My last uninterruptible power supply (UPS) was dead and buried. I had gotten by without one for a while. But recent, recurrent power outages, crashing my desktop computer system and taking my work down the tubes, had convinced me that it was time to get another.
This post describes how I decided which UPS to buy, some key points about its software, and how it worked out in testing. As discussed in another post, this was part of a larger inquiry into longer-lasting alternate sources of computing power.
Estimating Power Usage
It seemed that most UPS vendors included software that could send a hibernation or shutdown signal to Windows when the power went down. (See SuperUser for information on Linux.) For example, the User Guide for APC PowerChute 3.1 explained that, after a specified number of minutes, PowerChute would initiate hibernation or shutdown in response to a prolonged power problem. How-To Geek (Burgess, 2019) illustrated that PowerChute would let the user set the number of minutes that the computer would continue to run before automatic shutdown or hibernation commenced. Of course, I could tell Windows to go ahead with a sooner hibernation or shutdown, if I saw no reason to keep using the UPS battery’s reserves.
In the UPS world, time is money. If I wanted a UPS that would keep the desktop running forever, I would have to pay for it. The budget solution was to choose a UPS just big enough to let me decide whether this seemed to be more than a momentary outage. I wouldn’t want to go through the hassle of shutting down and starting right back up again if that wasn’t necessary. A Spiceworks discussion offered a variety of thoughts on how long I should wait before shutdown. I decided that a five-minute wait-and-see period should be long enough to see whether it seemed likely that the power outage would continue for a while.
So the question was, how much UPS power did I need, to keep my desktop running for, say, seven minutes (i.e., five to wait and see; two to shut down)? I might eventually decide to allow a minute or two above or below, but that seemed like a reasonable starting target.
To estimate my power consumption, I ran a search and looked at several different UPC power calculators. Some required me to state a number of watts; some asked what kind of devices I would be connecting; some went both ways. I will describe how I worked through that process — but I will also mention that, for me, the more accurate solution was to buy a Kill-A-Watt device (below), and let that tell me how much power I was really using. But maybe that solution won’t work in every situation.
The calculators that asked what devices I was connecting would probably be most convenient for those users who had a stock system — a computer bought straight off the rack and not significantly modified. If the user’s manual didn’t specify a power usage figure, several of these websites could help. For instance, TrippLite listed a large number of desktop PCs, with specific model numbers, to identify the wattage required, and likewise for monitors and a few other categories of devices; Kohler’s kVA Load Calculator asked “Do you know your load and load characteristics?” No > “Is it IT or Non IT?” IT > Typical workstation volt-amps (VA) = 200, etc.; and APC’s UPS Selector > Home Office > Configure by Device > PC/Workstation > Desktop Computer estimated a power draw of 100W. But even these few examples varied enough to suggest that this was not going to be an exact science.
I had built my system from scratch, in the sense of ordering motherboard, CPU, RAM, and other components according to the best combination of power, features, and price. So I couldn’t use a simple choose-your-system power calculator. For me, it seemed that, other than inspecting product labels and literature, the most accurate and informative approach would be to use a power supply calculator. MakeUseOf (Enticknap, 2021) said the five best calculators were from OuterVision, Cooler Master, Be Quiet, Power Supply Calculator, and PCPartPicker. To this list, I added Newegg. Among these, the Cooler Master calculator did not seem to be working. The others allowed me to specify (to varying degrees) the motherboard, CPU, Thermal Design Power (TDP: I went with their 90% recommendation), RAM (including FB DIMM, which was apparently not used in anything newer than DDR2), video card (if any), and other devices.
Those details about system hardware were available from system information tools (e.g., Speccy), supplemented if necessary by more specialized tools (e.g., MiniTool Partition Wizard), Google searches, and my old receipts. Their conclusions: I needed a maximum of 302W (OuterVision), 176W (Be Quiet), 235W (Power Supply Calculator), 261W (PCPartPicker), or 263W (Newegg). MakeUseOf (Enticknap, 2021) said that these calculators all seemed to be suggesting the bare minimum, and that caution called for adding “a couple of hundred watts on top.” But I wasn’t a gamer. It was unlikely that I would simultaneously be stressing the CPU and using all of the disk drives and other devices they asked about. One of these calculators said my idle power needs were only 98W. (I think the site that said that was Power Supply Calculator. Unfortunately, their webpage was not responsive when I went back later to verify.) I was using a 500W power supply, and had not had any power-supply-related issues, so 500W seemed to be a fairly solid ceiling. I’d had an even smaller UPS previously, and it had actually been useful. So I doubted I would actually need more than 400W now.
It seemed that printers should probably not be included in the system components. I saw various warnings against trying to run a printer from a UPS. I think the problem was that printers have motors that may briefly demand more power than an ordinary consumer UPS can provide. If the printer had to be available during a power outage, it seemed that finding a UPS for it could require some exploration of factors that I didn’t get into here. Note also the option of buying more than one UPS, to supplement a UPS that was too small, had too few outlets, or otherwise did not suffice for some systems or components.
