I have a couple of systems with 1600 watt power supplies. While I don't think they are drawing 1600 watts, I would like suggestions on both the correct sizing and your favorite brand of UPS.

During a storm a couple of days ago I had power outages twice in twenty minutes. Had to reboot all my BOINC boxes twice.

Tom

---

A proud member of the OFA (Old Farts Association).

I have a couple of systems with 1600 watt power supplies.

The load should be no more than 75% of the UPSs rating. Keep in mind UPSs are rated in VA (Volt Amps), their wattage will usually be around 60% of that.
eg my 1600VA Eaton UPS is rated for 960W, so a maximum load would be around 720W.
The problem with running much higher than 75% of it's rating is that it doesn't take much of an increase to get close to it's limit. If several things were to peak at the same time, you could hit the overload point of the UPS and it will shutdown.
Plus the runtime of a UPS under full load can be less than 1 minute: if the power does go out, you want time to shut things down gracefully. Even using the UPS software to automatically shut down the system, with the UPS at it's limits the batteries could still go flat before the system has finished shutting down.

---

Grant
Darwin NT

Hi,
Yesterday (Saturday morning) my house experienced two "brown outs" where the lights wobbled and then came back on.

It blew my Seti boxes offline.

One of the boxes is pulling about 800 watts off a 1600 watt PSU. So my now installed Cyberpower 1500va ups seems to be perfectly happy to support it.

The other box almost certainly is drawing more than that. I presumed it might be pulling near the top of the PSU and started looking for UPS's that would support 1600 watt draw.
It seems like I was looking at $800 or so if I am not confused.

Since I am price sensitive it occurred to me that a power conditioner that merely kept the voltage close to 110 volts might be cheaper even at that amperage level.
So if I presume that any extended power outage (more than 30 seconds?) my machines will be stopping then the only thing I need to worry about are brown outs.
I THINK I found a line conditioner for maybe $200 that will handle a full 1600 watts.

Is this a "good" idea?

Tom

---

A proud member of the OFA (Old Farts Association).

The majority of "power conditioners" are designed to "cap" any peaks above the nominal maximum mains voltage (typically to nominal plus 10%), and "boost" drops only for a very short periods of time (very often less than 1 second). So they will not help very much in the case of your longish brown-outs, to cover those you need active "buck-boost" systems, and they are not exactly cheap.
One thing to look out for on any of these "UPS-type" systems is the load-surge rating - what current they are capable of for the <1second while the computer PSU starts up. Many computer PSUs will draw several times their rated current during start-up so the "UPS" needs to be able to cope with that as well as the constant running load.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Although my current UPS isn't upto supporting it, it did give me 1,170 watts draw for my current (no pun intended) seti (gpu) wcg (cpu) mix of crunching.
https://setiathome.berkeley.edu/show_host_detail.php?hostid=8696615
I am running a 1600 watt PSU from EVGA. I am hoping I am not exceeding the efficiency power curve for this PSU.

I know a 1600va UPS won't cut it. Would a 2000va be an acceptable (eg. reliable/fairly efficient) choice?
I am running 110 volts AC on the input side. The circuit I am on breaks at 20 amps. And the extension cord is rated to 1600 watts.

My last choice(s) include reducing the total # of gtx 1070's reducing to lower drawing gtx 1060's (free, I have a surplus of 1060's). And/or going to lower power draw cpus (costs a little money).

Tom

---

A proud member of the OFA (Old Farts Association).

If your rig is consistently pulling over 1000 watts, then the typical 1500VA UPS' won't cut it. Most of the typical consumer UPS are only for limited loads and use a cheaper modified sine wave output that doesn't have the capacity of a true RMS sine wave output. Two identically rated 1500VA UPS will not have the same load capability. The BackUPS BRG1500G 1500VA rated UPS can only handle a 760W load. The SMT1500C 1500VA rated UPS can handle 980W.

