ
FSB Termination Voltage: [1.40v]
North Bridge Voltage [1.5v]

3.5Ghz. i went to

-skyyless
 
I am really hoping to have time on Sunday to see whether I can hit 356Mhz FSB pushing my system to 3.2Ghz and keep it stable and cool enough. I didn't like it that I had 2 cores occasionally hitting 62C while running Prime95-64 and LINX.
 
If you don't mind, I'd like to post some of my results here too, I seem to be in the same boat as you, as I have the same, CPU, mobo, and cooling fan looking to run a stable OC around 3 or 3.2 ghz. My ram is Dominator at 5 5 5 15 1033mhz.

I actually followed this thread yesterday and was able to POST instantly. Right now I'm running 3.2ghz@400FSB with 8x multiplier. I was able to get my vcore down to 1.18 but the system instantly crashed when I ran Prime95...If only ha!

I'm trying to get mine stable now. Here's my settings:

CPU: 8x
FSB: 400
PCIE Freq: 100
Adv Clock: [Defaults]
MCH Latch: 400
Mem Multiplier: 2.66D
Mem Freq: 1066
Timings: [Defaults, unadjusted, 5 5 5 15]

Vcore: 1.24375
MCH Core: 1.050
Dram V: 2.1

All other voltages are set to 'normal' right now. Gonna play with these.

I ran Prime95 with this for 11 hours and core 1 got an error after 3+ hours. The other 3 were still running when I came home. Just tried to play some L4D and the system locked up within 5 minutes, so obviously I'm not stable.

I'm going to put the vcore up one and MCH Core back on 1.1v

Hopefully you can benefit from some of my info. Do what you want, but I recommend using the 8x multi and 400 FSB to get to 3.2ghz...it keeps temps WAY down.

BTW, with these settings, my temps are 23-27 C idle and 44-47 with Prime95. Not bad! This is a GOOD HSF....
 
Have you tried running both Prime95 (64bit version) and LINX in 64 bit mode at the same time? There is another thread about running both and it being a killer. Im probably going to try 356Mhz with a 9x multiplier to start.
 
Do any of your guys enable "Load Line Calibration"? How does it affect overclocking?
 
I just did a quick overclock on this box and then disabled it. I had 355FSB with 9x multiplier and 3x on the memory with CPU and MCH core voltage's set to Normal. I ran over two hours worth of various tests (including 30min of memtest86) without any errors BUT on one of the reboots between tests my bios reset my settings and then complained about a problem. Is it normal for something like that to happen? Im guessing I should probably up the cpu voltage slightly.
 
I would say to up it one....IMO Prime95 doesn't mean shit unless you run it for HOURS...I ran mine for 11 hours and it took over 3 hours before 1 core failed. If you run it for 15-45 minutes, that won't tell you shit, unless your settings are really bad and your system crashes too easily.

I'm tired of dicking with settings and being expected to run it over night...I don't have time for that. I like to game on the weekends and so far I haven't gotten SHIT played or SHIT done...except for a hair cut today by my hot barber..but that's another story.

Anyways, I've been testing my system with just live game play...if Crysis crashes in 5 min...something's wrong! It would take hours for Prime95 to tell you that. Same thing with COD and L4D.

I've also been running 3D Mark 06 and Vantage, which simulates a game without actually having to go in. Plus these tests have CPU tests built in, not just GPU, which uses some CPU anyways. If the test locks up, I reboot and adjust something.

Right now all my settings are oh normal except for Vcore, which is 1.3 (will be lowered once stable) and CPU Term which is also 1.3. so far, it seems the tests either don't crash or don't crash as early as I've been increasing the CPU Term. Once I run my tests and things seem stable, I will dick with my MCH and ICH.

BTW, I saw a 20 FPS boost in Crysis...hot damn!
 
One of the things I was trying to do was validate the memory at the FSB speed. Memory usually fails after the first or second test if it is going to have a problem. (At least at stock speed it always has for me).
 
Maybe I should run that test then. I've been able to pass both 3D Mark 06 and Vantage and run a little P95 with success. But after 10-15 minutes of actual game play the game locks up and I have to hard reset. The mem voltage should be fine but I'll up it to 2.2 just to see. I'm not sure which voltage is causing these lock ups.
 
I have found that it is good to run it for awhile with a new set of memory but I figured it would work the same with a new FSB speed. One thing I did notice was that UBCD needs a PS2 keyboard for its prompts, glad I keep one in the basement.

I will play with it more tomorrow night. Probably up my cpu core from 1.3 to 1.31 volts and boot windows (I just backed my main drive off to a USB drive, just incase.
 
There was one time were whatever voltage I adjusted caused a BSOD that would say 'memory bad or corrupt'. I reset the settings and it still got the error. I had to put in the Vista disc and repair the OS for it to work again.

I had my Vcore on 1.3 and CPU Term up to 1.34, starting at 1.1, and I still couldn't get my system stable. I started adjusting MCH and ICH and that made it worse. This is rediculous. I upped the ram voltage and that didn't help. This is getting silly.
 
Try dropping the FSB speed to 355x9 with a 3x multiplier on the memory. That works out to 3.2Ghz (same as you have) and 1065 on the memory. The lower FSB should make things more stable (I would think). Reset the voltages to "normal" and start again.
 
Back
Top