Another adjustment: in these calculations, I forgot about my laptop. I always had my laptop plugged in when I was at my desk. Over time, keeping it plugged in had killed the battery: it was an Acer, and apparently Acer didn’t build in the charging shut-off software or circuitry that I had enjoyed with my old Lenovo. (I think ASUS also offered such software.)
At this point, without a working battery, the Acer laptop was effectively functioning as a small desktop computer. In the case of power outage, it would need the UPS to continue to run. Ideally, I would shut the desktop down sooner rather than later, and use the remaining battery power in the UPS to power the laptop. (The UPS offered only one USB port for its automatic shutdown system, and that cable would run to the desktop, so presumably the laptop would not be shut down automatically in the event of a power outage.) I believed that the laptop would not consume much power, so forgetting to include it in my power estimates (above) was probably not major.
For those who had learned, as I had, that watts can be calculated as amps times volts, some websites offered a correction. According to Schneider Electric (2015), watts were “the real power drawn by the equipment” while volt-amps (VA) was the “apparent power.” The W:VA ratio was the “Power Factor,” expressed as a number (e.g., 0.8) or percentage (e.g., 80%). For instance, APCGuard offered an APC Smart-UPS Runtime Chart displaying these W:VA values, among others: 200:280, 300:420, 400:560, 500:700, 600:840, 700:980. In each case, the Power Factor was 71%. N1C seemed to say that this was due to inefficiency in the UPC’s circuits: watts was supposed to equal amps times volts (i.e., volt-amps), but power loss was apparently unavoidable, to such an extent that (as that Power Factor example suggests) you might actually get only 70% of the expected power.
In other words, I should rely on the watts figure, not on the VA figure. Which I was going to do anyway. But if a UPS didn’t specify its watts output, apparently I would want to reduce its stated VA capacity by some such percentage to get its probable wattage capacity.
Choosing a UPS
With what I had learned about likely power needs, it seemed that now I could start to choose among specific UPS models:
- APC’s UPS Selector said that, with a load of 500W, a runtime of seven minutes, and a power expansion of zero (i.e., their fudge factor, allowing for future expansion to use more power-hungry devices, which I didn’t think I would be doing, at least not with a 500W ceiling), they could recommend a number of their power supplies, starting at only $370. That was the APC Smart-UPS 1000VA, 600W tower. Yikes! If I backed off to a 400W load, they were willing to cut me a break with their Back UPS PRO BR 1350VA for only $250.
- Moving toward more realistic prices, APCGuard’s chart said that I could get seven minutes at 400W from several models (i.e., SMT750, SUA750RM2U, SMT750R2-NMC, and SUA750, all of which presumably used the same 750VA battery) and seven minutes at 500W from the SUA1000RM1U. Searches at Amazon suggested that those models were all discontinued. The closest I got was the SMT750C, for $191. But at least it made sense that I would be looking for models with 750 in their names, short for 750VA, which apparently tended to produce somewhat above 500W.
- Eaton said that, for 7 minutes at 400W, I could start at the Eaton 5S 1000 LCD (600W) for $160. At 500W, it would be at least the 5S 1500 LCD (900W) for $215.
- Even with multiple filters, Tripp Lite gave me a plethora of models, with no obvious logic and no stated prices to facilitate comparison.
That effort was not too enlightening. I decided to switch to merchants’ filtering tools. Newegg’s power search was my favorite, so I started there. Those tools brought other UPS features into the mix. First, among the available features that Newegg offered for my consideration, it seemed that the number of outlets could be important because, according to Wirecutter (Witman, 2021),
Do not ever—ever, ever—plug a UPS into a surge protector or plug a surge protector into a UPS. Aside from potentially overloading either unit and tripping a fuse or breaker, you also risk canceling out the surge-protection function instead of doubling it up. Tim Derochie at CyberPower told us that UL, the independent company that evaluates electronics safety, strictly forbids the practice, and you’ll find strong warnings against doing so in owner manuals.
Similarly, you should not plug a UPS into an extension cord, because the excess load can cause it to overheat and melt. If the cord on your UPS isn’t long enough, you’re better off moving your workstation closer to an outlet or spending $100 or so to have an electrician install a new one—either of which is preferable to replacing damaged electronics or accidentally starting a fire (PDF).
The provided link was not very persuasive: it issued standard warnings against abuse of extension cords. As the linked document said, “Choose the right extension cord for the job, and use it as specified.” Witman seemed to be overstating the reality on that one. I didn’t need an extension cord this time around. I had used one in the past; I just made sure it had adequate capacity.
