If you're going to spend the money to replace with a better one, you may as well buy a better CPU that can still profitably mine.  Otherwise, there's no reason to use anything other than stock.

It's true that you should never mix connections to the motherboard, but that's not at all what is happening with USB risers.  USB risers do not send power to the motherboard - they send it to the video card.  The only thing transferred between the riser and the motherboard is data.  Yes, it's using electricity to do that, but at very low levels - for enough power to be sent over that connection to cause damage to hardware, something else would have had to catastrophically fail.

Mixing PSUs in general is a really, really bad idea, but if you have to do it, each GPU + Riser should be considered a whole unit.  Voltage imbalance between two sources can fry whatever is in between.  The video card is what is in between the sources, not the motherboard.

You're not taking 240V and dividing it - you're taking 12KW and dividing it.  That's going to require 60A minimum at 240V, which is a lot of amperage.  There are 60A PDUs available, but they tend to be a lot more expensive than 30A units.  eBay has TONS of 240V/30A PDUs.  Like most datacenter-grade enterprise gear, it lasts way beyond its warranty, so refurbished PDUs are a great option.  If the wiring has already been run and outlets installed for 240V/60A, your options are narrowed quite a bit by what outlet/plug you need.

Also, please don't make your own as another poster suggested unless you are really willing to risk electrocution, fire, and death, along with being totally out of code and likely to have any insurance claims rejected due to that.

Having a small area actually makes it easier, because it makes for a smaller volume of air, but just having a door that is slightly open won't be enough (not to mention that this is very insecure).

The key is going to be moving enough air - cooling the air isn't really necessary unless you're dealing with temperatures well above 100F on a regular basis.  If you're using open air rigs, you'll need fans capable of displacing all the air in the unit (~4000 cubic feet depending on ceiling height) a few times a minute to keep temperatures steady.  I'm no expert, but this doesn't sound like you're going to be able to easily vent that much air safely to the outside (not just into the ceiling - it will build up if there isn't good ventilation out of the roof.  If you use enclosed rigs, you could create hot/cold aisles, which drastically reduces the amount of air that you'll need to move.

220v @ 100 amps will still give you about 18kW of power to play with (technically 22, but you should avoid maxing it out in order to accommodate spikes and other fluctuations), which would be roughly 15-20 rigs depending on their configuration.  You'll probably be spending at least 1kW on fans, if not more, so keep that in mind when budgeting your power.

I'm running 4 rigs in my garage, enclosed in a fairly small area (~20 cu ft), with a 3500CFM exhaust fan directly to the outside, and a 6" intake vent feeding fresh air.  It's massively overkill since even with inefficient sealing it's refreshing the entire volume of air almost once a second, but even when dealing with 100F intake air, my GPU temps have been completely stable, with no more than a 1-2 degree fluctuation.

I've looked into getting a dedicated space as well, but the only way I can justify the extra monthly expense vs home hosting is if I can densely pack the unit - so I'd want at least 200A service (preferably 400A total) for a 400 sq ft space.  Otherwise, it's simpler and cheaper to just use my existing electrical service and ventilation options.

The title says less than $0.05...

The GPU is not going to be of any use - it's way too old.  If it could even mine, it would never make enough to cover the cost of power used to run it.

You could run 4 GPUs off that motherboard, but keep in mind that the CPU is power hungry compared to a newer proc.

I'd be very wary of an older PSU like that - it's at least 7 years old, as BFG Technologies closed down in 2010.  Even if BFG were still running and had been known for quality (they weren't), a PSU from 2010 is going to be really inefficient.

Don't use cheap PSUs, and don't use dual PSUs on one system if you can possibly help it.  Unless you're just making use of some high quality, lower wattage PSUs you had already, there is absolutely no reason to be using more than one PSU in a rig.  I have never seen a pair of PSUs that cost less than a comparable wattage single PSU of the same quality and efficiency.

If you still insist on using dual PSUs, keep them electrically isolated from each other - a GPU and riser combo should only receive power from a single PSU, never split.  Same for the motherboard.  Use only powered, USB-connected risers -- never the ribbon cable ones.  Failure to isolate can kill even the highest quality power supplies.

With a 7-GPU rig, you're making $35+ per day.  Saving $50 by using a cheaper PSU is pointless if you end up having to RMA it or wait for a replacement.

Pretty close to 100%.  Other than occasional algorithm configuration changes, and the rare power outage, mining is a lights-out operation for me.  I took a two week vacation, with no access to a computer or to my rigs, and everything ran smoothly.  I can remotely monitor via my phone, but it's really not needed.  I get notifications if a rig is down, but by the time I've checked the status of the system in question, it's either already running again, or is in the midst of rebooting.  I've got monitoring software running on a Raspberry Pi connected to the rig network, and if it detects that a rig has not been responding after a certain amount of time, it sends the command to the PDU to power cycle that outlet.

