Clevo 860/870CU GPU Thermal Pad Size?

Does anyone know what size (thickness) the thermal pads are on the Clevo 870 or 860 GPU heatsinks? To specify, I am looking for pads that will fit the 5870 heatsink but I believe it is the same as the GTX280. My pads are pretty much toast from swapping thermal paste so much (tried 5 diff brands) and I'd like to replace them. So far I have found 1.0mm, 1.2mm and 1.5mm in thickness but no clue which I should try.

Thanks for any help.

 

Give me a moment, rancid. I'm making a diagram for the most optimal sized pads for the GTX 285M in GIMP right now. I got you covered.

 

I knew you would come through, I should of just pm'd you

 

Sorry for the long wait. ImageShack is being a real butthurt today.

Anyway, here is my little guide pertaining to the GTX 280M and GTX 285M thermal pad sizes. Each of the components are color coded for retard proof reading ease. Below is a picture of my GTX 285M when I first received it months ago. At the time of that shot, it has never been used as seen by the unfaded inscriptions still present on the faces of the chips.

GPU Memory: These need 2mm pads. Anything thicker will result in lifting the entire cold plate too high causing the GPU die to not make contact. Anything thinner will result in the memory making poor contact with the heatsink, and thus having the chips fry themselves in the long run. I'd like to say that 1.8mm will still work as it puts less leverage against the heatsink, but if you go low towards 1.6mm and thinner, then you will have problems.

Chokes (ferrite cores): These need 0.5mm pads. I don't suggest using 1mm pads. I tried that before and it ended up slanting my cold plate offset with the GPU die.

Inductors and controllers: These originally came with 2mm pads, but 1.5mm pads will do just fine. I suggest using 1.5mm since they put less leverage against the back of the cold plate, which will less likely influence an offset angle when mounting the heatsink. Though it's not really that major since the pads are really small.

Capacitors: This is completely optional, but I decided to put pads here anyway because these parts still generate a good amount of heat. Originally, these did not come with pads, but the row of MOSFET's in the back can use 0.5mm without problems, and the single MOSFET inbetween the GPU memory can use a 2mm pad. Author's note: highlight the yellow text if you can't read it you lazy bum.

The MR 5870 most likely uses the same size pads for it's components as well. I'm not 100% sure as I do not have the card, but it should be fairly close enough. But the GPU memory on the MR 5870 definitely uses 2mm as well.

 

Stellar guide Soviet. I will give it a go on the 5870 and let you know. I found a guy on ebay that sells all kinds of sizes so I will most likely just get them from there. Only downside is the 7-10 day turnaround time he takes.

 

Let me guess, you're talking about new-think. I ordered pads from him a long time ago for my M570TU. They're very good pads and have better durability than the ones that originally came with the card. new-think takes around two weeks to ship to the west coast US from Hong Kong.

 

Yup, he is the one. Is a standard pad that he sells big enough to cover the entire card or should I get multiple pads of each size?

 

Send him a PM for a price quote on a certain size. For example, ask him how much would it cost for two pieces of 52 x 15 x 2mm pads, which is the exact dimensions for each bank of memory chips on the card plus 1mm to length and width. Then you can send him your payment through Paypal manually, (not through eBay). Order as many as you need. I don't think you will need anymore than two sets (four individual pieces) for the memory unless you frequently remount your GPU heatsink.

 

One more dumb question I am sure someone will ask anyways, is it better to cut the pieces to size and place them over the desired chip and then lay the heatsink down on top or try to cut the pieces to size and place them on the heatsink then put it over the chips? Seems a lot more room for error placing them on the heatsink vs just laying them over the chips.

Thanks again.

 

Whichever method is easier for you is what I suggest. As long as the thermal pad gets there, it doesn't matter where you put them on first. Personally, I put them on the card first.

 

Actually the parts in yellow are Tantalum capacitors. They are needed for the onboard buck power supply output filtering.

The green parts are inductor coils that are part of the buck regulators.

The blue parts are MOSFETS and the multi-pinned parts are the actual controllers.

Is it possible to get the numbers off the multi-pinned parts, would like to see what they are using.

 

Thanks for the proofread, othonda. I'm falling behind on my EE knowledge; too much DSP crap to study for in class.

It's really difficult to read the inscriptions now since they have worn off from the pads, but I'll get try to get a picture in next week. I have to prepare a field trip to the Robotics Society for the kiddies.

 

No problems with that. Are you studying DSP hardware or software? At work we are putting an ARM cortex DSP processor as a coprocessor on a board that I am in the process of getting running. This board also has a TI micro as the main processor, There is also an Epson video controller, a fujitsu touch screen controller and 3 pic processors for some analog circuit redundancy.

 

Right now we're researching for a project as to why the iRobot Roomba Wet/Dry cleaner's color navigation sensors controlled by a cheap DSP fail to maintain accuracy, or even operation, over the course of it's use aside from dirt and sediment forming over the optics lens. But my job is to figure out why the Roomba's "check tank" light keeps saying that the tank needs to be cleaned when it really doesn't. When I disassembled one of the Roomba's that we bought, the water tank has two DC charged pins in it that checks the water level as well as the pollution in the water. The pins are made of stainless steel, so of course it will eventually corrode over time and the corrosion rate is accelerated when charged and when exposed to dirty water. A basic textbook solution is to replace the pins with a better material like titanium or gold, but that would drive the cost of that cheap $0.30 pin to $1 or so. From an engineering standpoint, it's the best solution, but from a sales and economic standpoint, it's very expensive. Recalling thousands of those units to replace the pins would be a costly decision. So I'm working on a tentative solution to present to my team to submit to iRobot that will solve the problem at literally no cost. And my proposed idea is to instead charge the pins with AC instead of DC. The stainless steel pins are far more resistant to corrosion and exposure under a AC charge. It will not affect the mechanism and the inverters on the robot can output the small enough amount of AC to the pins. All that needs to be done is create a separate line from the robot's propulsion pinions. All of this can be done at the software level, meaning that all the end-user would need to do is dock their Roomba and download an updated firmware that addresses these changes. Actually programming the firmware will be up to iRobot though, and our SE doesn't know how to fiddle with it. But in the end, when the tank thinks needs to be cleaned or replaced, the Roomba will travel a short distance when engaged, then break down because of the warning. This as well as the sensor modules are what need to be addressed in this project.