nF2 chipset lapping - how far?

[trodas]

I figured out I better ask twice (to be sorry once) before I start.

I have one old DFI LP2 board and it is unstable in the long run (15h of folding and crash)... I noticed how bad the NB chipset contact with cooler on is (almost hole in the middle of the plastic on chipset!) and going to lap it. How far I can go?

I read sometimes as far, as I see the cooper?
True?

http://www.slibe.com/imagestumb/188c...ipset_temp.jpg
(the nF2 chipset has no visible metallic part ATM, DFI NF2 mobo)


Even w/o the lapping it can score fairy high O/C, so I wonder how I can take it with the lapp


edit: find some pics like in this guide: http://www.techarp.com/showarticle.a...=70&pgno=0
But looks like they just flatten out the surface? I would like get most of the temperature-resistant PVC out of way, so, what is why I asking how deep can I go. Till the little holes near corner disappear? Till I see cooper?

[Gizmo]

If you go down to the copper, I think you'll be getting into the chip, which would be a Bad Thing.

I've lapped mine down to the point that the plastic corner holes are non-existant, but that's as far as I dared go.

You might also want to give thought to lapping your south bridge.

edit: You might take a look at this thread I wrote a while back.....

[trodas]

I see. You lapped the NB nicely, however you did not dare to go much deep... The question is like if one can go deep. I was told I can...

And I also believe that see the cooper inside mean to see the bottom cooper plate on witch from other side the chip is. Hence no harm should be done... But I wanted to get that confirmed and seems no-one dared...

I have a big Zalman on SB already glued with Arctic Silver epoxide. It is not going anywhere...

http://ax2.old-cans.com/s.php?p=badt...8&c=8&d=1&v=v2

[trodas]

SAE: "there must be a spreader, t would be nonsense not to integrate one, good luck, mate^^"

So, the lapping experiment is done, mobo still works, and I gained some knowledge. First and worsest of all, there is NO heatspreader inside of the nForce2 chip in these later revisions with completely plastic chip - unbeliably as it sounds, but true for the nForce2 SSP Ultra 400 in my case.
Let pictures speak for themselves:

http://www.slibe.com/fullimage/97b91...2_lapping_.jpg

http://www.slibe.com/fullimage/d0539...2_lapping_.jpg

Unfortunately nVidia showed me again what I would call "nonsense" at best.

Testing lapping (pictured) was on JetWay N2PAP Ultra. I dubt the mobo will work, but it was dead one for sure before, so... (two caps in the Vdimm powering explode and took the mosfet with them...)

The chip thickness (measured over the green bottom) is exactly 1.25 mm.
When I reach 0.475 mm the resistor at one end and then these wires begin to surface - so, I lapped my DFI LP B nForce2 core to 0.50 mm and I hope for highly increased overclocking of the chipset

When I get there a good heatsink, we see

[Gizmo]

Originally Posted by trodas

Unfortunately nVidia showed me again what I would call "nonsense" at best.

Eh, I'm inclined to disagree.

There are very few chips that integrate a heat spreader, and none that I know of that integrate a heat spreader beneath the plastic. If you have enough power dissipation to have need of a heat spreader, then it would be monumentally stupid to hide it under an insulating layer of plastic. That would be defeating the whole purpose of the thing, don't you think?

[trodas]

Well, the whole nonsense is about isulating the core from heatsink by 0.8 - 1 mm thick layer of plastic. Even some claim that this is "special plastic" that "conduct heat reasonably well" it is still nonsense.

If you remember, earlier versions of nF2 has metl heatspreader. Then nVidia commited the "nonsense" (very polite word he used...) about removing it, as "not necessary" as my czech overclocking pal remember clearly.

It turns out that it is necessary to remove most of the stuff... In fact, before was my heatsink on NB cold. Now it is hot. Finally!

I can't wait to get serious cooling and try what can I now clock out of the recapped and lapped mainboard now


PS. I was at first really expecting that there is circle of cooper, under it is the real silicone core. Hitting the wires was shocking to me... But I have to say that when I shaved so much of the plastic, I begin to worry that the silicone is, in fact, mounted on the bottom and that mean that I hit it directly, probably... I did not expected the wires, but yes, when one think about it, is is logical
Nothing ventured, nothing learned.

[Favu]

Trodas, you are crazy! But I have to say it is cool that there are people still getting right down and dirty with their hardware, you haven't seen as much hardware modding recently as in previous years

[Áedán]

Originally Posted by trodas

If you remember, earlier versions of nF2 has metl heatspreader. Then nVidia commited the "nonsense" (very polite word he used...) about removing it, as "not necessary" as my czech overclocking pal remember clearly.

"Not necessary" as in "The motherboard will work fine within specifications", which is what manufacturers aim for. It isn't what overclockers aim for, but the overclocking market is not one to build a company of nVidia's size on. It is also cheaper to use plastic rather than a metal heatspreader.

[trodas]

Favu - yea, I'm! But it feels so good... And you are right. People went downright lazy. Time to shake the grounds a little


Áedán - you are right, except one thing. The mobo did not work with the specs. The machine crashed after like 8h of folding. With 7V 120mm fan blowing the air directly to the NB it can last like 12 - 13h, but still crash. Now it is running for 15h straight and no sign or weakness

I running not at 200Mhz (x12.5) so 1.6V - lowest possible settings.
I must say the heatsink was used to be kinda cold. I say very cold, despite the Arctic Silver 3 and two screws, that tightly holding it to (okay, to most of it, 60 - 70% for sure) the chip.
Now the heatsink is pretty heated up! Definitively not bellow room temp (25°C) as before! 26°C now, gotta check temp... 47 - 49°C, depend on fin. That is at least 20°C difference now...

So, the bad shape of my chipset caused me endless suffering and a lot's of problems. So IMHO nVidia failed badly.

Futhermore I get slight suspiction that they might used the plastic, because it is not only cheaper, but make the heatsinks - colder! And they perhaps want that, since at the nF2 times the chipsets aren't used to enerate that much heat, so to not scare ppls and save bucks - they simply used the plastic, that "fake" the real temperature at the expense of the core is getting hard beating with overheating...
And, of course, limited performance.