With prices at their current highs, even a half hour of downtime is costly, so I really like having as much monitoring and recovery as possible completely automated.

I had a power outage for a couple of seconds here, and when everything came back up, none of my rigs would connect.  I had to manually start claymore, with the Minerstat client closed (for some reason, if node was running, claymore would close after a minute or two).  I was a bit panicked, thinking that the power outage had messed something up on my end, so I was relieved when I realized it was the minerstat site.

Between the issues that Nicehash had, the flash crash on Ether today, and now this, it's been a tough day for mining.

The only reason I think mining specific GPUs will come out is because of the number of businesses adopting the Ethereum blockchain for various uses. I don't see them being optimized for machine learning or other sorts of tasks, because that would cut into the professional market too much, but a blockchain mining optimized card does make sense for enterprise use.

That sounds about right - the card is similar in architecture to a GeForce GTX 580: https://www.techpowerup.com/gpudb/563/tesla-c2075

A UPS is generally going to provide the most reliable power, but you're going to need a pretty substantial one for each system, and don't expect it to give you more than a minute or two of run time at full load.  If you expand beyond 2 GPU/machine, you'll need to look at datacenter-grade UPS systems.

You should not use a different PSU for the riser than you use for the GPU.  The powered USB-cable risers should never have power beyond that necessary to carry the signal, and will not short something out (outside of a catastrophic failure).  The GPU will draw some power from the riser, and some from the 6/8-pin on the card.  These two sources should be the same.

Don't turn off Update, Defender, or Firewall.  Even if nothing is ever running besides your mining software, a Windows machine connected to the Internet that is not continually patched will quickly become vulnerable to remote exploitation.  Especially if you aren't using a commercial-grade firewall or threat management device at the network edge.

The local Windows Firewall should remain on, and only open the necessary ports.  You don't just get rid of the locks on your house because you don't want to deal with them.

It's better to have some reboots once in a while due to security updates than to have your entire mining operation become part of a botnet, or need to have an OS reinstall on every machine.

http://thehackernews.com/2017/05/smb-exploit-cryptocurrency-mining.html

I'm absolutely amazed at the number of people who still suggest disabling Windows Update.  You don't have to actively do anything on a computer for it to become infected.  No amount of "being careful" is going to keep an unpatched computer secure.  A single compromised machine anywhere on the network is enough to compromise every other computer.

That's about $0.10 USD.  That likely would not even cover the electricity cost of charging those laptops for 3 days.

XMR can still be profitably mined with some CPUs, but just about everything else needs a GPU or dedicated hardware.

Monero CPU Mining Performance

https://www.monitortests.com/forum/Thread-AMD-ATI-Pixel-Clock-Patcher

Have you tried using this?

I'm running one of these in my garage.  It's loud, and pulls ~ 500W by itself, according to the label on it, but it moves the air through the exhaust hose very quickly.  I also have one of these pushing air over the equipment, with everything enclosed.  This allows the ~30 cu ft of air that the miners occupy to be completely refreshed every few seconds.  It's way cheaper to move a lot of air than it is to cool the air, so AC is a total waste of money and energy for this kind of thing.

DO NOT DISABLE WINDOWS UPDATE!  Do you want your mining rigs to end up as botnets?  Because that's how you get botnets. https://arstechnica.com/security/2017/05/an-nsa-derived-ransomware-worm-is-shutting-down-computers-worldwide/

You can block driver updates, but still allow windows updates: http://www.thewindowsclub.com/stop-windows-7-from-installing-drivers-automatically

Had the same issue yesterday.  The fix for it is to use DDU, reinstall the good drivers, then (and this is crucial) block any further driver updates/changes.  The easiest way to do that is via one of these options: http://www.thewindowsclub.com/stop-windows-7-from-installing-drivers-automatically  (Note:  the first option does not seem to be valid anymore with Windows 10, but the Group Policy fix worked perfectly for me)

In my case, the driver that seems to be the most consistent is 16.11.5, though I've had other rigs work just fine on different versions as well.

Windows expects to have monitors plugged in, or they put the cards into a sleep mode, making them perform poorly and sometimes not show up correctly/at all.  Because RX4x and RX5x GPUs have only digital outputs, they can detect if a monitor is present.  That wasn't really possible with analog connections, hence if DVI-I was on the card, it had to be assumed that a VGA monitor (via adapter) could be plugged in without knowing.

There's no known software fix, unfortunately.  Usually you can get by with one dummy plug per rig, but I've seen cases where it takes more than one card being connected like that.