Looking for TIM info

[diluzio91]

So, i'm looking for info on 2 different ends of the TIM spectrum

Best Overall to replace lapping a cpu

Best Non insanely expensive to replace lapping a CPU

I'm not really comfortable lapping my 2500K at the moment, but I want to overclock for the UW-LAN party, so I have till about Monday to order to make sure I have it by friday. I'm looking for a good not stupid expensive tim, that can make up for what I can assume are very poor surface mating from an intel heat spreader. My i7 would idle at 3.5ghz (i7 870) at about 30-35*C, and this 2500K at stock clocks is idling at 40*C with the same case, room, ambient, and cooler. Lapping will happen but not quite yet. Looking for a stopgap measure. thanks guys

[farlo]

sort of related, but interesting nonetheless: http://www.maximumpc.com/article/features/geek_tested_17_thermal_pastes_face

[diluzio91]

Yeah, I keep hoping Kayin will stumble on this... he seems to know everything about paste.

[Kayin]

You rang?

Best is Indigo Extreme, best cheapish is Prolimatech PK-1, last I checked.

I have something in the lab that might make people happy, though. Need to test it sometime.

[blaze15301]

I have always used artic silver 5 but I herd they have this new diamond stuff.

If you can find it it is called ic diamond 7 carat thermal compound. When I herd about it a year or so ago it looked promising.

[diluzio91]

should i put up extra? i think the problem is that theres an air gap that doesn't make contact.

[TLHarrell]

More thermal grease is not a good solution. Check the flatness of the top of your die and the interface on the cooler. Best way is with a high precision steel rule or other flatness gauges. A set of automotive feeler gauges would work well too. Place the edge across the die top and see if there are any visible gaps.

[diluzio91]

More thermal grease is not a good solution. Check the flatness of the top of your die and the interface on the cooler. Best way is with a high precision steel rule or other flatness gauges. A set of automotive feeler gauges would work well too. Place the edge across the die top and see if there are any visible gaps.

I already know that the cooler is flat, and the chip is warped. I'm looking for a stop gap solution before I lap the processor.

[Kayin]

PK-1 ought to have enough bondline strength to handle that, and IX will as well. It's really best if you have a severely warped IHS, but you can also RMA it to Intel citing out-of-control thermals.

[diluzio91]

PK-1 ought to have enough bondline strength to handle that, and IX will as well. It's really best if you have a severely warped IHS, but you can also RMA it to Intel citing out-of-control thermals.

I'm getting insane thermals. lol... can I send it back to intel if i'm not using the stock hsf? I didn't even know you could do that honestly

[d_stilgar]

I use IC Diamond (http://www.innovationcooling.com/overview.htm) thermal compound. It's 92% diamond, and diamonds are the best conductor of heat (from a material science standpoint). I like it a lot. Not as good as Indigo, but it works on stuff that the Indigo cards don't work with.

I would say if you know the chip is warped to first try for the RMA as long as it's under warranty and then go ahead and try to lap it. There are plenty of guides out there so it's hard to do wrong at that point.

[Kayin]

If you're getting that bad of thermals, contact Intel, tell them what's going on and that you measured the IHS as being nowhere near flat. If they don't want to exchange it, your only hope is to lap it down.

Expecting a TIM to fill too large a gap will actually insulate the processor, making temps even higher. We don't lap for geek cred, we lap because it needs it.

[Cale_Hagan]

If you're getting that bad of thermals, contact Intel, tell them what's going on and that you measured the IHS as being nowhere near flat. If they don't want to exchange it, your only hope is to lap it down.

Expecting a TIM to fill too large a gap will actually insulate the processor, making temps even higher. We don't lap for geek cred, we lap because it needs it.

that's why i lapped my 950... :whistler:

[Twigsoffury]

You know i've noticed that the way you apply it almost matters more then what the stuff is made of in itself.

But if the top of your processor looks like a stadium dome then there isn't much you can do.


RMA or get to upside down table sanding with jewlers rouge.

[diluzio91]

You know i've noticed that the way you apply it almost matters more then what the stuff is made of in itself.

But if the top of your processor looks like a stadium dome then there isn't much you can do.


RMA or get to upside down table sanding with jewlers rouge.

Lol... anyone remember that lapping starter kit? :D I'm gonna hit up intel first but something tells me that to them 40*C isn't anything they would worry about

[d_stilgar]

This video shows the importance (as well as the correct way) to apply your thermal compound.

EyXLu1Ms-q4

[diluzio91]

http://a2.sphotos.ak.fbcdn.net/hphotos-ak-snc7/432007_10150558652823426_601993425_8987741_7497778 47_n.jpg

Lapped... look how bad that was. lol...

i'll update when i put it back in. i rinsed it off, put it in alcohol, and patted it dry. then left it to dry on its own for a bit...

I also only have 1000 grit, its actually very shiny...

[diluzio91]

Update. whell. we're looking good here. I had to use ceramique, but we're at 35*C idle, and i havn't tested load temps yet. but I survived my first lapping. I'll get some more temps and such when I have good thermal paste and time. Which should be tomorrow.

[diluzio91]

lol... so after lapping cpu temps are.... 26*C idle... now to OC and check temps.

[d_stilgar]

Good to hear!

[Konrad]

From that Youtube vid (http://www.youtube.com/watch?v=EyXLu1Ms-q4):

"Theres no need for the paste to cover entire CPU.

It is not necessary to cover the whole CPU/heat spreader. Because most of the heat is concentrated around the core area which is at the center. Hence the center of the CPU is more important."

While this is true enough, it seems to me that maximizing surface contact between the proc and heatsink is going to maximize thermal transfer. A circular blob will only cover π/4 (~78.5%) the same area as one which fills the entire square. Maybe 90% of the heat comes directly off the core area in the centre, but ~22.5% of the remainder (ie: ~2% of the total) heat will not have any TIM coverage at all; it'll either sit where it is (in the proc) or eventually dissipate towards the center (where the core is already producing heat). Maybe 2% isn't much, but then again maybe it is.

Me, I like to make sure the "circular" blob is big enough to nearly touch all the corners, it almost always drips out the sides when I first apply pressure.

[d_stilgar]

I agree. I'd rather get full coverage than not, but I think the video shows that you would rather get that 80% coverage in the middle bubble free than have a ton of bubbles by trying to spread it all out.

[diluzio91]

true, I've had a ton of luck with the 5 dots method, one big dot in the middle, and a tiny dab on the corners of the proc. havn't had an issue yet so i guess that's good. So many opinions on thermal paste... someone needs to do a hard core study on application methods.

[Neodymium]

Some say one blob in the middle and when you press down the heatsink, if you see a perfect or kinda perfect square, then you're golden...there are other types of applicators for example like a nail polish brush o_0 etc. You cant put too much either, like stuff dripping off the sides of the die...