Overclocking Q6600 on GA-P35-DS3P

[xx04201987xx]

I'm trying to OC for the first time and I need some help from experts.
I have Q6600 (B3)
and MOBO Gigabyte P-35-DS3P
I'm seeing all these voltage control and don't really know how much to up the voltage for each part.

I see things like...
DDR2 OverVoltage Control
PCI-E OverVoltage Control
FSB OverVoltage Control
(G)MCH OverVoltage Control
CPU Voltage Control

I'm trying to get to 3.0 GHz.
It would be really nice if someone can help me with these voltage control.
About how much I need to increase them

for cooling I have Zalman 9700.
After putting the fan in I get ideal temp of about 42C @ stock speed.
Do I need to place my cooler better before I can OC?

I left voltage control to auto and tried running 3GHz
ran Prime 95 and can't go over 5 mins without temp going over 60C
was going up to 63C before I turned the test off.
Ideal temp at 3GHz around 43-45C

 

[SpunkyJones]

Check out the thread by Greysky pinned to the top of this forum. I used it a couple of weeks ago, and have my q6600 running stable at 3 Ghz.

 

[Idontcare]

With a B3 on air you should be happy with mid-60's for temps at 3GHz from what I have gathered.

I run my B3 on vaporphase cooling to improve my overclocks. My G0's run on air with a Tuniq120 and loaded temps are in the low 60's for a 3.3GHz clock.

I don't know how your HSF compares to a Tuniq120, but unless yours is much better than a Tuniq120 then I think you can be happy with your temps given that you have a B3 quad.

 

[xx04201987xx]

in the sticky thread it only talks about DDR voltage and CPU voltage.
well at lease I don't see other stuff...
and CPU-Z doesn't show me other voltages so I don't know where to start for other voltages.

 

[xx04201987xx]

bump

 

[VirtualLarry]

To get to 3.0Ghz, the only voltage adjustment that you should need is the CPU voltage (Vcore). Start by increasing the FSB by 5 each time, and run Prime95. If it errors out, increase the Vcore one notch and re-test. Repeat until you hit your desired overclock, or your vcore or temps exceed a safe amount. To be extra safe, once you reach that point, lower your OC back down by 5.

Keep your load temps (while running Prime95 Small FFTs - use CoreTemp to measure) under 70C.

 

[BonzaiDuck]

I can't be sure that an average VCORE for Q6600 @ 3.0 Ghz isn't dependent in some way on the chipset, but from what I've seen so far, you should be able to get there with VCORE around 1.32V with the B3 OR the G0 stepping.

My oc'ing adventures traced the following path with the B3:

1.32V and 3.0 Ghz
1.375V and 3.13 Ghz
1.418V and 3.15 Ghz
1.418V and 3.2 Ghz

Another source of information would be the thread that was "hot" late last summer on the Q6600 over-clocking -- asking people to post their voltages, maximum over-clocks and temperatures, with emphasis to specify the stepping as B3 or G0.

Of course, even for a particular stepping, the results are going to vary as the VID varies across individual processors.

I had seen some reviews that put the B3 through its paces -- showing that 3.4 Ghz was attainable at 1.46+V. One review said that 1.46+ was "safe" for the B3. I just never wanted to venture outside the 1.42V envelope, so that's where I stopped.

I won't beat up on you about the Zalman 9700 -- you could do worse. You could also do better. But if I had a piece of Acrylic or Lexan 5"-diameter tubing, I'd be tempted to see how much I could increase the efficiency of that Zalman cooler.

A lot of people here I'd seen with nice cases and the same ThermalRight cooler I'm using, and we compared room-ambients and peak core temperature under load. At 79F room-ambient, they were butting up against 70C running at 3.0 Ghz, and I was at 65C @ 3.2 Ghz. People over at "Cases and Cooling" will tell you that I worked like crazy all summer to trim my load temperatures.

Right now, I wish I had some extra ducats to spend on a later motherboard and a high-end Penryn. And with the good temps, if I were tempted to push the voltage above 1.44 or 1.45 on this B3-Q6600, and the lightbulb burned out, I'd also be in a pickle.

 

[RussianSensation]

This article has a lot of info

Here is the basic breakdown:

DDR2 - stock voltage at 1.8v; depending on your memory you can take it up to 2.2V+ but I wouldn't recommend above 2.0V at DDR2 800 as there is not much need to go beyond that at stock.

PCIE - fix this to 100

FSB - If the system doesn't pass at 6x400 (described below), increase by 0.1V (might help). Generally there is no need to increase FSB voltage on your board to get 400FSB.

MCH (northbridge) - This might require an increase of 0.1V (stock is 1.3V i believe)

CPU voltage - on Q6600 try not to go above 1.45V. Since B3 runs pretty hot, it's best to increase your cpu speed on stock voltages until you are required to bump them up. Just make sure voltage is not set to Auto Control or as you overclock the system will increase voltages automatically. Set it to Manual.

------------------------------------------------------------
Overclocking

Basically the idea is to first isolate the components 1 by 1 from the cpu to make sure it's neither the motherboard nor the ram that's holding the system. Luckily with B3, it's not going to be the memory (since it's hardly going to clock beyond 9x400 = 3.6ghz).

So to test the motherboard simply lower the multiplier from 9x to 6x in the BIOS while bumping the FSB to 400. This will give you cpu speed of 6x400 = 2400 (stock speed cpu), while still operating at stock DDR2 800 memory speeds (set the divider to 2.0 as on Gigabytes that's your CPU:FSB ratio of 1:1). Make sure you relax the ram timings to 5-5-5-15 to isolate ram as the culprit (you can tighten them later).

Start Prime 95. Here you are testing the stability of the motherboard (northbridge) at 400FSB. Use Prime 95 small FTTs to run it for 24 hours. If the system passes then you know that you can increase the FSB on your motherboard to 400FSB without changing anything else. If the system doesn't pass then:

1) Bump the MCH voltage by +0.1V

If it still doesnt pass

2) Bump the FSB by +0.1V

Once you know that your board passes 400FSB, you are almost done.

Now set the multiplier back to 9 keeping everything the same (for voltages), but now lower the FSB to stock 266. You are now searching for your top cpu speed. Increase the FSB to 300 to get 2.7, see if it boots safely. Keep going until you find instability of sorts or the system doesnt boot. At this point you add voltages slowly. At some point you'll find that the temperatures are either too high at load or the system doesnt boot even at 1.45V. From here you can start decreasing your cpu speed and testing at which voltages it is stable with Prime 95.

I also want to point out, almost no program out there will max out the quad as much as Prime 95 will. I personally load all 4 cores with BOINC since that's the most intensive program I run on the quad and there is no need for me to get Prime 95 stability since my system will NEVER reach those loads. However, it is the most bullet-proof method but not necessary the most realistic for the real world for a quad (different story for dual cores though).

 

[VirtualLarry]

Originally posted by: RussianSensation
I also want to point out, almost no program out there will max out the quad as much as Prime 95 will. I personally load all 4 cores with BOINC since that's the most intensive program I run on the quad and there is no need for me to get Prime 95 stability since my system will NEVER reach those loads. However, it is the most bullet-proof method but not necessary the most realistic for the real world for a quad (different story for dual cores though).

I'm disturbed by the fact that you participate in distributed computing, and yet don't think that Prime95 stability is important. How do you know that the results you send back are correct? You can't know that without stability testing!

I run Seventeenorbust, and it easily gets to the same level of loads as Prime95. Only it doesn't have the internal consistency checks that Prime95 does. So without testing Prime, I could be returning invalid results, and destroy the integrity of the project.