Supermicro PSU efficiency
- Thread starter Fritz
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
You may be able to extrapolate a decent estimate from their 80 Plus reports:
https://www.supermicro.com/products/powersupply/80PLUS/80PLUS_PWS-1K28P-SQ.pdf
https://www.supermicro.com/products/powersupply/80PLUS/80PLUS_PWS-741P-1R.pdf
Gut feeling, I'd expect that by using the larger PSU your system's load would move further from the efficiency sweet-spot.
https://www.supermicro.com/products/powersupply/80PLUS/80PLUS_PWS-1K28P-SQ.pdf
https://www.supermicro.com/products/powersupply/80PLUS/80PLUS_PWS-741P-1R.pdf
Gut feeling, I'd expect that by using the larger PSU your system's load would move further from the efficiency sweet-spot.
In my experience, no. I've actually tested similar levels of loads with exactly both models of power supply, and the 1280 uses a few watts more.
Both are 80+ platinum, so same efficiency requirements. The 740 watt you're running at 20% load. The 1280 watt you'd be running at 10% load. These things are simply not very efficient at 10% load -- and they're not required to be. At 20% they need to meet a minimum standard at least. If your PMBus numbers are accurate you're already hitting 91% efficiency which is pretty good.
In a household, the watts use will be a little more on the 1280, another 1-3%. In a datacenter, you may also be hit with poor power factor -- using more amps per watt than you would with a better power factor -- adds another 1-3% to the operating cost.
no, Unfortunately haven't found Time to cleanup/bring everythig in Shape. pmbus and measured at the wall on some psu is quite a difference, if i remember right the 721 are not as bad as reported by pmbus, but still not great on low load.
Efficiency levels of the SQ and non SQ Platinum PSUs are fairly similar. For lower utilization the smaller PSU's are almost always better since you get to the most efficient range easier.
So In your case (175W) the most efficient PSU will be 500W Platinum, then the 741 then the SQ
So In your case (175W) the most efficient PSU will be 500W Platinum, then the 741 then the SQ
I haven't tried the 500w platinum, so I can't say for sure. Looking at the specs, it doesn't look very impressive. The efficiency may be within bounds for the 80+ P rating, but the power factor is atrocious by comparison. In a datacenter this is very relevant where you are billed by the amp, less-so for home use where you are billed by the watt.
are the amps of the reactive power included in the accounting?
I fortunately have accounting by metered watts, so didn't care about when deciding on psu's
I'm not sure about accounting. The power supply test reports show amps, watts input, watts output, efficiency %, and power factor, at 10%, 20%, 50%, and 100% load. The 740 watt seemed like the sweet spot for low power use situations. The 10% load figures on the 740 were about as good as the 20% figures on the 500w, and the 20% figures on the 740w looked better to me than the 50% figures on the 500w. Again, the biggest difference was power factor and this is less relevant when you care about watts and not amps.
In many datacenters, you pay for a circuit based on its size (amps and volts), not based on how much you use. A power factor of 0.99 therefore gets you roughly 10% more watts per dollar than a power factor of 0.90.
In some ways this makes sense. UPS and generator capacity is often limited by volt-amps and not watts, and that equipment is a significant expense for the datacenter. In other cases, it's just like any other "unmetered" service -- the real profit is made when customers use less than they expected to.
yes, i know that UPS etc need to be specced for the amps/apparent power, wether it is used in total or not. Cooling would be different, as only power that is actually converted to heat matters.
I've made measurements on the 501 and the 741 with a True-RMS power meter, can't access the results atm but found the 741 not so appealing. but will put the numbers in here soon. Unfortunately i didn't write down the apparent power but could have a quick look at this for those two psu, too. (still have all psu and that exact box i used as consumer)
if i got it right PF / Power Factor is what is labeled with PF in the specs PDF, and that is the ratio of apparent power and effective power ?
(Sorry, need to get the translations from the net, so maybe terms are not exactly right)
Our colo rack is not accounted by amps but actual usage of kWh, what for me makes more sense and also gives a motivation for optimizing psu. so i put most interest for measurements and comparsion on the actual draw
I've made measurements on the 501 and the 741 with a True-RMS power meter, can't access the results atm but found the 741 not so appealing. but will put the numbers in here soon. Unfortunately i didn't write down the apparent power but could have a quick look at this for those two psu, too. (still have all psu and that exact box i used as consumer)
if i got it right PF / Power Factor is what is labeled with PF in the specs PDF, and that is the ratio of apparent power and effective power ?
(Sorry, need to get the translations from the net, so maybe terms are not exactly right)
Our colo rack is not accounted by amps but actual usage of kWh, what for me makes more sense and also gives a motivation for optimizing psu. so i put most interest for measurements and comparsion on the actual draw
A Xeon D 1518 system idle power using a 920SQ was around 80W. It went down to 36W when I installed it in a Cse-505-203B chassis with a 200W PSU.
If you are willing to give up the hotswap capability, you probably can use the PWS-341P-1H which is the smallest platinum rated PSU from SuperMicro with PMBus.
I'm Currently building out 4 nodes that will have a peak power consumption of around 220W. They idle at around 90W. I Havent done the math to see if the cost of swapping out the PSUs will be worth it for them. I have been watching but havent seen any used one for a good price.
If you are willing to give up the hotswap capability, you probably can use the PWS-341P-1H which is the smallest platinum rated PSU from SuperMicro with PMBus.
I'm Currently building out 4 nodes that will have a peak power consumption of around 220W. They idle at around 90W. I Havent done the math to see if the cost of swapping out the PSUs will be worth it for them. I have been watching but havent seen any used one for a good price.
Maybe in terms of efficiency you are better off with the 200W (i guess it's a gold) PSU, especially when the nodes are idle a lot.
With 36W Idle you would trade running at ca. 18% on a gold PSU vs little more than 10% on a Platinum.