Things not to do (if you think you are an overclocker)!

Status
Not open for further replies.
Nice tips. They do come in handy. All I would like to specifically mention here is that in my opinion, insulation is the most crucial thing in any LN2/DICE session. From whatever little experience I have with LN2/DICE, I have always observed that whenever my insulation was good and the pot contact perfect, those sessions went well.
Hence take as much time as you want for insulation, never skimp here. Good insulation is not only necessary to protect your hardware but also to ensure that you have a trouble free session, where you just need to worry about clocking and dodging cold bug.

Just my 2 cents. :)
 
Amey said:
Nice tips. They do come in handy. All I would like to specifically mention here is that in my opinion, insulation is the most crucial thing in any LN2/DICE session. From whatever little experience I have with LN2/DICE, I have always observed that whenever my insulation was good and the pot contact perfect, those sessions went well.
Hence take as much time as you want for insulation, never skimp here. Good insulation is not only necessary to protect your hardware but also to ensure that you have a trouble free session, where you just need to worry about clocking and dodging cold bug.

Just my 2 cents. :)

Finally you comment and add to it, Darky! :)
 
Sudarshan_SMD said:
Excellent write up. Thank You.
I have read various guides and they mention that, increase clocks in steps(say by 5mhz), after system crashes/does not boot lower clocks by 10mhz. But, I don't know why it is to be increased in step by step.

What happens if we increase clock by 50mhz?

What if we increase clock by 50mhz(assuming we take care of Vcore) and once system is unstable, reduce clock by 100mhz and then increase clocks by 5mhz? What difference could it make?
pardon my noobness.
Good point... but this again you feel. Its instinct man.

I usualy give it a go on air to check it's potential with how much volts for what freq..

But then I sometimes dont even test the chip on air before going LN2. Recently 1090T and 980x both were done that way, direct under LN2 at my end. Both got easy 6GHz+ in less than an hour from starting with GIGABYTE boards which I prefer over other (also the fact that they sending me their good stuff).
 
Sudarshan_SMD said:
Excellent write up. Thank You.

I have read various guides and they mention that, increase clocks in steps(say by 5mhz), after system crashes/does not boot lower clocks by 10mhz. But, I don't know why it is to be increased in step by step.

What happens if we increase clock by 50mhz?

What if we increase clock by 50mhz(assuming we take care of Vcore) and once system is unstable, reduce clock by 100mhz and then increase clocks by 5mhz? What difference could it make?

pardon my noobness.
Amount of frequency increase and respective vCore depends on the user. Many like to up the vCore to the maximum permitted on official sites (say 1.34v for my Q9550) assuming this will run any frequency till ~3.8Ghz. Then you can increase the speed in 100Mhz steps. You just bounce the OC really fast, but are overtly stressing the chip. Then once they hit the speed they want, you can decrement the vCore in 0.05v steps. Usually I prefer to find the lowest possible vCore my system will remain stable at (IBT/Linx/OCCT Linpack) on stock, then starting from base stock peddle up the speed in 5-15Mhz increments. This way I will never kill the chip. Once I reach a unstable point I go back into the BIOS and notch up the vCore 0.05V again. And keep looping. Its gets difficult once you cross ~3.8 Ghz territory. The FSB has taken a massive bump and the NB (for LGA 775) starts to get stressed. This is when the real fun starts. True OC actually begins here. Cause now you have to play with multiple parameters.The CPU GLT, vMCH, MCH GTL, VTT FSB, vICH all need to be fine tuned. The MCH specially needs to be strengthened perfect. For example if my MCH is 1.496 I get stability. If I move it say to 1.50V the system will triple post and reset the CMOS. Now if I am at 1.1840 for the MCH again I am stable but a degree below this, causes the system to reach the desktop but most drivers do not load. Weird stuff starts to happen. If my vICH is above 1.50v my CMOS resets..! See, how critical the voltages become. Usually at this point I take out a pen+paper and start to note settings. Of course once I hit any thing stable greater than 3800Mhz I will commit it to one of the BIOS storage banks. It is a good start up point if things fail.

As Harshal, mentioned it is instinct and the feel good factor you have with your equipment. Just pushing in a massive vCore is an easy way to hit the desired high speeds, but I feel, not the correct way. You may lessen the lifespan or be running a system which is receiving far greater power than it should be...! You had mentioned "take care of vCore". Not that easy. It is extremely dynamic and fluid, and should not just be "increased" on the whim to get more speed. Careful graduations are must. Just my thought and style. Obviously open to debate and critique.
 
This thread is more of a check list of things not to overlook when you are going extreme cooling way, guys.
Technicals and other few things I dont think I can just write n explain to you all. Comes with experience. So keep at it, play around with your hardware n learn. Just remember I am not responsible for your play time mishaps! :D
 
I think I've like read this too many times , every time I read it . makes more sense :) Great write up sir :)

You must spread some Reputation around before giving it to Harshal again. :|

You cannot give Reputation to the same post twice. :|
 
l33t said:
I think I've like read this too many times , every time I read it . makes more sense :) Great write up sir :)

You must spread some Reputation around before giving it to Harshal again. :|
You cannot give Reputation to the same post twice. :|
Thank you for being kind with words!
I think I have a new point to add to it will do so after finalising the write up! :)
 
Status
Not open for further replies.