To start off, when I first got my system, OC'ing seemed easy till 3ghz range, dual or quad core wise.  When I set the FSB to 333 or 1333mhz (for my settings), the system simply would not boot causing me to reset my bios.  There were times when I could OC to 3.1ghz and then the bios screen would freeze.

When I first starting thinking about it, I thought I was hitting some FSB walls because of my chipset (NV 650i).  Well it turns out I was wrong, but on the right track.  It was the chipset, but not like I was thinking till I found out another cause that was keeping my OC's to a min.

So, I was thinking about the amount of heat going on with the MB's HS over the voltage regulators, down to the North and South bridge:

Since I had a IR Thermometer, I decided to really look at the temp readings I got from it.  Keep in mind I couldn't get a steady temp, so I ranged it.  On idle, for my 2.880 OC, temps were:

1.) Voltage regulator mosfet - 48/50C
2.) NB - 48/50C
3.) SB 41/46C
4.) RAM 41-46C

With it under a load, using prime 15mins +:

1.) Voltage regulator mosfet - 63-71C
2.) NB - 62/63C
3.) SB - 62/63C
4.) RAM 46-49C

That really got me thinking.  As you can see on the load temps, the heat was pretty much affecting the SB allot, and I know its not going to get anywhere near the temp, and really doesn't need a HS, imo.

So my plan was to break that chain HS on my MB.  I brought some parts:

RAM - even though it wasn't part of the HS on the MB, I thought why not.  

Thermaltake CL-R0026 Aluminum Heatsink

This product was easy to install, but I had heat spreaders on my ram that had adhesive to keep them on.  I spent a good hour cleaning off all the sticky crap, mubbling... why did they do that?    After I cleaned them up, I was kinda worried if I damaged anything, especially when I ripped off the heat spreaders:

Turns out my RAM was okay on Memtest, which I ran 2 hours, and without the heat spreaders, was doing about the same load temps when Memtest was running.

This is what the MB looks without it's aluminum, copper spray painted HS:

It was kinda tough getting the HS off the South bridge.  Turns out MSI used an adhesive on that chip.  And yes I was mumbling again.. why.. why... why???

So I bought the COOLER MASTER RR-NCW-L4E1-GP Copper (Skived fin) Fan&Heatsink for the South Bridge, even though I didn't think it really needed it.  I had to modify the HS to fit on the MB though, it went as far as breaking a fin off to drilling holes in it to mount it.  I also had worries when I had it on, if the video card would go on correctly.  To my relief, it did.  

Next is the NB.  Which I brought the Thermaltake CL-C0034 Copper Fan&Heatsink.

I wanted to get a different one, that had the heat pipe exposed, but had an idea that I'd have problems centering it on the NB chip which is really small, and I use a huge tower HS for the CPU which doesn't make things fit like I want.  So this NB cooler works okay with my MB which swivels to fit.

Last but not least, the voltage regulators.

Thermalright HR-09U Mosfet Cooler

I wasn't really sure about getting that, but I did.  It basically covers all the voltage regulators, but again I had to modify the mounting since its system didn't fit my setup.  I basically used some of the mounting hardware from the cool master NB/SB HS.

So my rig looks like:

After all the work, and money spent into it (90 bucks)... The system never booted again.  

Okay.. I had to plug the power back in the PSU.  

And it basically worked.  I was able to push the system to 3.2, but it wasn't stable.  I haven't had the time to tweak everything, but I'm still happy with the 2.880 OC atm.  The temps are now no longer linked together.  On load, with my original OC:

1.) Mosfet - 50-52C
2.) NB - 47-49C
3.) SB - 36C
4.) RAM - 35C

I'm not really sure about the RAM HS I got for the temp reading since they are really big, and its hard to say if its really dissipating the heat, or hiding the heat signature from my IR.

Here is the CPU-Z on the 3.2 OC that wasn't stable:

Q6600

Heh, just in case, I using speed step.  My multiplier was resting at x6, on load it would be at x9.  On load temps were still tolerable, but my ambient temps were 83/84F which I didn't like for the system to run at.  

I did get a BSOD during the short testing.  Guess I have some more OC'ing to test out to see if I can get this thing to 3.6ghz. My main conclusion to why some NV 650i chipset may not do so well, is that its basically over heating with their current cooling on the NB chip.

You couldve bought a cheap p35 mobo for slightly more than 90 bucks. The nvidia chips arnt good for ocing, especially quads

True... but I didn't want to go through the hassle of reinstalling Vista 64, and dealing with MS for a different MB (reactivation), not to mention being stuck with a MB with no use.

Edit:

Also, the money spent was on the RAM, 14 bucks x 2 = $28.  So it was actually around $60 bucks for the separate cooling.  I perhaps should have looked at cheaper prices for the NB/SB chips, but what is done is done.

It could just be your chip doesn't OC well...

Heh... well how could it?
 
The voltage regulators get pretty freaking hot, and its linked to the north and south bridges.  That was the reason I made this thread, to test it out.
 
I wasn't looking for thoughts on how bad the NV 650i is, but to help it, without having to buy a new chipset/MB.

Let me re-clarify myself here.
 
When I was using the stock HS for the chipset (copper colored), I couldn't even get to the desktop at 3.1ghz to even log into Vista 64.  I couldn't even run safe mode to check anything out.
 
Now with separate HS on the chipset, I was able to get into my desktop, to even send CPU-Z a validation, and test some with Prime 95 at 3.2ghz.
 
So I hope that help everyone understand what I'm trying to accomplish.

Good job and nice review!
I knew the linked heatpipe systems would share the temps but I never knew by just how much!  With everything this cool, you should have no problems dumping enough voltage into your Nb to get it stable.