The APC SmartUPS SMT1500C will only handle a 980W load without complaining about being overloaded. My 4 gpu Threadripper system often is running around 1150 watts so the SMT1500C didn't cut it. I had to move the next level up and get the APC SMT2200C UPS which can handle the load easily and still have 20+ minutes of runtime on that load. Runs off a 20A circuit using the standard 20A receptacle.

Amazon used:
https://www.amazon.com/APC-Smart-UPS-2200VA-Battery-SMT2200/dp/B0036RDRXG/ref=sr_1_4?keywords=smartups+2200+c&qid=1571585000&s=electronics&sr=1-4
Amazon new and latest:
https://www.amazon.com/APC-SmartConnect-Interactive-Uninterruptible-SMT2200C/dp/B077YDWR3Y/ref=dp_ob_title_ce

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

If your rig is consistently pulling over 1000 watts, then the typical 1500VA UPS' won't cut it.

As an experiment to see what my Cyberpower 1600va would take I am pulling upto 950? odd watts.

That translates into 3 Gtx 1070ti's, 1 gtx 1070. 1 gtx 750ti was a "bridge to far" :)

887 watts and/or as high as 102% of the maximum load according on some peaks. So far it hasn't stayed up that high.
It starts complaining if the load stays above 100% for very long.

Tom

---

A proud member of the OFA (Old Farts Association).

The majority of "power conditioners" are designed to "cap" any peaks above the nominal maximum mains voltage (typically to nominal plus 10%), and "boost" drops only for a very short periods of time (very often less than 1 second). So they will not help very much in the case of your longish brown-outs, to cover those you need active "buck-boost" systems, and they are not exactly cheap.
One thing to look out for on any of these "UPS-type" systems is the load-surge rating - what current they are capable of for the <1second while the computer PSU starts up. Many computer PSUs will draw several times their rated current during start-up so the "UPS" needs to be able to cope with that as well as the constant running load.

Thank you for mentioning "buck-boost" as a search term. If found this: https://ustpower.com/power-conditioning-technologies/ which lead me to the term "sag ride through power conditioner" which describes what I am looking for (if it is cheaper than an equivalent UPS) and that lead me to this:

Tripp Lite LC2400 Line Conditioner 2400W AVR Surge 120V 20A 60Hz 6 Outlet 6-Feet Cord

which is going to the tune of $206 on Amazon.

Tom

---

A proud member of the OFA (Old Farts Association).

that is a line conditioner. no battery backup capability at all. only useful to smooth out voltage fluctuations

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

As the article says - Buck-Boost TRANSFORMERS are manually controlled, which is OK if your source voltage is, say, 100V continuously and the attached equipment fails to operate below 105V. They are NOT automatic.
Unlike "proper" Buck-Boost POWER SUPPLIES, which will automatically regulate the output voltage to a set value when the input goes up or down - but again within a finite range. Neither of these work when the input voltage drops to near-zero as one gets in brown-outs. It is quite possible to use a laptop PSU driving an inverter to get a very stable output voltage with quite substantial input voltage swings, but that comes at quite a price to get a 2kW output device (compared to the typical laptop supply delivering about 100W).
And none of these work if the voltage drops to zero for more than a couple of cycles, at which point you are into needing a battery somewhere in the system to keep tings going - and that is essentially some sort of UPS, and a 2kW (which is probably going to be a 2500kVA unit), 10 minute run, UPS is almost certainly out of your budget :-( (and don't even think about second hand as the first thing you will probably require is a new set of batteries at $$$)

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

that is a line conditioner. no battery backup capability at all. only useful to smooth out voltage fluctuations

The original presenting problem was a 2 brown outs within 20 minutes of each other. Would this probably fix that?

Tom

---

A proud member of the OFA (Old Farts Association).

No. You need a battery backup. An UPS.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

@Ian,
Question. What kind of power/load is this system with the low powered cpu's drawing? https://setiathome.berkeley.edu/show_host_detail.php?hostid=8433872

Tom

---

A proud member of the OFA (Old Farts Association).