But the case against surge protectors was more convincing. To cite another source, “Schneider Electric recommends against the use of any surge protector, power strip or extension cord being plugged into the output of any APC Back-UPS and Smart-UPS products.” That source reiterated that the circuitry in surge protectors could interfere with or confuse the battery backup, and vice versa. In other words, let the UPS be your surge protector, for the devices plugged into it.
It seemed, then, that (as per consensus in e.g., a Reddit discussion) a power strip without surge protection could safely be plugged into a battery backup (though doing so might still invalidate your warranty). This mattered because Newegg’s power search posed a question of whether the UPS needed to have a certain number of outlets. The answer seemed to be that no, it didn’t.
Turning to other criteria, Newegg’s search disappointed me in not allowing for any subcategories within the 1-500 watt group. Within the 400-500W range, it yielded 35 UPSs. I was still not too sure what I was looking for, so I went on to an Amazon search. Unfortunately, Amazon’s filters were even less helpful. I had to conduct individual searches by each major manufacturer. Moreover, Amazon mostly reported VA rather than watts, so I had to manually inspect many items that turned out not to produce at least 400W.
I narrowed down the Amazon listings by comparing them against recent recommendations within my range from Thomas, Wirecutter, PC Gamer, Bestcovery, and Lifewire. I also refined my understanding of various UPS features:
- Some recommended models offered automatic voltage regulation (AVR). Wirecutter described that as “an important feature.” According to Lifewire, it “automatically corrects power fluctuations to protect the connected equipment against hazards like voltage spikes and brownouts.” CyberPower seemed consistent with these reviewers when it said that AVR was a feature of “line interactive UPS systems” that “significantly increases battery life and lessens the likelihood of data loss, memory freezes, and system crashes” when compared to the less expensive “standby” design.
- Another feature sometimes mentioned, but not found at this price level: pure as distinct from stepped sine wave. My reading led to the conclusion that my devices were not sensitive enough to require pure sine wave power.
- I saw that some units had LCD displays. I’d had one of those myself, on a UPS several years back. It was mildly informative. But it didn’t provide nearly as much information as the monitoring software. For instance, the User Guide for APC PowerChute 3.1 said that the APC software would display the things I would really care about (e.g., undervoltage, remaining battery charge, estimated time until shutdown) as well as other potentially interesting things (e.g., energy usage rate). An LCD display might be worth spending a few extra bucks on, if that was an option, but it wasn’t important to me.
- CyberPower offered Eco Mode. Apparently this feature would shut off power to peripherals when it detected that the computer was turned off or in sleep mode. I generally left my computer on, and anyway didn’t tend to plug power-hungry peripherals into the UPS, preferring to save the battery for computer power.
- Tripp Lite persuaded me that lithium-ion batteries were superior to the old sealed lead-acid batteries (with, among other things, faster recharging speed), but it didn’t look like they were going to fall within my budget.
AVR looked like the only must-have feature from that list. On that basis, it seemed that the reviewers’ favorite units, for my purposes — the ones to compare all others against — were CyberPower’s CP800AVR 450W ($112, 3-year warranty) and CP850AVRLCD 510W ($120, 3-year), and Tripp Lite’s AVR900U 480W ($98, 3-year), with honorable mention to the Amazon Basics 1000VA 550W ($100, 1-year). The Tripp Lite unit, in particular, suggested that I should be able to get a 480W AVR unit with a three-year warranty for under $100. It did not appear that the reviewers were relying on any special testing or insight beyond the units’ stated features.
Those selections left out APC, the other big UPS manufacturer. Apparently that was due to price. A Reddit discussion conveyed the sense that APC had a superior reputation in critical and server settings, while CyberPower was more commonly used on desktops. Not everyone in that discussion felt that APC’s quality was as good as it used to be. Some participants voted for Eaton over APC, but Eaton didn’t seem to have many competitive offerings at my level.
Against those recommended UPS units, I boiled down the many alternatives on Amazon to a few name-brand favorites:
- It would cost an extra $17 for a mere 30W, to step up from the CyberPower AVRG750U 450W ($83) to the AVRG900U 480W ($100). I decided that wasn’t worth it.
- AVR cost an extra $20 in CyberPower’s CP825AVRLCD 450W ($103) compared to the CP825LCD 450W ($83). Given the many power fluctuations we had been experiencing, I decided it was worth spending extra for the AVR: my system had held up, but I couldn’t know whether the fluctuations were taking a hidden toll.
- The CyberPower AVRG750U 450W ($83) would give me 12 outlets, while the CP825AVRLCD 450W ($103) offered only 8. In both cases, half were battery backup, half were surge only. The latter seemed superior only in offering an LCD display. That wasn’t worth $20 to me.
- Having just encountered an Amazon user who said that a 450W UPS wouldn’t keep a desktop computer going very long, I was almost spooked into abandoning my calculations and fleeing to the safety of the CyberPower CP900AVR 560W ($114), for a $31 premium over the AVRG750U 450W ($83).
- The closest I came to being tempted into buying an APC unit was the BR700G 420W ($117).