@ Blackwidow_rsa
 
Yes he could have, but it is much more impressive that he did not.  Modding an underperforming piece of hardware is much more rewarding than dumping unnecessary cash at it.  Besides, some of us enjoy the challenge    

Thanks!!!  
 
I'm always in the state of mind to try at least make older stuff run better, even though it's not even 3 years old yet... well I haven't had it for 3 years.  
 
I'll try to keep this updated as my progress goes on.  Right now I'm on my E4400 setup, my linux PC, while my Q6600 is testing.  Right now it's at 3.3ghz under memtest.  Just want to be sure the RAM is okay to some extent.
 
My BSOD on the first try was the IRQ_EQUAL LESS crap.
 
Not sure, but I may not be able to go to 1600 rated FSB, but I didn't cranck the juice to the limits I'm willing to go... yet.  

Nvidia chipsets = Fail with OCing.
 
Good luck though Grimmy. Maybe someday in the future you can grab a cheap P35/P45 to replace it with and get nicer more stable OCs.

Heh... Naaaaaaaa.  I guess I'd have to get a new MB if I blow this one up.  
 
I'm actually happy for what it can do now.  Just wanted to try to expand it's OC potential.  If I'm getting a new MB, it will be with perhaps the latest chip..  Which maybe in the next few years.
 
But ya, I know this 650i is a pain in the butt, though I wanted to put my thoughts to the test.
 
Tanks on the luck.. I'll need it.
 
. o (do you smell something burning?)
 
. o O (gotta go.)

Update on my progress.
 
I can barely make it to 3.4ghz with the 650i chipset.  The actual speed is 3375mhz.  Main problem is blue screens.  So my max rated FSB is 1500, but unstable.  Pretty much cranked the juice up, and things did get toasty.  So my dreams are shattered for 1600 FSB.  It just won't post.
 
Right now my room temp is 83F, and the ambient case temp is... 91F.
 
However I did manage to get a more stable OC to 3.2ghz, with the FSB @ 1475, speed step disabled.  Prime ran about 20 mins, and temps were:
 
Tcase 50
Cores - 54 - 54 - 52 -52
 
After 20 mins, the room just got un-comfy just to sit in.  
 
Also I noticed that my vdroop is prolly another reason why my OC won't stick on the 3.375 OC.
 
Right now the OC, when the system is idle - 1.368 - 1.376 vcore
When the system is under load, it drops - 1.320 - 1.344
 
The temps on the new parts I got:
 
1.) Mosfet - 52-55C
2.) NB - 49C
3.) SB - 38C
4.) RAM 38C
 
Also decided just to run the RAM in single channel, rather then dual.  There really doesn't seem to be any difference running in dual channel.
 
At least I got a lil more out of my system, but I think I'll go back to 2.880 for now.  I do like running it cooler with speed step on.

Hmmm.... If its BSoD have you tried upping the memory voltage or even the timings?
 
When I was OCing my Q6600 to 3GHz, my mobo set my memory voltage to low and I had to set it to spec in order for it to run without a BSoD while Prime95 testing.

Yes...
 
I increased the memory voltage to 2.40 volts, and relaxed the timings to 5/5/5.  
 
The FSB VTT was increased to 16%, and my vcore was set to .0750+.
 
Seemed like memory problem as well.  I was able to play a game and such.
 
Oh.. heh at 3.3ghz, my system did super pi at 15 secs.  . o O (o well)

Just checkin. I have yet to try for over 3GHz. Although I know my HSF will be good for 3.2GHz and maybe 3.4GHz (Zalman 9700) but I am too lazy to try and set the voltages and Prime 95 and such especially when its getting to about 110+ during the day and my car gets 125+ if I don't roll the windows down lol.
 
Although it would be a good compariaon between the P35 and 650i chipsets that people can look at and see if they really want SLI or the ability to PC much easier....... I shall think about it.

No problemo... I'm open to any suggestions even if I tried them, but didn't mention them.
 
It would be cooler in my room if the AC thermostat went by the temp in my room rather then the living room.  
 
It would be a good comparison, but the P35 should handle it better, I think.  I prolly should have waited, but couldn't pass up the $80 MB at the time.  This experience also pretty much taught me to think more about the MB cooling as well.  Something I've seem to over look with this MSI P6N Plat.
 
Although the P5k MB you have has somewhat the same cooler for the chipset.  It would be interesting, but don't sweat it.  110F temp, I'd wouldn't want anyone to have ta do anything stressful.  Dat's just too hot, fer me.
 

Yea my AC wont cool it with 2 PCs running almost all day in the living room. But it still stays decent. But its monsoon season so 100F+ wirth 50%+ humidity just sucks. Heck even 80F with 50%+ humidity sucks.

Heh.. that's basically what it is here... 81F 79% humidity.

Nice grimmy.  3.6GHz out of chips runnng a 266MHz FSB is just a bit too much for most 'setups'.  Few make it to 3.6GHz stable with the Q6600.  The chip is not built for that kind of stress, nor is the 650i.  That's how I see it anyway.  Maybe someone running stable at 3.6GHz with their Q6600 can post their specs or perhaaps a 'link' to such a monstrosity LOL.  I use two of these on my systems.  Positioned on top of the RAM and blowing directly down on top of the chipset area it does a formidable job cooling down my A8N32-SLI Deluxe 939 16 x 16 , which is the hottest running MB made in recent histiory.  Newegg raised the price 5 bucks.
 
http://www.newegg.com/Product/Prod [...] 6835704001
 
This thing can be positioned directly over the RAM, video cards and MB chipsets too.  The Zalman 92mm fan is fast and powerful (same fan as on a lot of their CPU coolers).  I use this in conjunction with the Corsair RAM fan on the A8