One last update on lapping the stock cooler.
I took the card back apart to give everything a once over and once again slapped a bunch of paste on there. I removed the nylon washers, as it seemed to be interfering with screwing the backplate bracket all the way down. I made each screw extremely hand tight. Any tighter and I fear I'd damage the card.
Additionally, I spoke to the OCN user who first lapped his cooler and created the picture guide. After an in-depth discussion with pictures of my work on the cooler, we concluded that I'd actually done a pretty good job on lapping the cooler. We further verified this by running Firestrike at the same settings as his card and comparing scores and temperatures. Temperatures were nearly identical. Afterwards, I began to push the card to it's limits in Firestrike with Registry Mods. Final results before I went to bed:
Core Clock: 2122 MHz
Core Voltage: 1185mV
Memory Clock: 1200 MHz
Max Core Temperature: 68°C
Max Junction Temperature: 90°C
Fan Speed: 100%
Driver Version: 19.3.1
Firestrike Graphics Score: 31722
Some Observations:
While benchmarking, I get several 100 MHz dips below the average frequency, which I can't seem to mitigate. Perhaps I need a more powerful registry mod, but at any rate it's affecting my scores. Also, fan speeds get stuck at 100% even when I change them back to 50% in between test runs (purely for noise), which is pretty annoying. Biggest thing to note here is that these clocks are in no way even close to being usable as daily settings or for gaming. I even tried running 4K Superposition with the above clocks and I think I made it to Scene 3 before Junction temperatures hit around 103°C. I stopped the benchmark. Firestrike doesn't use as much power or push the card as hard. Realistically, for extended gaming sessions, my maximum clocks are 1900 MHz @ 1033mV with the memory at 1100 MHz and fan speeds topping out at 55% for noise purposes. Memory speeds also seem to barely have any effect on gaming benchmarks. I benchmarked The Division (DX12) yesterday several times with different settings. Going from 1000 MHz to 1200 MHz increased average framerates by about 2-3% (108 FPS to 110 FPS). This is margin of error stuff and not worth the extra heat.
All in all, I was enamored by the amazing temperatures that some users were achieving with the lapped cooler. What I didn't realize was that this was during Firestrike, which isn't nearly as demanding as 4K Optimized Superposition and maybe 1-2 minutes shorter. Lapping the cooler certainly has made a difference in junction temps, as I mentioned earlier in this thread, by about 6°C or so in some situations. I was aiming for a dramatic decrease in temps, but this cooler just isn't up to the task. I daydream of maybe strapping three 80mm fans to the heatsink instead of the stock fans and shroud to see if there would be any increase in performance while being dramatically quieter. If you want to push this card passed 1950 MHz or so without the ridiculous noise of the stock cooler, turn to water cooling.
Hope this helps someone in the future. It's been fun playing around with this card. I'm currently on the fence as to whether I'd want to water cool this puppy later on down the road. Perhaps Alphacool will release an AIO. That'd be ideal.
Facebook
Twitter