I did decommission the fans and heat sinks on all of my cards to check thermal paste as the MSI Gaming X were pre-owned. They have much bigger (Overall Thickness) heat sinks and have thinner fins designed horizontally so when the fans are on they blow out via the vents.
The Nvidia heat sinks are smaller (Overall thickness) but have thicker fins directed vertically which doesn’t make any logical sense because if they were in a PC housing the blowers would be blowing the hot hair around inside instead of out.
In terms of flow of air volumetric and velocity The MSI have bigger fans and have smaller outlets so at 30% fan rate I could feel it on my hand. On the NVidia the fans are smaller and the outlet has bigger holes and I could just about feel something coming out.
Yes, I’m a bit of a weirdo but then again that’s why I’m employed to analyse measurement uncertainties for fluid flow so it’s a win win to be in a niche job. I hate to think what I’d be doing with a liquid cooled graphics card I’d be using my work devices to measure the flowrates analysing flow profile, measuring the temperatures on different points for thermal dissipation and cooling efficiency. My boss would probably pay me to conduct the tests and then push the sales guys to visit the manufacturers to improve their quality standards
That it why I’ve set VDD to 875 as I made increments of 5 from 800 all the way to 950 and the same can be said with Core clock, Memory clock, DPM, Memory controller, Memory state I was literally happy to blow it but found my sweet spot.
Do you mind if I ask why you were so adamant I need to try these again? was there any particular reason why you were expecting? how about you try my numbers and let me see what you get… I’m a numbers guy so i find things like this interesting.
I’ve tried and tested your recommendation and hundreds of other variations and combination and found my current setup to be the best for “me” what works for you might not work for others as I’ve mentioned before I am 100% on this statement that subtle errors in manufacturing processes means you can NEVER reproduce the exact same copy of the one before or after on the production line. Even if they have a 6 Sigma Process.
Surprisingly it is very very very EASY! Just make sure you have some thermal GPU/CPU paste. Clean with electronic wipes reapply and done Literally a 5 minute job once you do the first one. There is a you tube video somewhere someone has already done to see an example and most have the similar setups.
I am sorry i dont remember sujesting the settings before tno you. With you settings i think my card wont do it because is asrock 570 power hungry shit and i barely made it work with 95 wats power and 31.11 mhs. But i often see these settings used in other rx cards.
i also tested your settings just to see what happens but with core voltage 800.
i got 2 watts reduction and 0.8mhs lower. i get exactly the same mhs like you do but with 95 watts consumption due to my shitty build GPU.
i have tested 1175 and 1200 cores with 875 volts, and memory clock 2080 with mem controler voltage 800. it does 31.85 with 1175 and 31.88 with 1200 but watts are to the roof.
so now i get 31.77 with 1160 core cloc , 850 core voltage and 2080 mem with 800 mem controler voltage.
I understand the frustration only because as I overlook the quality of the electronics manufacturing for my employer and have to carry out uncertainty budget. Unfortunately we always have to allow some +/- tolerance as a part of life nothing is 100%. I was getting frustrated that I bought 3 cards from the same supplier and 1 of them just consumes more energy than the rest regardless of the OC settings.
I had some time to fine tune last night and go through the numbers you suggested and found that I can only use 850 when Memory clock is not set, as soon as I set it the watch dog prompts GPU dead or something along those lines… However I did end up finding what does work and did some more fine tuning and reduced the wattage by another 10W, lost 1MHs and also helped to reduce the temps further.
With energy prices now up an uphill in the UK I went the extra mile and even changed suppliers to find the lowest price per kW with cheap night rates as I plan to Mine 24/7, with the lowest temps possible and the lowest wattage at a reasonable hash rate. My goal was to try and mine at >0.1 Ethereum per month so that by the Time POS kicks in I should have about 0.6. The Calculator estimates 0.1547 ETH per month so I’m happy with that because I worked out that the money I’ve spent on the rig it would take me 5 months to come out in the green. Now I just got to stop tinkering with it
Set your self a goal with some tolerances it’s help with the frustration. I’ve achieved what I wanted for now and I’ll be around on the forums to help if I can but as soon as ETH has crosses over to ETH POS I will move on to Raven / ECG / ETH Classic. Need to find out more about these three as I managed to setup for ECG:
Nicely done. Its strange that the core clock affects the memory. For my card it doesnt matter what the core clock is , the memory controller clock is alway running at 800 and i can go to 2090 without crashing.
but then again that’s why I’m employed to analyse measurement uncertainties for fluid flow so it’s a win win to be in a niche job. I hate to think what I’d be doing with a liquid cooled graphics card 
to improve their quality standards 
