Finally have something fun to share about immersion. This is a side-by-side comparison of two S17+ 76t in the same tank, using nearly identical total power. Why did I do this? Autotune just finished on the left
1) Because I wanted to show exactly what immersion can do, for something as unloved as the S17s. I've also been asked exactly how much efficiency is lost in overclocked. So thanks to the latest available release, @braiins_systems is able to show us what that looks like.
2) So we have two identical miners. Both have been stable for me for a long time, both are in the same physical tank, so, same oil, same cooling, same oil temps. And we can see exactly how well oil does at increasing potential.
4) Note that the reason one card is disabled on the right miner is because the stock power supply cannot handle the 4.8k load. Since Braiins allows the total power to be set, then divides it by three, I can send 100% of the power from the supply to only two boards
15% decrease in efficiency, 38% hash rate increase. Nearly 0% heat increase measured at chips or boards.
Boom! Science!
And all of this is what I'm working towards with tank design, especially with the tanks we will be making at @aiproenergy
Minor update, I rounded when I did the math at 15% less efficient. Going to two decimal places puts it at 11.23% less efficient.
• • •
Missing some Tweet in this thread? You can try to
force a refresh