So I guess I messed up putting the sapphire card block on a few months back. EKWB blocks come with three different screw sizes. For simplification I'll call them small medium and large. I was supposed to have used the small screws. I used the medium screws so I had poor contact with the gpu. The reason I had better temps when I added a support post wasn't because I'd prevented the card's pcb from bending, though it served that purpose, it was because that was what was holding contact between the block and the gpu. When I took the block off the the sapphire rx480 the tim had pumped away from 1/5th the gpu. It was bare. That's why I was getting thermal shutoffs. It was the 1/5th of the gpu with no tim frying. Amazing that it lasted 3 months. Well 2 in this condition. Right before I took the block off I started getting thermal shutoffs on idle... so yeah. My Sapphire reference rx480 card is dead. This probably is why I kept getting degradation on the memory oc. 1/5th the chip was cooking.
I have the ASUS rx480 reference card in the system under water right now. It's a much lower ASIC score card and trickles memory errors at a higher range of ocs. From 2115 to 2135 I get 1 or 2 memory errors every 10 to 30 min. I have to go down to 2110 to get 0 ever. I have to run much higher voltage to get 1400 stable. I'm at 1.190 volts right now. It may need a little more, but Heaven and 3dmark firestrike ultra stress are stable at 1.18v. The extra seems to be necessary for deus ex mankind divided dx12 bench + I added some because I had some artifacts in GTAV 4k once. Temps are much lower with the block assembled correctly. I idle at ambient now. It's about 1 degree lower idle than my cpu. On load I'm about 25 degrees lower than my old setup. I never get out of the high 30s gaming unless my room heats up. Even 3dmark ultra stress never got above 38 degrees. VRMs are about the same but maybe 2 degrees lower. Way higher than I expected on water. The highest I've seen after many hours of gaming and my room heating up to ~28C was 64C. Mostly they are in the 50s. I used Arctic pads this time. I never got to test the phoyba pads as they were put on the dead sapphire card. The phoyba pads were the easiest the peal and assemble. The arctic pads came in 2nd. the EKWB pads were the most difficult to peal. Since I had my block assembled incorrectly, I don't really have a way to tell the difference between the arctic pads and the ekwb pads from a temp perspective. For $10 you get way more pad and easier assembly with the arctic pad vs. the EKWB pads from my perspective. I think they were 140mmX140mm squares. Maybe 120mm. I can't remember. Thermals seem about the same.
I moved my bottom 240 60mm rad to the outside of my case along with my pump\res. I also now have my tower case laid flat. I did this so the wight of the waterblock wouldn't stress the pcb. I found that the bend of the tube going from my raystorm to my EKWB gpu block was pushing on the pcb and causing a little bit of flex still. So I "complicated" my loop and have much longer tube runs that follow closer to the natural curve of the tubes. I also go pump/res->raystorm->360 rad->420rad->EKWB gpu block->240RAD->pump/res in my flow now rather than more logical pump/res->EKWB-GPU->raystorm>360 rad->420rad->240RAD->pump/res.
I didn't bother drilling holes in the back of the case. I just laid the cover on top of the tubes to keep dust/cats out. It kinda looks a little like Bane's back criss crossed over the back of the case. I also have the back as the new front so I can get tot he ports easier. 3 of my four 200mm fans came with my TT Core V71 case. They only light up if they are hooked up to the built in fan controller. They don't light up hooked in to my lamptron. Unfortunately, the built in fan controller doesn't remember settings. It only has fast and slow speeds. After returning from sleep it goes back to fast. Fast is too loud and i got tired of having to hit the button four or five times a day. So they're hooked up to the lamptron now but no longer lit. The "top" and "front" sides are off the case. It's just pointless to have them on anyway other than as dust collectors.
As my 240mm coolgate has 5 ports, I now have a much easier time filling my loop. Draining is still a pain. Due to me not drilling holes in the case, the tubes go high enough that I can't just tip the case and get it to drain out the bottom of the 240 more than 1/3rd the volume of water. It's still necessary to flip the case over and disconnect fittings to get the 360 and 420 rads to drain. Pain. I used EKWB EVO fluid this time. I had been using straight distilled water with a silver coil. The silver had started to react with the nickle in the ekwb block and cause some tarnishing. I also saw some light scaling on the tubes/ block. The EKWB evo fluid ate the light scaling right up. Everything is clear again. I"m only using 50% evo fluid and 50% distilled right now. My loop holds about 2 liters. That's 32$ to use full evo EK fluid. I don't put fluids that expensive into my body, seems a little much for the computer. However the results so far are nice.
Here's the question: would it be wrong for me to RMA the Sapphire card? It seems kinda dishonest. I'd gotten permission to RMA it previously for other issues, but now it died by my own hands. The card has been reassembled to stock. I don't know.
Edit: I should mention that I also over tightened one of the terminal screws on the sapphire block and cracked the plexi. It was like my hand palsied or had a mind of it's own. It cracked and the screw turned an extra 1/4 rotation. As a result, I leaked water on the Sapphire card. While I used a hair dryer and waited several hours, that may well be the real reason it died. It never had water on it with power going to it, and I've had luck with drying cards in the past, but at this point I don't know if I fired it by thermal death or by water.
Edited by greytoad - 10/3/16 at 9:59am