- The Tripp Lite AVR750UTAA 450W ($121) was not competitive against the CyberPower AVRG750U 450W ($83). The Tripp Lite VS900T 480W ($95) came closer, but still no cigar.
For the reasons just stated, none of the reviewers’ choices stood up against the CyberPower AVRG750U 450W ($83) — which, incidentally, did boast GreenPower and Energy Star credentials. I went back to compare that unit against the offerings at Newegg. That unit was the same price there but, as at Amazon, nothing else compared. So I ordered a CyberPower AVRG750U.
Problems with UPS Software
When the CyberPower unit arrived, I noticed certain instructions: charge the battery for at least eight hours before using; keep the power load below 80% of UPS capacity; recharge every three months while in storage. At their website, I saw that a two-year extended warranty could be had for as low as $39; I registered the unit and verified registration; and I viewed the available software, including specifically their PowerPanel Personal and Business 4 software, utilities, and firmware updates. From these, it appeared that all I needed or wanted was PowerPanel Personal (available for Windows, Mac, and Linux).
The Windows version of PowerPanel Personal could reportedly provide current information, send me emails to notify me of power conditions, and manage shutdowns. After installation, I ran it. It oddly started in the Windows system tray rather than the taskbar; I had to hunt around to find it. The User’s Manual didn’t say anything about connecting a USB cable in its Getting Started section. I had a cable connected nonetheless; but the software’s home screen didn’t seem to know anything. The User’s Manual seemed to say the problem was not the cable; it was that “[t]he PowerPanel Personal Service is not running.” Whose fault was that? A search in the manual did not produce any hits. I didn’t see the Personal Service listed in services.msc. A Google search led to no answers, nor did searches of their FAQs or resources. I clicked on their website’s Chat Now button; it was unresponsive. The webpage said tech support was open 24 hours on weekdays only, so I submitted a ticket in their Support area.
Eventually I noticed that my search of the manual had been defective. Its brief Troubleshooting section did say that, if the PowerPanel Personal Service was not running, I should run C:\Program Files (x86)\CyberPower PowerPanel Personal\startService.bat. But that produced an error: “The specified service does not exist as an installed service.” Continuing with another suggestion in that Troubleshooting section, I uninstalled and reinstalled the software. But no, that wasn’t the solution: it was still “not ready.”
CyberPower’s tech support replied to my ticket rather promptly, with a PDF containing additional troubleshooting tips. It said I should take a look at Device Manager. Contrary to what they suggested, in devmgmt.msc I did not see CyberPower Battery Backup as an item under the Batteries heading, and in services.msc I still did not see either PowerPanel Personal Service or PowerPanel Personal Service Monitor. The PDF said that antivirus software could interfere with installation, but my Bitdefender installation would normally give me a pop-up notice in the event of a problem, and that hadn’t happened. Uninstalling and reinstalling with antivirus turned off did not fix the problem.
So I replied to CyberPower tech support with my new situation. This time around, they did not reply promptly. In fact, they went dark. They lost power. You know what I’m saying? They unplugged; their batteries died.
Since CyberPower was compelling me to provide my own tech support, I cast about for solutions. I decided to uninstall the Personal version and download the PowerPanel Business 4 software and its User’s Manual. I installed it only for Local (not Remote) use. After installation, I ran it. It worked immediately. Apparently there was something about my system that the PowerPanel Personal software was not developed well enough to handle.
UPS Software Setup
When I ran the PowerPanel Business 4 (PPB) software, it opened an HTML page in my Firefox web browser, with a URL address of localhost:3052/local/. That page required a login. A popup informed me that the default username and password for the PPB software were both simply admin.
The menu across the top of that HTML page offered choices: Dashboard, UPS Setting, Setting, Reporting, Help. The Dashboard led off with a status indication (i.e., “The UPS is working normally”), an update request (i.e., “Please set battery replacement date”), and indications that voltage was Normal at 124V; output was 124V at a load level that seemed to fluctuate between 15% and 25%; and the battery was fully charged, with around 40 minutes of runtime. The manual (p. 23) said,
Battery Replacement Date is the date that batteries have been replaced. This date should be changed immediately after the battery replacement or when the unit is first purchased. The battery lifetime varies by UPS models. Once battery replacement date is changed, the software will alert the customer when the battery age has reached the lifetime.
But now, in the first of several such instances, I could not figure out where the desired item was located. They wanted me to set the Battery Replacement Date — but where could I do that? The manual didn’t say, and the Dashboard didn’t provide a link. I had to go through the software manually to search for it. I found it in UPS Setting > Configuration (below).
