I lost a PC PSU to a couple of micro-power cuts and (no doubt the associated voltage spikes etc.), so I decided to get a UPS to protect things.
I only want to support the desktop PC, Monitor and my 4 bay NAS.
The next questi "Add the PC PSU rating (850W in this case) to the NAS PSU rating (90W) and the Monitor (78W) to get the max load (1018W) then multiply by 50% (some say 100%) to give some headroom. Some even pointed out that (e.g.) the PC PSU rating was its DC output, so its AC input would be even greater so...
Things seemed to be getting completely out of hand, so I decided to measure, not calculate.
I put those three items (PC, Monitor & NAS) on a separate supply circuit monitored by a power meter set to record max power draw in watts (VA would have been better, but that's what I had available).
Ran things up. Pulled video down from the NAS & ran it on the PC. Ran up a few more apps at the same time - I did pretty much anything I could think of to max out the loading.
Maximum recorded power demand 159W.
The difference between the recommended calculated load and the measured load is so vast that I'm beginning to doubt myself.
I'm certainly not the first person to go down this route so... Comments please.
You have done the right thing. The rated power consumption of an electronic product is based on worst case conditions. That means that every USB connector will be delivering its maximum output power, the power supply will be driving the most power hungry motherboard that it is rated for, every disc drive slot will be filled. Back in the real world, these conditions hardly ever happen all at the same time. Adding a decent safety margin - maybe 50% to 100% - to your measured results will give you a reliable but not unduly over-specified system. Load currents are generally highest at startup, but the UPS will normally be delivering power to a system that is already running when the mains is interrupted, so you don't need to worry about this too much. You mentioned that your measurements are W not VA. Rated power is measured in W, but UPS ratings are generally VA because it is often the output current that is limited rather than the power. Devices with power supplies rated at less than about
75W may not need PF correction, but larger ones will have it. Maybe add 10% to 20% to your measured power to allow for less than perfect PF. John
Even a fairly powerful PC is unlikely to draw more than 250W continuously. Probably less than 100W most of the time. The other things not much - unlikely to be more than 100W all together. (note that inrush can be much higher - so turning kit on while running on battery can glitch the output enough to drop a load or reboot a PC)
So a 650VA UPS would likely be plenty, and leave some capacity to spare. That kind of device is also fairly cheap.
Having said that you often find that as the VA rating goes up, at some point so does the battery capacity. So if you only have 500VA max load, it can still be worth going from the (adequate) 600VA to the (overkill)
1kVA to get the bigger batteries and double the run time.
(I recently wanted one to hold up a micro form factor PC (fairly high end i7 box - running accounting software in 4 VMs), a synology NAS, and various bits of network kit including a 24 port PoE switch).
I was looking at a CyberPower BRICs unit, and the 1kVA/600W would have been plenty. However at half load the unit was only rated for about 6 mins. So going for the 1.2kVA version which has twice the battery size gave plenty of time for the orderly shutdown of the NAS and all the VMs if required. (in reality, according to actual load shown on the the power meter on the UPS, it will likely get 60+ mins of runtime)
Something else to consider when supporting multiple bits of kit is how you will synch the shutdown of the attached loads. Most UPS devices have a USB connection these days that lets you connect a single PC or other device and monitor it. That also allows you to specify a charge remaining threshold when the device should power down. That is easy with just one PC.
However it gets a bit more complicated with multiple devices. There is an open source suite of software called NUT (Network UPS Tools), that allows one device physically talking to the UPS to act as a server and share information about the current power state with multiple clients. I found the Synology box actually uses this for its default UPS monitoring capability (although it does not seem to name it directly). That made it easy to get a NUT client for the main windows machine running the hypervisor, setup the NAS as the NUT server, and point the windows box at that. That the whole lot can be gracefully shutdown in sequence, and then finally switch off the UPS itself as well once the available capacity falls below the desired capacity threshold.
(on restoration of mains, everything is configured to come back up can carry on were it left off)
I have an extension lead made up with exposed individual wires, so that I can loop them through a clamp meter:
formatting link
Makes a handy way to get a feel for the actual size of UPS required.
When you've finished your calculations this is the modern version of the one I have:
formatting link
I've always lived in rural areas with overhead cables so power cuts are frequent. I have it set up so as soon as mains power is lost the PC shuts down, I don't expect it to run the PC in the event of a power cut.
You are nuts to put an 850W PSU in a PC that never exceeds 100W?
