If you’ve ever dealt with mining cryptocurrencies, you certainly know how much this process generates extreme positive temperatures – especially in summer weather like now. I have a small rig with a power of about 1kW and hashrate for ETH ~ 250MH /s. It doesn’t make a lot of profit for now at a price of $ 0.17/kWh in the region when I live – about $300 after deducting electricity costs of around 800kWh for one month [about $135]. The mining adventure has begun, so now it has to be continued.
Today I was spending my time looking for some suitable coins to mine. As I’ve only been mining Etherum so far – temperatures have reached a maximum of 70C [one GPU] with fan speeds around 75%. Today I added two algorithms [CCX AND TRTL] – in one task – to take full advantage of the available memory available on all GPUs.
After a few attempts at the “out of memory” error, I was able to run three algorithms on seven cards with maximum memory usage. Maybe cryptonight algorithms need different OC settings, I went it to OC for ETH. As today it was damn hot with the windows open and the card temperatures reached 70C with 75% already around 10AM. After starting task they jumped to almost 80C.
Every now and then I check the condition of the cables and the heating status of the components – especially the power supplies. I have a 2x750W Volcano from Modecom. For some time I have noticed that one of them is much warmer than the other. I powered with it the MOBO, 3xAMD8GB, SSD, additional fans and 1xNVIDIA8GB connected directly via full pci-e slot [the card took the rest of the voltage from the second PSU, which additionally fully powered 3xAMD8GB via risers and CPU.
I decided to turn the device off for a while, took it out into the air, cleaned the components with compressed air [surprisingly they were not so contaminated after about 4 months of use]. I made some minor changes to the powering of the cards from PSU’s. On the warmer power supply so far I left MOBO, 3xAMD8GB via risers, SSD and took off NViDIA card from full pci-e and added a CPU. I powered the other 3xAMD8GB and 1xNVIDIA8GB plus FAN’s with the second power supply via risers.
I get a lot of air circulation by removing quite a large card from the inside of the device and mounted it on the edge of the rack with screws after drilling a few additional holes.
After reconnecting the device around 4PM – still hot as hell – on the same task – the temperatures did not exceed 70C with the autofan function set to a minimum of 73%.
Merely mining three algorithms in one task did not turn out to be profitable – a difference of about $100 less and 100W more compared to ETH alone.
Below are photos of the installed card.
I know that I should not power graphics cards through SATA connectors [I have 4 of them ; -]. I managed to buy cheaply in the local store three modular ones with a P4 tip matching my power supply, and in turn I also bought two quite expensive in the store on online auction, which after connecting began to melt at both ends of one pin. Maybe it’s a little my fault because not all modular wires are compatible with each power supply, as well as I could try to connect a card to 8PIN with a 6PIN tip. SATA wires used in above rig have no traces of overpass or damage. They are not even warm during a large load.
And finally, an alternative way to quickly cool the room.