I just went through an extreme experience with my GTX Titan Hydro Copper graphics card, I had already mentioned the sorry mounting of the water block to the card with the GPU die itself only covered with thermal compound about 85%, with one corner completely bare of TIM.
Discovering that made me extremely grateful I had never pushed it too hard overclocking it, I was also thoroughly disgusted at the gelatinous crap thermal padding that was used between the memory chips and VRs, I had never even seen thermal padding that crappy, it was like 1.5mm thick gel, is my best description of it.
I would have known right then the problems I was looking at if I had investigated the memory chip and VR modules impression depth and contact to the thermal padding, as I would have discovered the serious discrepancies of the memory chips out of a flat level plane.
The memory chips were hardware mounted to best describe it as a roller coaster, of course I did not discover this issue until I bought a Watercool Heatkiller water block to replace the Swiftech Komodo that actually comes on the GTX Titan Hydro Copper.
When mounting the Heatkiller and testing for fit my initial goal was excellent GPU die contact so I wasn't even concerned with memory chip issues until I got the dies coverage I was after, it was then I discovered the hidden in plain sight nightmares the graphics card was hosting.
Before I continue I want you the reader to understand that the Heatkiller water block was excellent, however you cannot get good contact when a flat surface is being mounted to a roller coaster, Heatkiller provided .5mm thermal padding for memory chips and 1mm for the VRs.
The VRs were not a problem but the memory chips not being manufactured level and using the supplied using .5mm thermal pad 6 were squished flat and 6 barely even touched the thermal padding at all, so 6 of the 12 memory chips on the card were barely touching the thermal padding at all.
So the initial fitting of the water block to the GPU die was literally a total waste of time as when the thermal pads were added the GPU contact was lifted and not making good contact, that's when it all turned into a nightmare!
One of the worst things about me is I am a perfectionist, and I could not stand what I was looking at, I fully understood why the card was limited in it's performance, because full contact with everything that needed cooling just wasn't happening.
There are 24 memory chips on this graphics card 12 on the front side with active cooling and 12 on the backside with passive cooling, I had already decided with the Heatkiller block to bypass the cards back plate and just mount a set of 12 ram sinks and cool the backside with air.
Even though we are all at some point of buying a graphics card especially one the cost $1300 US dollars after performance but it seems we're getting looks and fluff over raw cooling for best overall card performance and overclocking.
Chock this experience up to another lesson learned the hard way!
So is that the end of the story?
No!
I was extremely discouraged at this point of attempting to get excellent coverage and it forced me as a resolution to run regular TIM on the 6 memory chips that made contact and the .5 thermal padding on the 6 chips that did not, that achieved excellent contact with the GPU die again but but revealed another problem.
The outer 2 memory chips across the top only had about 2/3rds partial contact and required a lot of TIM to touch.
Pissed Off as Hell, comes to mind as I knew I was just not going to be satisfied with what was happening, and this was all because EVGAs manufacturing process of the graphics card in the first place was quality controlled by monkeys!
I am saying this because both of my past 580GTX EVGA cards were memory chip flat, and I had mounted Heatkiller blocks to them with just TIM only and had excellent contact, that's why I stated the Heatkiller water block was not the problem.
So what now?
I had zero intentions of running the card like that, so I had considered getting a universal fit GPU water block and thermal bonding Ram Sinks and cooling the memory and VRs with air.
However I had an additional water block an Alphacool that I had intended to flow modify for the peltier setup, the only problem was it was a clear 1366 socket mount acrylic top CPU water block, and we all know you cannot mount a 1366 socket on a 1155 socket, much less to a graphics card, so some serious modifying was in order.
So end result that I am happy with as for the first time since the GTX Titan left the EVGA factory door the GPU die is fully covered, and every memory module and VR has it's own Ram Sink..
Leak Testing:
Back in Business:
This is my resolved issue, and I shudder to think what may be under some of your graphics cards, thus the reason for posting this, as I thoroughly recommend checking it out for yourself, (if you of course have the hands on ability to do so!), to see what's actually going on under the hood of that graphics card you are running.
Edit: Gaming load has dropped to 28c with a 24c ambient.
Edit: 1, Ran an hour of benchmarking various programs Uningine Heaven, and Valley (let Valley just run for 30 minutes before starting the benchmark), 3DMark 11, etc.the highest temperature reached was 33c at a 26c ambient.
Memory chips on the back of the card are 26.4c, memory chips on the front of the card are 26.2c, (obviously the copper hedgehog memory sinks on the front are the best, taking into consideration they are thermally bonded and the ones on the rear are aluminum and just stuck on).
The VRs are 25c with a 26c ambient? (I have to admit I did not expect that, nor do I fully understand why but I did change out one of the front intake fans to a low speed San Ace 120mm x 38mm, which is a little louder but outputs quite a bit more air volume.)
Heat sink temperatures were taken with a point and shoot laser thermometer.
Facebook
Twitter
Instagram