Bronze versus Platinum PSU's for Gaming / Cryptocurrency Mining

The most expensive motherboard I've ever owned.

Introduction

I purchased a kill-a-watt and, with the market being down, decided it was a good time to optimize my mining rigs by first starting with my 'home' computer to balance performance with power use. For each test I loaded the PSU with 40 to 50% of its max powerload paying close attention to the power curve on the box with 115 VAT power.

Document differences between configurations of case/caseless setups.
Compare savings of a Platinum PSU versus an underloaded bronze PSU.
Decide wether or not to use a Platinum PSU for a home 'gaming' rig that uses the fastest GPU for gaming and a secondary PSU for mining; and both when not being actively used.

The Computer

Motherboard	CPU	RAM	OS
MSI 370 Gaming Carbon Pro	i5 8600k (8th gen)	Crucial 8GB Single DDR4 2133 MT/s	Ubuntu 16.04 LTS 64bit

YES, I GAME ON UBUNTU.

The Cards

Card Abbreviation	Name	VRAM
1080FE	GTX 1080 Founders Edition	8gb
1060SSC	EVGA 1060 SSC	6gb
1070GX	MSI 1070 Gaming X	8gb
1070XLR	PNY 1070 XLR8	8gb

Open Air Control Comparison

cards	Power	Draw	equihash	lyra2v2	neoscrypt	Day	Mon
	860 PSU
1080FE 1070GX	125w 115w	38 280	482 76° 433 60°	43.87 78° 35.83 60°	919 77° 775 60°	$1.34	$40
	650 PSU
1080FE 1070GX	125w 115w	43 290				$1.39	$41.70

Grouped by computer configuration. kWH charge is $0.20, 30 days per month. Temperature in degrees. Cost in USD.

Hashrates in bold, load power draw in italics.

This clearly shows a $1.70 savings per month. The power savings at idle is ~5 watts, and 10 watts under load. But the results are a little inconclusive, but were created to reduce over-all power cost and not maximize hashrate. Ideally I should have used a 860 bronze PSU for better comparison... but for my purpose this is sufficent; as the power curve on the platinum psu offers a wider range of efficenecy below 50% than the bronze psu.

Conclusion

Platinum PSU's offer little power-savings over an appropriately loaded bronze psu. They seem to offer the most savings/usefulness if you operate them at more than 70% load. You can achieve near-platinum PSU performance by operating a bronze PSU at ~42-48% load. Since I do not require even half of what the 860 watt PSU can output for home computing I have decided to use it in a mining rig with more than 3 GPU's.

Next Step

For this study to be complete I would like to compare operating the maximum number of graphics cards the motherboard can accept, 3 or 4 if using a riser, in addition to using a PCI-X extender card (which adds 3 additional slots), putting the maximum number of cards at 7, or 9 if no cards are plugged into the motherboard and the system is using only pci-x risers.

Other Details

Here are some other numbers from other configurations I ran:

Grouped by PSU/case.

cards	Power	Idle/Load	equihash	lyra2v2	neoscrypt	Day	Month
In Phanteks Eclipse Case	860 PSU
1080FE 1060SSC 1080FE*	135w 110w 155w	38.3 448	485 82° 310 70° 499 81°			$2.15	$64
1080FE 1060SSC 1080FE*	115w 90w 125w	39 375	464 79° 295 65° 486 73°			$1.80	$54
1080FE 1080FE	135w 155w	35 332	485 83° 499 82°			$1.59	$47.70
1080FE 1080FE	135w 115w	36 287	484 83° 475 72°	44 84° 43 74°	932 82° 882 72°	$1.38	$41.40
1080FE 1080FE	115w 115w	34 270	468 77° 477 72°			$1.30	$39
Open Air	650 PSU
1080FE 1060SSC	125w 90w	37 261	481 77° 294 64°	44 24.32		$1.25	$37.50
1080FE	125w	33 162	481 75°	44 77°	920 74°	$0.78	$23.40
1080FE 1070XLR	125w 115w	39 290	480 79° 425 66°	44 82° 35.25 67°	918 81° 825 68°	$1.39	$41.70
1080FE 1070XLR	115w 105w	39 268	467 77° 414 66°	42.34 80° 34.43 66°	864 78° 780 66°	$1.29	$38.70

'*' - powered by PCI-X Riser.

The MSI 1070 Gaming X performs better at equihash and worse at neoscrypt versus the PNY 1070 XLR; but the PNY 1070 runs slightly warmer. That being said my preferred configuration is with the PNY 1070 as the secondary card to the 1080FE, mostly because the MSI 1070 is a behemoth of a card and I mine equihash a little more than neoscrypt.