The Failure Continues

At this point I took my findings to GELID and I was told at first that maybe I got a pre-production sample. I think to myself that this is very feasible, so I tell them let's go ahead and try this again. In the mean time, I also asked if they could show me where they had the cooler mounted to a 470 for testing. I was told it fits on an Inno3D card and had seen in forum posts that it fits an MSI version as well. So what is the deal; why not the EVGA?

In the micrometer is the part B standoff which is the one the instructions say to use. These measure 4.34mm and GELID says there is a +/- of 0.1mm for these.

Just for giggles, I measured the part C standoffs. I figured I would need to check if maybe it was a simple typo. I found these to be 2.60mm and in no way will the add support to the card. The overall length is shorter, so tightening them all the way down will severely bend the card and possibly cause damage.

Since I had this handy little add on to the micrometer, I figured I should go into a battle of numbers with all the information I could find. I set the depth gauge and locked it in place so I could get the next image.

So my IHS on the GTX 470 sits 3.82mm taller than the PCB. I also got into contact with someone in the know at NVIDIA and was told the height of all GTX 4 series cards have a specified height of 3.98mm (+/- 0.17mm). This leaves my IHS on the extremely low end of that measurement, but still within NVIDIA spec for the card. Now, at this point GELID argued that their coolers base sits 0.7mm taller than the riser. So in essence 4.34mm minus the 0.7mm should leave for 3.6mm of height. As the image of the base showed, 0.7mm of clearance is on their best milling day, and mine obviously wasn't made that day.

So along comes the second Icy Vision sample and right away I pulled out the standoff and got a measurement. Now, I had already asked how this was going to magically work, as I knew I already had the production sample. I mean why wouldn't I? - The cooler was on retail shelves as I got my sample. When I placed it in the micrometer there was no change; actually, these standoffs are a bit taller than the original. This cooler isn't going to fit either.

The only way I was able to get some sort of contact was to use the shorter part C standoffs. Even when I got contact in this configuration, you can see my dilemma with the base of the cooler. The edges are clearly making contact and the center I can fill with compound, but if you look closely the standoffs don't support the card around the core. I don't like to have to guess at the correct pressure needed for the cooler to function properly, and I sure as hell don't like coolers that bend your card like a taco.