I've finally started looking into the better parts for the next mass-air cooling system. This is to be the second phase--the first already being in use. I have some Flexhaust ADC duct tubing to purchase, and about 300 lbs of HDPE (High-Density Polyethylene) sheeting to pick up and mill. The HDPE will have to wait. I'm limited on build-time, so I'm choosing to upgrade the intake and exhaust ducting, and cooling.
All that aside, I'm going to attempt to post daily stats updates for anyone interested in what sort of performance I'm getting from my miners. So, here is the first:
As of 08-Feb-18 09:49 EST:
The over-all mined value in USD has increased by nearly 25% in the last 32 hours due mostly to the market price recovery. I'm not banking on it staying this high at the moment, but it won't matter until the 15th (distribution day).
By replacing one of the intake blowers with a higher capacity, I was able to offset the negative pressure the filter was causing completely and start drawing in a total of 1,100 CFM--I'm still 300 CFM short of my goal, but the upcoming duct upgrades will help me keep the pressure issues at bay. I'm pretty happy with the 17°C intake temp, but too much of that is the temperature of the air coming into the cooler from outside (last night I saw an intake temp of 15°C). The answer may lie in a larger a/c condenser, but more on that later.
The 3.7 GH/s makes me very happy--the 0.7 MH/s over default clock-speed results is very nice. Due to the "chip lottery" the Miner Control Program (that's right, folks. The MCP) has two primary, active functions; One it looks for heating and cooling trends and limits during the passive monitoring function and adjusts the machines' fans accordingly; Two, it watches for failing chips (chips that have hardware errors) and finds the nominal clock speed for the heat conditions and tries to keep the chip failures to zero.
This is very effective when you consider the PSUs are not exact machines, either. The performance varies slightly. Some of this is even due to the fan speeds in one or more of the other miners. At any rate, the MCP is written to record that data and prevent faults while running as fast and cool as possible.
Each miner has its own personality based on it's PSU, the quality of each chip as well as each PCB (Printed Circuit Board or blade), and generally acceptible flaws in manufacturing. One of the upcoming changes will be to group the best performing blades into two miners, the averages into three, and the worst of the lot into the last (based on the counts I have of each). The last will eventually be replaced and will serve as an experimental machine (within reason, I don't want to potentially brick it).
The miners are currently clocked as follows:
Miners 1 - 5: 475 MHz
Miner 6: 500 MHz