@Ian,
Question. What kind of power/load is this system with the low powered cpu's drawing? https://setiathome.berkeley.edu/show_host_detail.php?hostid=8433872

Tom

I don't have it on any kind of power meter (240V with C13-C14 plugs), so I can only estimate.

200W power cap on the 2080
200W power cap on the 2080
225W power cap on the 2080ti
2x 60W CPUs @ 15% load
1200W Platinum PSU
throw on 25W for Mobo/RAM/misc

I figure probably around 700W from the wall.

---

Seti@Home classic workunits: 29,492 CPU time: 134,419 hours

I've been thinking about UPS' recently. I assume that for a desktop computer, BOINC doesn't recognize UPS as a 'battery'? I have an old UPS that I used to have my desktop plugged into. I took it off the UPS when it would start to beep with what appeared to be no power fluctuation. I haven't gone back to see if there is something fixable with the UPS, or if it just had old components that needed maintenance or repair (if even possible). The few times the power did go out, I don't think I was home and either it rebooted or just turned off, I don't remember.

The one nice thing about a laptop is that when wall power is interrupted, it runs off batteries, and suspends Boinc. I haven't done the calculation, but I'm guessing to keep a desktop powered AND running Boinc would be pretty big, especially with a GPU. What experience does anyone have with this?

---

Seti@home classic: 1,456 results, 1.613 years CPU time

Bill - that all depends on the operating system and the UPS in question. Some combinations work, others don't and (worse) some that should don't.

As to your second question - are you talking about keeping a laptop with a GPU running for a few minutes, or are you talking about running a desktop from a laptop PSU?

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Bill to add, if the O/S sees the UPS as a battery, then BOINC will (should) also see it as a battery as BOINC reads the O/S for that info.

Bill to add, if the O/S sees the UPS as a battery, then BOINC will (should) also see it as a battery as BOINC reads the O/S for that info.

But unfortunately, BOINC doesn't in all cases. That is why I have my threads in BOINC Client over at the BOINC message board and on Github and here in NC. I have only one report that BOINC suspends computation in Windows when the computer switches to the battery backup in Windows. It does not recognize when the host switches to battery backup in Linux even though the OS knows that has happened. BOINC is only considering laptops that shift to battery power to suspend work. It should recognize the OS signaling that the host has shifted to battery power for desktops also and suspend work. That would shed the compute load from the desktop and preserve more of the UPS load runtime. The client needs to be changed to recognize the other OS' signaling of a shift to battery power.

Boinc on Ubuntu 18 LTS not detecting "on battery"
Add support for Windows 10 Battery Saver #1471
What module contains the code to detect when the host computer has shifted to battery power?

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

The trouble is Keith, not every UPS uses the same API, and not every operating system respects all of them, so there would have to be several variations on the theme to cover the whole spectrum :-(
This is unlike laptop batteries which, thankfully, now use a common API call.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

The trouble is Keith, not every UPS uses the same API, and not every operating system respects all of them, so there would have to be several variations on the theme to cover the whole spectrum :-(
This is unlike laptop batteries which, thankfully, now use a common API call.

No that is not true. No matter what software any brand of UPS uses, it ultimately connects to and sets the upower flag which is the only power API in Debian/Ubuntu. That covers probably 80% of all *nix variants. Might not cover Fedora or BSD. There still is some sort of power management API running on all *nix variants.

The problem is that the code that controls suspending compute when on battery in Linux is out of date and hasn't been updated. It is checking for flags that don't exist anymore in modern Linux. The signaling in Linux has changed to a different interface.

https://setiathome.berkeley.edu/forum_thread.php?id=84719&postid=2014182

https://github.com/BOINC/boinc/blob/master/client/hostinfo_unix.cpp#L195

I am not a programmer but I think all that needs to be done is add a new detect Method mechanism named Upower so that it looks for the upower interface and reads its flag.

#elif LINUX_LIKE_SYSTEM
static enum {
Detect,
ProcAPM,
ProcACPI,
SysClass,
Upower,
NoBattery
} method = Detect;
static char path[64] = "";

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)