The Dashboard also provided a log of Recent Events. The first one was, “Communication with the device has resumed,” as of a few minutes earlier. I decided to use that log for guidance. I began by observing a remark in tech support’s additional PDF (above): “We recommend using CyberPower supplied US [I think they meant USB] or Serial cable, whenever possible.” But now I saw that the log said, “Communication with the device has resumed” when I replaced their stock cable with a red USB 2.0 cable that I preferred. When unplugging and plugging things, the red cable was easier to trace, among the various cables under the desk. So I decided to use the red cable instead of theirs. I don’t think I encountered any situations where the cable affected functionality.
Their PDF also warned me, “Do not use a USB Hub or any other intermediate connection. The supplied cable must be plugged directly from the UPS to the port on the computer.” But that was stronger than their FAQs page, which did contemplate the use of a hub. Here, again, the log said communication resumed when I plugged the USB cable into a USB hub. I could understand that a USB hub would be problematic during a power outage, if the USB hub was powered by an AC adapter. This particular hub was powered solely by the USB connection. I doubted the UPS was so sensitive that, unlike my other devices, it was unable to function reliably through a hub. Nonetheless, I did have a spare USB 2.0 port on the computer, so I used that as advised.
Around this time, I let the webpage sit for a while, in the Firefox browser, and apparently it logged me out. When I tried to log back in, it was unresponsive for a moment, and then said, “Error: could not authenticate you.” I tried killing the webpage and restarting it from the PBB software’s Start Menu icon. But no, now Firefox was saying, “Unable to connect.” The problem didn’t seem to be with the browser: I was able to view news at CNN.com. In Chrome, the message was, “localhost refused to connect.” In services.msc, I saw that PowerPanel Business Service was not listed as Running, so I right-clicked > Start. Now the webpage was willing to run in Firefox. I wasn’t sure what caused all that.
In the PPB software, I went to the Dashboard’s next menu pick at top: UPS Setting. This one led to Diagnostic, Configuration, Connected Equipment, Scheduled Shutdown, and EnergyWise options. Diagnostic led to a Battery Test, which was successful, and a Runtime Estimation, which promised to discharge the battery “to near zero capacity with the current load.” I postponed running that test, pending completion of this tour through the software.
Within UPS Setting > Configuration, there were options to set power failure conditions (i.e., upper and lower output voltage limits (default: 132 and 100 VAC) and detecting sensitivity (default: medium). There were also options to disable the Utility Power Failure Alarm, and to revise the default low battery thresholds.
That was also where I found the option sought above: Battery: Set Replacement Date. I entered the date on which I had first plugged in the UPS. But that didn’t fill in the Replacement Date field, as expected; instead, it filled in the Last Replace field. The Replacement Date field remained blank. They should have labeled it, “Set Last Replace Date.” But then I saw that I was supposed to click the Apply button to make it stick. When I did that, the software generated an estimated Replacement Date three years after the Last Replace date.
It would have helped if CyberPower had made sure that terms appearing in the software were described in the manual. Unfortunately, a search for the PPB software’s term, “Low Battery Threshold,” turned up nothing in the manual. Then I saw why not: believe it or not, the manual had not been run through OCR. The term was there, on page 36: I just couldn’t find it with a search. I had to use Adobe Acrobat to OCR the manual before I could search for stuff in it.
For Low Battery Threshold, the manual (p. 36) said that the UPS would sound an alarm when its capacity or remaining runtime dropped below this level. The 300-second value was the lowest of those offered, in a range up to 600. There didn’t seem to be a way to turn it off. So I was going to get an alarm when the UPS had only five minutes’ worth of juice left, and also when it dropped to 10%. I left those values at their defaults for now.
Under UPS Setting > Connected Equipment, there was a list of the UPS’s 12 power outlets — six labeled BATTERY, six labeled SURGE. It looked like I could edit that list to indicate specifically which devices (e.g., computer, monitor) were plugged into which outlet. I didn’t have any plans to use UPS Setting > Scheduled Shutdown, which was blank by default. The manual (p. 55) said that UPS Setting > EnergyWise was intended for use with Cisco EnergyWise-controllable devices, which I didn’t have.
Back at the top level of the menu, the next main item was Setting, leading to options for Shutdown Events, Notifications, Notification Channels, Recipients, SNMP, Security, Advanced, and Network Configurations. Here, again, not all of these were relevant for an individual user in my situation.
In Setting > Shutdown Events, PPB allowed me to set Required Shutdown Time. The manual (p. 49) said, “This setting will set the amount of time which Local and Remote computers will take to shut down. … This setting should be configured to allow for the normal shutdown time of the computer.” Possible values ranged from 1 to 30 minutes. PPB also offered a choice between the default Shutdown and the alternative Hibernate (visible only on systems that supported hibernation). On my system, one minute would probably suffice for shutdown, but I could see where hibernation might require more like the default two minutes, so I left it at that.