What you calculated was the gold standard 'this will *have* to work'.. Not the 'well this will, in practice work'
However there are *VA* ratings. The *peak* inrush current on your power supplies may well be 4A if you switch them all on at once.
In short while the 'heating' power may well be down at 150W, the 'fusing current' might well be up around 4A at mains voltage, or 80 amps from the e.g. 12V car battery.
The question is whether or not your UPS can handle that for a second or two
The simplest advice is to switch to a laptop with external display, keyboard and mouse. I did that a very long time ago and have never lost data for any reason.
For a home setting, in town with reliable electrical supply, and not doing critical work on a computer, there is only a marginal benefit to having a UPS.
In a rural location with overhead supply where short power outages are common, it becomes far more important. Same goes for if you need to WFH often, or have other kit that may not take kindly to having the power dropped on them without warning (NAS box, Hard drive PVR etc)
For business use (like that described above) where trashing a critical database is not high up on the agenda of desirable outcomes" it would be daft to not have one.
If you run a home server, you don't want to risk corruption of the storage array due to power cuts, brownouts or spikes, so a UPS is a very good idea. Also useful to avoid having to restart the server if you need to knock the power off for a few minutes during DIY.
As we are switched over to digital telephony, a UPS to keep the phone working may become a good idea too.
Do you actually have it set to shut the PC down immediately or with a couple of minutes' delay? The latter would allow for times when the power cuts out, but reconnects shortly after.
Definitely overkill - although on some (gaming) machines, little power is used while watching a video, but when playing a game, the video card alone can take 250W ... and some machines have a pair of video cards working together. Unlikely for the sort of machine that might need a UPS, at home, though.
For a PC, I use a copy of Furmark to max out the GPU. And a copy of Prime95 or use 7ZIP compression, to keep the CPU busy.
On this machine, that would be 300W. Without the goading, a "typical busy value" would be 136W, which is railing on one CPU core, without the GPU in usage. That's the difference a "power test setup" makes to the demand result.
On the other machine, it will draw 400W, 180W for the graphics card. Most of the time, it runs 100W-180W if not goaded.
But the power meter is the right approach. Measurement as a means to reflect some degree of reality. The NAS and the display device, they might have more static loads for all I know.
I originally budgeted for one PC on the UPS, but eventually had two on it. You would not expect too many cases where they would go-to-max on their own. The UPS is disconnected at the moment, because the battery is dead, and I did not like the failure mode it displayed.
I doubt I could fit 4TB disk into a laptop. My main server died yesterday. It was my oldest and worst *86 machine, so I popped its disks into my second worst *86 machine and rebooted. Apart from it insisting on using a DHCP network address because the ethernet came up as a different device, it was in fact a straight reboot.
No data was lost.
Unless you have a *very* busy machine, power loss wont corrupt your disks.
UPS is more about having internet access during a power cut, for me.
It is very wise to either keep a PC up and shut it down gracefully, or if you have limited battery, shut it down immediately. And keep it down.
The *only* time I ever lost data during a power cut was when a PC lost power, the power came back on and it stared booting automatically, and then a brownout crashed it again *as it was booting*. The boot disk never recovered.
These days my machines go down and *stay* down. Until I am sure the supply will last long enough to reboot them properly and check their filesystems.
Shutting down an idle machine by pulling the plug on it almost never corrupts its data. But a UPS makes sure anyway.
Indeed. If I were to do all that I'd need to power a router, a server, a PABX and a desktop PC in the office here with a separate UPS for the NTE downstairs as well
That would still not be ideal, but I would have internet. But not a central heating system...
You do NOT want to mess up RAID arrays, by doing dirty shutdowns. Yes, there are kinds of equipment you definitely want a UPS.
I've had people report all sorts of funky failure states for RAIDs. Like the poor individual who had a RAID Mirror of two drives. One drive died, the other drive had apparently "stopped mirroring" three months prior, and he lost three months worth of changes to his data. And people do not set up Mirrors, just for the aggravation when they didn't work. They expect them to work.
RAID arrays still need backups. And a UPS helps prevent them from becoming de-synchronized.
I do not really know what keeps consumer SSDs alive on power failures. It's supposed to be a "firmware scheme", but since they don't have advanced power fail detection (and a Supercap), it's not hardware that protects them against critical data loss. It's hard to imagine the flash has sufficient life for any sort of firmware scheme to make the device "reliable".
The UPS can have a cable, which plugs into the PC, and it tells the PC to do a shutdown. Typically this fires off at the two minute mark (before the UPS has run out of juice). And it allows the PC to do a clean shutdown.
HomeOwnersHub website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.