(update) As far as the original question goes...
or, its direct page: http://pe.amsdell.com/ippsspec.htm
The catch is, it's probably useless for any modern PC, because it's only 200 watts & appears to lack ATX12V power.
Here's another. Unfortunately, like the first, it too lacks ATX12v and has only 200 watts:
http://www.indocomp.com/IND-UPS200-ATX.html (kind of wimpy, though)
...and my personal favorite that I'm probably going to burn my weekend trying to find somebody on earth who actually sells, because it explicitly advertises 45-65hz tolerance for input power and generator-compatibility. It's not built into the ATX power supply (it's built into a box that fills a 5-1/4" drive bay), but unlike the first two, it would actually be useful in a modern PC:
insofar as another answer mentioned...
They are designed to allow switch over to a generator
Actually, 99% of the non-Enterprise sub-$1,000 UPSes on earth are completely dysfunctional when sitting between a computer and a generator... and most of the Enterprise-grade ones, too. There are plenty of war stories about people who breathed a sigh of relief when their new generator kicked in... then had their day ruined when the UPS(es) refused to switch over, and stubbornly remained on battery power until they unceremoniously ran out.
The main thing that keeps UPSes from working with generators is the frequency of the power they make. As a practical matter, the frequency of power from all non inverter-type generators is going to drift as the load changes. With small and cheap generators, it lurches and thrashes around. With large (100-kilowatt or more) generators, it drifts and wobbles. As far as most UPS logic boards are concerned, none of them are good enough.
Making matters worse, the UPS problem gets lots of people to junk a perfectly good (or at least adequate) generator and buy a new inverter-type generator, only to discover that inverter-type generators bring an entire host of NEW problems with them.
Most cheap inverter generators don't output sinusoidal power -- they approximate it with stepped square waves. And most UPSes won't tolerate it, and will reject it as firmly as they rejected the frequency-unstable power from the conventional generator that came before it. The sad irony is that the pseudo-sinusoidal power put out by cheap inverter-type generators is actually superior to the dirty square-wave TOTALLY non-sinusoidal power put out by most UPSes.
Anyway, back to the drama. Frustrated with generator #2, you junk it, and buy a $3,000 pure sine inverter-type generator... then discover that it doesn't handle radical load changes (like when the UPS kicks in or out) well, and ends up in yet another death spiral battle to battery death with the UPS. The UPS kicks in, and the generator fires up. The UPS sees line power, and switches the load over to it. Hit with an instant surge of several hundred watts, the total harmonic distortion of the inverter's power goes through the roof for a fraction of a second... which, to the UPS, looks like power that's deviating from 50/60hz. So the UPS kicks in, and the cycle begins anew.
Stir, rinse, repeat, and you'll see why everyone who has to try and make a UPS play nicely with a generator ends up getting frustrated. In the end, it all keeps coming back to the complete intolerance of most UPSes for deviant line frequency.
The sad thing is, there's NO TECHNICAL REASON why it has to be this way.
UPS makers say they fail over to battery power when the frequency is out of spec because they can't fix it otherwise, then mumble something about AC motors not working well when the frequency deviates. The problem with that logic is that 99.9% of the computer hardware likely to be plugged into a real-world UPS doesn't actually care (much) what the exact frequency is or how stable it is from second to second, as long as it's somewhere between ~45 and 65 hertz.
It's a classic case of optimizing the wrong thing for the wrong reason just because it's easy to do, and might have made sense at some point in the past. It's easy for a UPS to determine whether or not the line power is within 1hz of 50 or 60hz. All you have to do is trigger interrupts on the zero-crossing, and compare the time since the last one to your estimate of how long it should have taken. Twenty years ago, before UPSes had microcontrollers with faster CPUs and more ram than a Commodore 64, it might have even made sense. But now, it's just plain stupid, counterproductive, and causes more grief and harm than it prevents. We need to get UPS manufactuers (APC, are you listening?!?) to give us an option to have it just disregard the line frequency, and sync up to whatever it finds (if it finds anything at all) as long as the voltage is within some reasonable range.
Most of APC's UPSes (even their lower-end ones) actually have a setting that reduces its sensitivity to voltage spikes & surges, including models whose official free monitoring app doesn't include the option to change the setting. Just connect at 2400/8N1 and send two characters with a brief pause between them: 's' and 'L/M/H' (L=low, M=medium, H=high). Make sure the PC you're using isn't actually running from the UPS at the time you connect, because some models will interpret DTR's change as a legacy "shutdown" command. However, this setting won't do you the least bit of good in most cases, because the generator's frequency will still keep it from working.
This is purely anecdotal, but from what I've read, TrippLite and Cyberpower are the most rigidly-intolerant of frequency-deviance. APC will tolerate +/- 1hz, but no more. According to the official specs, the Ultra (CompUSA/Tiger Direct) 1000AP will tolerate +/-10%, which means (in theory) it should tolerate anything between 54 and 66hz. I own that particular UPS, but I haven't been able to verify it or its generator-friendliness. It appears to be a rebadged Powercom BNT-600A, but the BNT-600A officially tolerates +/- 5%. Then again, I believe frequency-tolerance is entirely a function of firmware, so it's quite possible that Ultra might have had Powercom tweak the firmware to make it twice as tolerant as their own model.