The Shutdown Events screen also gave me the option of setting Shutdown Delays. For example, the first item on the list, “Battery capacity is critically low,” was set by default for Instant shutdown. Most of the items on the list were set to Inactive. (Other delay choices ranged from 5 seconds to 60 minutes.) The manual (p. 70) said that one interesting Inactive item, “UPS fatal internal abnormality,” meant that the UPS could be set to shut down (instantly or after the desired delay) upon a variety of problems inside the UPS (e.g., output overload or short circuit; reversed battery connection).
In Setting > Notifications, the situation was similar, though the only options were Instant or Inactive: PPB was set by default to send instant (and, optionally, repeat) notices upon the occurrence of the items already listed in Shutdown Events, and a few others as well (e.g., battery replacement recommended). Setting > Notification Channels allowed me to specify whether these notices would be sent by email, SMS, and/or XMPP Instant Messenger. Setting > Recipients was self-explanatory. Setting > Advanced allowed me to specify commands that would run in response to various energy events.
So, for my purposes, that pretty much exhausted the options in PPB. But had I missed something? I didn’t really mind the admin username and password, but I would prefer not have a password; yet I had not seen a place to change that. I soon saw why not: the manual said I could change it in Setting > Security. PPB’s Help option likewise seemed to say that there should have been a number of topics within that Security page, but mine was missing Login, Account, and others.
A partial solution was to click on the person icon at the top right corner of the screen, next to the flag, and choose Change Password. In the resulting dialog, I left Session Timeout at Never, entered admin as the current password, and left the new password blank. I clicked to finish that. I was promptly logged out, and PPB refused to log me back in with a blank password. It required the admin password. This suggested that a blank password was unacceptable. I emailed CyberPower tech support to inquire about these issues.
Testing the UPS
My research and the UPS software left me, at this point, with perhaps two questions: how long would the battery keep my computer going, and how would the PPB software behave (e.g., what messages would it send me, what values would it display), during a power outage?
As noted above, at UPS Setting > Diagnostic, I had already run a successful Battery Test, and I had considered running a Runtime Estimation. Then I realized that it was going to take the UPS some hours to recharge its battery, after I ran it down in a test. After all, they had asked users to charge the unit for eight hours before first putting it into service. So I wasn’t going to be able to run multiple tests, not unless I wanted to drag out this project for another day or more.
I decided to run just one test, but to make it as real as possible. So there wouldn’t be a comparison of multiple tests of UPS performance with light vs. heavy loads, and with different shutdown settings and trigger levels. In place of the Runtime Estimation test, I figured I would emulate an actual power outage by unplugging the UPS and watching what happened.
I could nonetheless get some insight into light vs. heavy loads by using a different source. In the process of researching these matters, I had discovered and ordered a Kill A Watt P3 P4400 Electricity Usage Monitor. I chose this over the Kuman KW47-US after a search for the word “accurate” in user questions and reviews: there seemed to be pretty consistent agreement that the Kill A Watt was accurate (which was what Consumer Reports also concluded, back in 2009) and substantial indications that the Kuman was not. While I was reading reviews, I noticed that people tested against various items for which they had wattage ratings (e.g., light bulbs). I also noticed, incidentally, that some of these power testers could be fried by plugging them into a modified sine wave inverter or generator, or by leaving them connected to heavy power users (e.g., refrigerator) too long
I powered down the desktop and laptop computers and unplugged the UPS from the wall outlet. I plugged the Kill A Watt (KAW) unit into the wall outlet, and plugged the UPS into the KAW. The UPS was unplugged for only a few seconds; its recharging needs probably did not significantly affect subsequent readings. (The KAW was hard to read in low light, which may be the actual usage circumstances for many users. Shining a light onto or near its old-style LED display barely helped me to figure out what it was saying.)
With everything except the laptop computer plugged into the UPS but turned off, the KAW said it was drawing about 8.7W (i.e., watts). (This and other readings fluctuated, but mostly less than 1-2%. Where there was fluctuation, the values reported here are my impression of the approximate average.)
When I turned on the computer, the KAW jumped up to 129W with no software running other than the basic system and miscellaneous items (e.g., Bitdefender antivirus) that were always running in the background. But that was apparently just startup load: 15 minutes after startup, the KAW said the idle system was using only 63W. This was for a desktop Windows 10 system with an Acer G226HQL 21.5″ LED monitor, Intel Core i7 4790 CPU, 24GB DDR3 RAM, onboard graphics, and a Samsung 850 EVO M.2 500GB solid state drive (SSD).
After another 15 minutes, during which I had three browsers running ~100 open tabs, power consumption had only gone up to 66W. So far, I had connected only the Samsung SSD, a PNY SSD, and a WD internal hard disk drive (HDD), and was not overtly using any of them. Now I turned on five external USB HDDs — one WD Passport, one WD MyBook, and three other 3-4TB WD internal drives, each in its own external dock, each getting its power from the UPS. After ten minutes of idling, the KAW said the system was using 86W. This suggested that plugging in an external HDD might be worth 4-5W.
