I no longer trust my CyberPower UPS

[Rajahal]

A while back I purchased a CyberPower 700A 350W UPS because, well, it was cheap ($40 shipped).  I never bothered to hook up the data port to my server, I just used it as a dumb UPS.  At first I tried to hook up my server and my whole TV/HTPC system to it, but it beeped loudly telling me that it was overloaded.  Since then, I ran just my server on it, and it seemed to work fine (I tested it, and it would keep the server running for about 15 minutes during a power outage).  We tend to have more power blips than outages in my area, so I figured it was fine to just keep running it in 'dumb' mode.

 

Later on I bought a good quality APC 1000VA 600W UPS on sale for $87 shipped.  More than twice as much, but a much better deal I've realized.

 

We had a severe lightning storm in my area last night.  The house's power blipped probably 5 or 6 times.  I had my TV/HTPC setup powered by the APC model (since the other isn't powerful enough to handle it), and my server, desktop PC, and networking gear powered by the CyberPower.  Well, every time the power blipped, the server would restart and kick off a parity check.  I eventually realized that I was overloading the pitiful little CyberPower again, but for some reason this time it didn't beep at me to tell me so.  Instead it just let the server die.

 

So basically this thing is only powerful enough to handle one component at a time, which makes it fairly useless.  I'm thinking now of devoting it to running just my router and switches, and buying a third UPS dedicated to the server and desktop.  Lame!

[chanders]

Thanks for looking out Raj!

[Arioch5]

Um... were you running a 20-Hard Drive server with a 520W power supply on a 350W UPS?

 

I'm no professional on UPS but I do own one cyberpower and one APC UPS.  Both work great, don't discount CyberPower just because a 350W unit can't run 400W+.

[madpoet]

Any recomendation for a good UPS for a 650w PSU?

[Arioch5]

To properly size a UPS you really need to know how much power you're actually drawing.  If you don't know that you're just guessing.  There are two stats you really need to look at on the UPS the Wattage and the VoltAmps.

 

Wattage = maximum instantaneous power draw, if you don't use the full Wattage of your power supply you don't need the UPS to surpass it.  However, if you don't know buying a UPS with a larger Wattage than the power supply of the computer will (should) mean you're safe.  There is still a small chance that you have a Power Factor Correction supply which can cause the power drawn to be higher ...

 

Volt Amps = the total stored charge in the UPS, IE how long it will run.  Since you're supply is pulling current at 120V the current draw can tell you how long the supply should last in theory.  Here's an example.

650W supply at full load would pull 650W/120V = 5.417 Amps at 120Volts.  A 1100VA supply would then last for 1100VA/5.417A = 203.06seconds or 3.38 minutes at full load.

 

I think you'll find if you aren't running a lot of hard drives you aren't likely pulling the full 650W.  But Hard drives and optical drives pull a decent amount of power so a full server with a large number of drives can start to add up.

 

Hope that helps.  As for brands, I've been happy with CyperPower and APC so far, haven't tried any others as I haven't needed to.  The lats APC I owned lasted me about 8 years with one (or was it two) battery replacements.  It has recently bit the dust however, always registering an over load condition even with nothing plugged in.  I can't really complain after ~8years of service.

[Rajahal]

Um... were you running a 20-Hard Drive server with a 520W power supply on a 350W UPS?

 

I'm no professional on UPS but I do own one cyberpower and one APC UPS.  Both work great, don't discount CyberPower just because a 350W unit can't run 400W+.

 

My 20 bay server has only 11 drives in it.  I've tested it, it is well under 350W (peak usage is 220W).  Still, my complaint isn't that it isn't more powerful, it is that it didn't warn me when I had overloaded it like it is supposed to.  I consider that a defect.

