The chart says it all:
https://bitinfocharts.com/comparison/ethereum-mining_profitability.html

My setup is the following:

- Regular PSU to power only the motherboard+CPU
- Server PSU to power all GPUs, both riser+8pin
- Regular ATX motherboard
- Not connected the grounds of both PSU
- Both power supplies are connected to the same power outlet

The rig works perfectly well when I disconnect the power from both units, wait few seconds to discharge the caps, connect the secondary (server), wait few seconds, and then connect the main one. However after a power outage, both supplies start at the same time and the rig is unstable (boots fine, detects GPUs but hangs after I put load on them) until I manually disconnect both and repeat the procedure.

I know about Pico PSU, but I'm CPU mining as well and those mini boards are not designed for a high load.
How do I solve this issue without replacing either of the units? Perhaps connecting both grounds would help, but I haven't tried that yet.

After lots of researching, trial and error, I have managed to run 10 GPUs on my ASRock X99 Extreme6 using two PCI-E multipliers for $12 each off eBay or AliExpress. It's likely that I'll be able to run up to 13 if I had more cards.

You need:

• Linux or Windows 10 with 1709 update
• Above 4G Decoding enabled
• If using Windows - install it on UEFI mode with GPT partition table
• PCI-E set to Gen1 or Gen2
• Intel motherboard on H61 chipset or newer; AMD motherboards with UEFI.

If you are missing either of these prerequisites, it is quite possible that it won't work. Motherboards without Above 4G decoding can run up to 6 GPUs, more are unlikely. Older motherboards with a legacy BIOS can run up to 4-5 cards max, assuming that you don't encounter IRQ issues. The hard limit seems to be 13 regular cards, unknown more mining cards.

Potential issues:

• If your install is on Legacy MBR mode - you can convert it without data loss using the built-in MBR2GPT.
• If some of the cards are missing - make sure that they're not in Device Manager - Other devices and waiting to have their driver installed.
• If the cards are recognized but output error codes in Device Manager - try nuking your driver using DDU and then reinstall it. Make sure that you patch the AMD driver if you are BIOS modded.
• If the doesn't boot or doesn't recognize the cards altogether - try switching the PCI-E slot. My motherboard refuses to boot if I plug the multiplier on the first PCI-E x16.

More troubleshooting tips - try updating the BIOS to a newer version, disable unneeded peripherals (audio, additional SATA controllers, whatever), disable iGPU, set PCI-E to a different Gen, even reset the CMOS. HackFlags registry trick might work but I've never tried it.

YMMV, of course - please let me know if it worked for you. So far I've tried it on ASRock X99 Extreme6 and Gigabyte GA-H110-D3A (>6 cards), MSI H110M PRO-VD and several other mATX motherboards for 6 cards.


Donations:
BTC: 1P4oRbB76QGw3dzmhcxSFo1B9pxSswipfs
BCH: 1M6TaEKzKYQEQdoBXRbPhzm3m23iYxFhyU
ETH: 0x2211d972934ec7a66225883355e9d23f848f634a
ETC: 0x64b30f6f1acd4ab9f1e5d6da7e7d9d22d506c8c4

What do you guys think about a self-made PCI-E cable that powers all 6 006C extenders in a rig? Daisy chain all 6 connectors with 3xGND + 3x12V 12-14AWG wires which should be able to handle a power draw of 10A per wire or 40-50W per card (less than 300W actual power draw). I'm using a modded 2400W server PSU, so the power is not a problem.