Next, I powered up the laptop, running something like Windows To Go (WTG) from a 128GB Samsung USB flash drive. The laptop was an Acer Aspire 5 A515-51-563W 15.6″ with a Core i5-7200U CPU, 20GB RAM, two internal SSDs, and a Samsung 850 EVO M.2 SSD. The laptop’s battery was worn out and could not hold a charge; it was running entirely on power supplied from the wall outlet through the USB.
Two hours after starting the laptop, the KAW said that the whole system was consuming only 81W. That was with both laptop and desktop plugged into the UPS and running but idle, aside from this blogging effort in one of the desktop’s browsers. Presumably the reduction in power usage was due to automatic powering down, for the plugged-in but presently unused drives on the desktop.
Now I started using the desktop system. I began by setting up several programs that would keep the drives at least somewhat busy. These included Beyond Compare, poised to begin a copying process involving two drives; DoubleKiller, ready to start looking for duplicate files on a third drive; and chkdsk /b on the two other drives. I also set up several other programs that would occupy the CPU and RAM. These were VMware Player, to run a Windows 10 virtual machine; Adobe Premiere Elements, to produce a video; and Handbrake, to produce another video. I also set up a video editor to produce a video on the laptop. This probably approximated my most extreme usage scenario: it was conceivable but unlikely that, in real usage, I would have all this happening at once. The mere process of getting all these tasks ready to run increased power usage on the KAW to ~110W.
I started with the drive tasks. With all drives busy as just indicated, KAW read ~130W; but within a few minutes that settled back to more like 120W. Then I ran the other programs just listed. From this point onwards, Moo0 System Monitor reported the CPU as being 98-100% busy. The programs raised KAW to ~173W. So that was a surprise: a heavy software load raised the power demand by one-third.
With both laptop and desktop computers thus occupied, the PPB Dashboard continued to say that the battery was Normal, Fully Charged, Capacity 100%, Remaining Runtime 0:27 (h:mm). Now I unplugged the UPS from the wall socket and focused on PPB Dashboard. At the top, the usual message, “The UPS is working normally,” was replaced with, “The utility power fails. Battery power is being used.” There was a similar but not identical message in the Recent Events panel down below. The Input area ceased to say Normal, and instead said Blackout, and its Voltage value dropped to zero. The Output panel stayed at Normal; its Voltage dropped from 122.0V to 120V, and stayed there; and its Load value rose from ~20% to a pretty steady 40%. The Battery panel likewise changed from Normal, Fully Charged to Discharging. Its Capacity began dropping from 100%. Remaining Runtime had been ~35 minutes; now it almost immediately dropped to 13 and then 12 minutes, and continued to decline from there.
With the UPS unplugged from the wall, all software continued for several minutes to run with no visible change. Then, at about 3:45 — that is, not quite four minutes after I unplugged the UPS from the wall — the Firefox browser that I was using to view the PPB Dashboard crashed. Firefox was not in the habit of crashing for me, these days; I have to think the crash was due somehow to the power situation. I was able to restart it, but it refused to accept the admin login credentials. That may have been when the Recent Events pane acquired a new message: “Local communication with the device has been lost.” But the Dashboard ran fine in Chrome, and all the other programs kept on running until the end.
The UPS lasted longer than expected. At 6:00, we still had a reported 57% capacity and an estimated 7 minutes of runtime remaining. The unit continued giving two beeps every 30 seconds for the first eight minutes. Then, with 4 minutes of Remaining Runtime and capacity of 40%, the Battery pane indicated that power was Critically Low; the beeping became constant; and the Recent Events pane said, “Battery capacity is critically low, power could be lost immediately,” followed by, “Shutdown process initiated.” Program windows started coming to the foreground and then closing.
When Remaining Runtime reached 3 minutes, at around 9:15 (i.e., a little more than nine minutes after unplugging the UPS from the wall outlet), PPB abruptly put me at the Windows 10 lock screen. That didn’t seem to make sense. I was able to log back in. But by then, all the program windows had been closed. Windows said, “Shutting down.” The continuous beeping continued until the computer turned off, at approximately 10:30. At that point, with only a few LEDs still running on other devices plugged into the UPS, the beeping reverted to twice every 30 seconds.
It appeared that the UPS still had enough juice to continue to power the entire computer for another couple minutes. Therefore, I have to assume it could have kept beeping and powering those little LEDs for hours if not days. Having had enough of that, I plugged the UPS back into the KAW into the wall outlet. Now, with all computers and components turned off, KAW said the UPS was drawing 122.7W. That was apparently how much the UPS would allow itself to draw, for purposes of recharging: it seemed to stay at that value even after I powered up the desktop computer. According to the Recent Events pane, the UPS was fully recharged within 1:40 (h:mm). At that point, it was back to saying I had ~35 minutes of runtime.