 

(Atrioch5 said the rest of this as well, but I'll leave it in)

 

Also, don't confuse the watt rating of the PSU with the watt rating of a UPS.  The watt rating of a PSU is the max wattage it can possibly support.  In most cases, you will only be using a fraction of the max.  You need a voltage meter of some sort (I use a Kill-a-watt) to determine what you are actually pulling from the wall.  The watt rating on the UPS tells you the maximum wattage it can support, but if you use the max you are likely to only get a few minutes of battery life out of it.  If you use less than the max, you can get much longer battery life.  Good quality UPSs such as my APC one will tell you all kinds of useful data, such as how much power your devices are using, and how many minutes you can expect the battery to last (although I don't find the latter number to be terribly accurate).

[Arioch5]

My 20 bay server has only 11 drives in it.  I've tested it, it is well under 350W (peak usage is 220W).  Still, my complaint isn't that it isn't more powerful, it is that it didn't warn me when I had overloaded it like it is supposed to.  I consider that a defect.

 

Ah well in that case there is a problem.  Two things to be aware of, if it's a Power Factor correction supply the 220W apparent power could be pulling more due to that.  I'm not familure with the Kill-A-Watt so I don't know how it handles that.  Additionally, at least some my model of CyberPower has a 'silent mode' where it won't beep at you so that you can just log things with the software and not listen to it beep.

 

I do agree though that resetting the hardware with out a beep and at what looks like significantly lower power draw sounds like it's a defective unit.  Since most come with 3-year warranties I'd contact CyberPower about it, bad units do happen they might replace it for ya all together.

[Rajahal]

Thanks for the advice.  The PSU is an Antec NEO ECO 520W, which it appears does have PFC.  I believe the Kill-A-Watt just reports exactly whatever passes through it (it sits between the server and the wall).  I'll do some more testing, and if I determine that my UPS is actually defective I'll definitely get it replaced.  Maybe I inadvertently put it on silent mode or something silly like that.

 

I first started this thread with the title 'I no longer trust CyberPower UPSs'.  I then rethought it and added in that crucial word 'my'.   Sample size is fairly small, after all.

[JackBauer]

I got a good deal on a Cyberpower 1500VA UPS a year or two ago.

 

I paid somewhere around $150 - $180 total.  Can't recall exactly.

 

It is SOOOO much nicer than the 1500VA Belkin it replaced.

 

My belkin died somehow, and actually fried some parts of my MB.  I could no longer use PS/2 keyboard/mouse inputs...  Sometimes it wouldn't boot up properly, etc...

 

The Cyberpower is much better made.  I could end up having the same thing happen some day with it - but just the quality of construction makes me feel better about it.

[bubbaQ]

Volt Amps = the total stored charge in the UPS, IE how long it will run.  Since you're supply is pulling current at 120V the current draw can tell you how long the supply should last in theory.  Here's an example.

650W supply at full load would pull 650W/120V = 5.417 Amps at 120Volts.  A 1100VA supply would then last for 1100VA/5.417A = 203.06seconds or 3.38 minutes at full load.

 

That's very, very wrong.

 

VA = Volts*Amps = Watts/PF where PF is the power factor which is a way of saying how much power you loose due to inefficiency.

 

Watt-Hours is how you measure runtime.  Batteries are rated in Amp-Hours.  Multiply by the battery voltage to get Watt-Hours.

 

For example, the standard RB7 battery assembly for an APC UPS is two 12-volt 18Ah gel cells = 432 Watt-hours.  This can, theoretically, provide 432 Watts of power for 1 hour.  In practice, the best you can expect is about 70% of that, before the voltage drop is unacceptable.  Then you have to figure your power factor losses (30% loss).  So the actually amount of juice to your computer from 432 Watt-hour battery is 212 Watt-hours.  So you can run a 200 Watt load for an hour, or a 400 Watt load for half an hour, etc.

 

The Wattage rating on the UPS is the maximum Wattage the UPS can deliver.

[KYThrill]

A while back I purchased a CyberPower 700A 350W UPS because, well, it was cheap ($40 shipped).  I never bothered to hook up the data port to my server, I just used it as a dumb UPS.  At first I tried to hook up my server and my whole TV/HTPC system to it, but it beeped loudly telling me that it was overloaded.  Since then, I ran just my server on it, and it seemed to work fine (I tested it, and it would keep the server running for about 15 minutes during a power outage).  We tend to have more power blips than outages in my area, so I figured it was fine to just keep running it in 'dumb' mode.

