GA-P35-DS4 temp questions, overclocking

[Dewdman42]

Hi all. I am a new gigabyte mobo owner and also pretty new to overclocking. I have been playing around with mild overclocking of my Q6600(G0) together with the GA-P35-DS4. Very easy to do and this board has so many options that it is very flexible. Therein lies the rub though, its hard to know what is the BEST way to go because of so much flexibility. A couple questions: (my system specs below)

1 - What is an acceptable Northbridge temperature and how can I monitor it? I am touching the NB heatsink with my finger and its just barely warm, even when I'm overclocking pretty high. These coolpipe things work great!
My Ram isn't too hot either by touch. Stock CPU cooling so far and no NB fan yet... I don't plan to overclock to the extreme anyway, but I still might get a better cooler to lower fan noise.

2 - Trying to understand if there is any benefit to ensuring 1:1 mem clock divider versus a different ratio? With my system, in order to get into overclocking, I have found a bunch of different combinations that will work between the cpu clock multiplier, the CPU FSB speed and the clock divider(which in my case is either 1:1 or else actually multiplying rather than dividing (ie 2.40 or 2.50).

Should I try to get the most memory bandwidth I can get, even if the memory clock speed is higher than the fsb? Is there any advantage to using 2.00 instead of 2.40 or 2.50 mem clock divider or should I just try to shoot for the max mem bandwidth I can get, regardless of the CPU? I know about heat issues, for now, just wondering about performance considerations assuming the heat will be taken care of.

My ram is overclockable to about 900mhz(I tested it)

GA-P35-DS4
Kengsington HyperX DDR2-800 5-5-5-15
Intel Q6600(G0 stepping)

I will use this machine for audio production, which means that lots of data will be moving around. I think a higher memory bandwidth could be advantageous, but I am wondering if it makes sense to use a strategy where I use a lower cpu clock multiplier with a higher FSB speed and 1:1 divider...to acheive it....or.... should I use the maximum 9x clock multiplier, with a somewhat lower FSB and then a 2.40 or 2.50 mem divider to ramp the mem speed back up to what its capable of?

[Lsdmeasap]

Re: GA-P35-DS4 temp questions, overclocking

NB temps can be seen in Everest, or EasyTune. It is called "System" In EasyTune, and "Motherboard" in Everest Sensor section. And since you did the finger test, sounds very cool to me so no worries there

1:1 ratio is always best I have found. But everyone has thier own opinions. You can test the bandwidth with everest also. If anything you will get a few (FEW) extra points in bandwidth testing seen only in benchmarks by using a larger divider. 1:1 is the best, so you can either use you max fsb and max multi with 1:1 or you can lower your multi and thereby increase your FSB and ram, which ever you find faster. Testing is the only way to know for sure what works best for your hardware

[Dewdman42]

Re: GA-P35-DS4 temp questions, overclocking

Using HWMonitor, I see a couple temperature readings, TMPIN0 and TMPIN1. Is one of them reliably the NB? I had heard that this sensor was in a different area of the mobo.

You are not the first person to tell me that 1:1 is ideal, but I'm still trying to find out why 1:1 is considered ideal? Why do you like 1:1 the best?

I ran a series of benchmarks with Sandra. I did about 4 or 5 scenarios with the Q6600 modestly clocked to 3ghz, using different clock multipliers, ram speeds, etc.

A lot of operations performed the same, but I noticed a difference in performance with ram bandwidth as well as multi-core communication. The best scenario was with clock multiplier at 8, CPU clock speed at 375 and divider at 2.4. So this was not 1:1 and I could not detect anything that was worse because of not having 1:1. That is why I'm trying to find out why everyone keeps saying 1:1 is more "ideal".

In my case the higher bandwidth I think will be beneficial because of the kind of work I do, but I want to make sure I understand the tradeoffs. however, so far I can't see any tradeoff at all.

[Lsdmeasap]

Re: GA-P35-DS4 temp questions, overclocking

1:1 Is better because at any other ratio you have to look at it like this....At 1:1 your CPU thinks about a cheesburger at the same time as your Ram does. At 3:2 your CPU thinks about 2 chesseburgers, but before it can go to eat them it must WAIT on the ram because the ram is thinking about 3 cheesburgers and thus is thinking longer and slowing down your CPU's thoughts

As for HWmonitor, I have never used it to tell you, Do you have easytune installed? (I dont use it, nor suggest, but You can see the System temp, which I reliably think is directly by the NB) But since you did the finger test, you are fine.

One thing I personally do when I get any board is Redo the heatpipes or NB/SB/Mosfets Thermal paste. Sometimes there is way too much, sometimes not enough. All up to you, if you do do this though, Use NON conductive paste. I redid my whole DS4 pipes with TX-2, my temps are great

As for your tests, it also really depends on your actual rams performance and hardware as well. Sounds like your ram may perform better at lower then rated speeds, and this could have possibly only looked or been this way because of the lower timings allowed. Unless I am not sure what 2:4 made you ram though either

So again it is really up to you, your hardware, and testing. But, you would Idealy want your ram to be running at least at the rated speed. But yeah testing is the only way to know what performs better. As for real world difference you would likely not see much at any setting, other then your highest FSB

[Dewdman42]

Re: GA-P35-DS4 temp questions, overclocking

I guess I must be misunderstanding something then, because I thought that using a 2.40 ram clock divider with my DDR2 is actually making my ram run at a FASTER speed then the CPU. It should be the ram that is standing around waiting for the CPU, not the other way around.

For example, I'm currently running my machine with cpu clock multiplier at 8, clock speed at 375, ram clock divider at 2.4.

CPUZ reports 3.0ghz cpu, 1500mhz FSB and 450mhz ram speed (but its actually double that at 900mhz since its DDR2).

If I use the default of 2.00 for the ram clock multiplier, which is supposed to be the same as 1:1, then my ram clock speed is only 375mhz with 750mhz DDR2

What am I missing??

[Dewdman42]

Re: GA-P35-DS4 temp questions, overclocking

One thing I personally do when I get any board is Redo the heatpipes or NB/SB/Mosfets Thermal paste. Sometimes there is way too much, sometimes not enough. All up to you, if you do do this though, Use NON conductive paste. I redid my whole DS4 pipes with TX-2, my temps are great

Since the finger test seems to be ok, even when cranking up the FSB pretty high, I'm going to just leave it alone for right now, but I will remember that for the future. In the future I will probably install an aftermarket CPU cooler and I can re-paste the NB heatsinks at that time. Is TX-2 the paste everyone likes the best?

[Lsdmeasap]

Re: GA-P35-DS4 temp questions, overclocking

YOu got it right, either way one is waiting on the other...Thus slowing down the system.

You can use a lower CPU multi if you want to keep 1:1 and raise the FSB, which will then raise your ram freq.


TX-2 Or MX-2 Are both good, Arctic Ceramic also is good, but I dont think it is the better of the 3. I just suggest NON conductive, so that is why I said TX-2

[Dewdman42]

Re: GA-P35-DS4 temp questions, overclocking

I'm not sure how I understand that RAM being faster will cause the whole system to slow down, but I will research it more. Thanks for the tip about the grease.