Reactions in General, and to the CyberPower Unit Specifically
This testing yielded a number of observations and reflections that may be useful for others, and also for me, as we continue to learn about various UPS models and capabilities:
- My computer’s power usage was much lower than the estimation tools (above) had led me to expect. Even at an unrealistically high level of power demand, it peaked at around 175W. The suggestion that I needed another couple hundred watts above that seemed completely unrealistic. Of the calculators I tried, the only one that didn’t overestimate by at least 30% beyond my unrealistic peak — and at least 100% beyond normal usage — was Be Quiet.
- It was surprising that the power usage calculators (above) focused on hardware, when software might make more of a difference. Perhaps an ideal power estimation tool would use something like the crowdsourcing approach modeled in UserBenchmark‘s hardware benchmarking, where millions of systems are probed, with users’ consent, to see what they are running and how much it is demanding from various system components.
- I should have bought the Kill-A-Watt device — or, better yet, a competing model with a more readable display — before buying the UPS. It would have paid for itself in demonstrating that I would probably be OK with a $50 UPS.
- Since I had bought a UPS that had turned out to be bigger than I really needed, I had the pleasant experience of seeing that it ran longer than I had expected. The ~8 minutes of runtime seemed like more than enough to start the supplementary power system discussed in the other post.
- A desktop computer was sure to use much more battery power than a laptop. But this testing revealed that an energy-efficient desktop, doing low-key work, might use only one-third to one-half as much energy as a power hog running demanding tasks. A laptop would require even less. Thus, for users who needed to keep a process running, one solution would be to synchronize the main computer to a laptop or a smaller desktop, and plug the latter — not the power hog — into the battery backup. Then, in a power outage, the hog would die, but the smaller machine would carry on, with a data drive that would have the current or nearly current state of the files that had been in use on the bigger desktop. It could help, also, to separate relevant devices from the hog. In my example, that could mean running disk drives as external devices that could continue to draw power directly from the UPS after the big computer shut down.
- In response to the Firefox crash midway through the test, one solution might be to run a browser (perhaps in a portable version) with just one tab open, dedicated to the PPB software. Alternately, PPB should at least be bookmarked in multiple browsers, so that it would hopefully remain available during a power outage. Another solution, here, was to drag the handle from the browser’s address bar to the desktop — and, security permitting, label the resulting desktop shortcut with the username and password (e.g., “CyberPower PowerPanel – admin admin”), since the software apparently insisted on login data that might not come to mind in a moment of crisis.
- I was not sure why PPB began shutting things down with four minutes of runtime remaining, when I had specified two minutes. It was OK for my purposes — I had already enjoyed more runtime than expected, and draining batteries too deeply would apparently shorten their lives. Nonetheless, it seemed that either the CyberLink software had a bug or its time feature was more complicated than it had seemed to be.
- If they were going to shut me down prematurely, one useful feature might be a renewal button that would buy me another 30 seconds. I might punch that button and risk running out of time for an orderly shutdown, if that’s what it took to finish some crucial task or save some essential file.
- I wondered whether some of the amounts of time reported by PPB were understatements. It did not seem that the battery could have been so deeply discharged, if it was able to be fully recharged in less than two hours. I hoped the reason was not that CyberPower was deliberately depriving users of battery time, in a bid to persuade them to buy a larger UPS.
- It was confusing that Low Battery Threshold was set to 300 seconds (i.e., five minutes), and yet that didn’t seem to be when I started getting beeps. The alternate 10% setting for Low Battery Threshold was another mystery: as already noted, the UPS had shut the system down well before we got to that point.
- I would have liked to see a graph providing a historical review of the issues that might have cropped up, of fluctuations in power and voltage levels, and of atypical events, within the past 72 hours. That might have helped me to understand some of these things that didn’t seem to be working right.
- The constant beeping, as the battery ran down, was unhelpful and distractive. One or two beeps would have been sufficient. Unfortunately, it appeared that PPB allowed only an on-or-off choice: either Enabled or Disabled in UPS Setting > Configuration > System. I switched it to Disabled, on the assumption that I would generally notice when the power had gone down — though I might not notice if something accidentally got unplugged or switched off, or if the circuit breaker for that particular wall outlet was the only one that tripped.
- The shutdown process was pretty ragged. Aside from the mysteries of the Firefox crash and the shutdown two minutes earlier than scheduled, I didn’t know why the shutdown process put me outside the lock screen, such that I had to log back in to see what was happening. It was also not like a normal Windows shutdown, to see all those program windows flashing onscreen. Possibly that was necessary for a better program shutdown. It looked chaotic.
- The software did not learn from experience. Having just seen that my system would last only ~10 minutes in a power outage, the Dashboard went back to telling me I had ~35 minutes of runtime. A better solution would allow the user to choose a display of estimated runtime based on the average of the survival times from the last five shutdowns.
On balance, I felt that the CyberPower AVRG750U had turned out to be a legitimate unit, decent for its price. The software was obviously a mixed bag, but at least the hardware had worked more or less as expected.