 

Later on I bought a good quality APC 1000VA 600W UPS on sale for $87 shipped.  More than twice as much, but a much better deal I've realized.

 

We had a severe lightning storm in my area last night.  The house's power blipped probably 5 or 6 times.  I had my TV/HTPC setup powered by the APC model (since the other isn't powerful enough to handle it), and my server, desktop PC, and networking gear powered by the CyberPower.  Well, every time the power blipped, the server would restart and kick off a parity check.  I eventually realized that I was overloading the pitiful little CyberPower again, but for some reason this time it didn't beep at me to tell me so.  Instead it just let the server die.

 

So basically this thing is only powerful enough to handle one component at a time, which makes it fairly useless.  I'm thinking now of devoting it to running just my router and switches, and buying a third UPS dedicated to the server and desktop.  Lame!

 

Have you tried hooking the APC to your unRAID and pulling the plug to see if it also shuts off.  Some power supplies with Active PFC will not work on power supplies that use a stepped sine approximation (a square-wave with varying amplitude to simulate a sine wave).  No matter how large the back-up, the power supply will detect a fault and shut the equipment off.

 

So in this case, you plug in XYZ.  Your cyber power beeps to tell you it would be overloaded.  You unplug X and leave YZ.  No beep because it isn't overloaded.  Then you loose power.  You power supply detects a fault and shuts down, causing your computer to reboot.  You say backup is defective because it didn't beep to tell me it is overloaded.  However, it wasn't overloaded and shouldn't have beeped.  It was your power supply that was resetting, not the Cyber Power. 

 

Both the APC and Cyber Power use stepped sine.  So you might experience the same results even plugged into the larger APC.

[Rajahal]

Interesting, I didn't know that.  I have tested the server on both UPSs and both keep the server running while power is lost, so I guess there's no PSU/UPS incompatibility.  If I have only the server plugged into either the CyberPower or the APC, both work as expected.  It was just this one 'overloading' scenario that has put the CyberPower on my sketch list.

[bubbaQ]

Realize that simply "unpluging" a UPS to see if the equipment runs OK is not a 100% valid test.... PARTICULARLY with a UPS that uses a transformer for step-up "brownout" protection since the transformer is storing energy.

 

When the power grid fails, all the other devices in your home that are on cause the voltage on the hot lead to drop to zero FAST, as they will act as a short to other equipment.  In fact, power can flow backwards from the transformer in the UPS back into the mains in that situation.  That does not happen when you simply unplug the UPS.

 

Don't get me wrong,  I do an "unplug" test all the time as a quick check, and to do rundown testing.  But for critical testing, I have an outlet strip with four 150 Watt light bulbs, and I plug that into the wall, and the UPS into the strip with the four lights, and with the lights on and the outlet strip on a switched outlet, kill the power.  This more closely approximates the environment of a real power failure.

[PhilH]

Realize that simply "unpluging" a UPS to see if the equipment runs OK is not a 100% valid test.... PARTICULARLY with a UPS that uses a transformer for step-up "brownout" protection since the transformer is storing energy.

 

When the power grid fails, all the other devices in your home that are on cause the voltage on the hot lead to drop to zero FAST, as they will act as a short to other equipment.  In fact, power can flow backwards from the transformer in the UPS back into the mains in that situation.  That does not happen when you simply unplug the UPS.

 

Don't get me wrong,  I do an "unplug" test all the time as a quick check, and to do rundown testing.  But for critical testing, I have an outlet strip with four 150 Watt light bulbs, and I plug that into the wall, and the UPS into the strip with the four lights, and with the lights on and the outlet strip on a switched outlet, kill the power.  This more closely approximates the environment of a real power failure.

 

Never thought about that with the light bulbs. It would definitely test the UPS under load during a power failure.

 

Thanks for the tip.