Author

Topic: Antminer and GPU rig power and safety questions (Read 256 times)

newbie
Activity: 35
Merit: 0
December 12, 2017, 09:22:38 PM
#3
Thanks for the clarification....  Got the information from this thread www.geeks3d.com/20101108/tips-maximum-power-consumption-of-graphics-card-connectors. Must be talking about the power connector limits on the graphics card itself for the 6 pin connector and not the limit of the power supply's 6 pin connector. I just misinterpreted what end they were talking about i guess.

Thanks again.
legendary
Activity: 3822
Merit: 2703
Evil beware: We have waffles!
No idea where you got your info about the PCIe connectors but it is very very wrong. The connector is Molex mini-fit, re the specs from Mouser PCIe Connector Specs
Try 13A per-pin, 3x = 39A x 12v = 468W

A good designer will keep things under 300w per-connector for margin.
As for burnt connectors - virtually always the result of crappy or worn out PSU side connectors and/or using too thin wire. The max current rating only apply when using 16ga wire which will help pull heat from the contact areas.
newbie
Activity: 35
Merit: 0
Hi everyone, this is my first post. I am in the research phase of deciding which direction I want to go with mining. I research the heck out of everything before I jump in. I live in Canada (eh!).

So in all my research, it seems as though from an initial capital standpoint, the ROI is much faster with the Bitmain Bitcoin and Litecoin ASIC hardware compared to GPU based rigs which would take more than twice the time for ROI. Because of this, I started investigating ASIC hardware first. Setting up at home, noise was an issue so I am really leaning heavily towards the Antminer R4.

Several things I noticed in my initial investigation:

1) The APW3 and APW5 power supplies from Bitmain don't seem to have any electrical certifications that are relevant to North America (FCC is just for EMF emission standards). So for example: c(UL)us, CSA, c(ETL)us, c(TUV)us, or component specific electrical certifications like c(RU)us. So that was a concern for me. Solution? Buy a 3rd party PSU like EVGA that is electrically certified for Canada.

2) I've noticed users complaining about fire risk / hazards with various models of Antminers. The S7 had that issue where if it lost Internet connectivity it would overheat. I think I saw a similar issue posted about the R4, but I would hope all of this was solved with firmware updates.

3) Burnt PCI-E 6 pin power connectors.... I've seen a number of reports of burnt / melted 6 pin PCI-E power connectors on multiple models of Antminers, so this one required more investigation since I'm very safety concious. This one is the big one to which I'm turning to this community since the math I'm doing is giving me answers that I don't want.... The Antminer R4 has six PCI-E 6 pin power connectors for powering the hashing boards and one PCI-E 6 pin power connector for powering the controller... The unit is just over 800W overall (840 or 845?Huh). Anyways, I am assuming that the majority of the power for the unit is going to the hashing boards and not the controller. So even assuming the hashing boards are using 700W of the total power and the controller board is using 100W, 700W divided by 6 is almost 120W per 6 pin PCI-E connected to the hashing boards. Correct me if I'm wrong, but everything I'm reading is saying that six pin PCI-E connectors are rated for a max of 75W. If that's the case, the 6 pin PCI-E power connectors are being driven way beyond their rating and that concerns me. I don't have any solution to this one because it's the design of the unit that I can't do much about. Thoughts?

GPU:
From a GPU standpoint, I've noticed people mentioning issues with the molex to sata power connectors and to avoid them. Also not to power more than 2 risers per PSU power cord... Any recommendations for a super safe GPU setup?

Thanks,
Jump to: