Other Power consumption and DC-DC/"DC-ATX" from Batteries?

[zovc]

Hey,

I'm (again) looking into moving towards a mobile living situation like an RV, and I'm trying to figure out if it'd be feasible to account for a lot of computer usage off of mostly something like solar power. I like playing games and I like having a powerful computer, so I'd prefer to not give up that if I don't have to.

(For what it's worth, I've given some consideration to getting a very energy efficient laptop/NUC-style machine and utilizing cloud computing, but I am waiting to see how 4G coverage grows, how 5G pans out, and how unlimited data and mobile latency ends up.)

I don't know much of anything about how battery systems work, but I've been trying to do some homework and it seems like a lot of things aren't compatible with the numbers I'm used to. (RV) Batteries seem to be rated in amps and amp hours, where as computers are typically rated in watts and there's a general assumption that you're converting AC power to DC power.

So, I'm trying to learn more about this stuff (because it'd obviously be useful to know if I'm going to be living an independent/mobile lifestyle), but I'm having a hard time piecing everything together since it's presented as very different worlds. Here are some basic 'principle' questions I have, and I'd appreciate any help sorting them out.

• What should I be looking at to figure out how much power a computer would draw from a RV's batteries? Do you have any suggestions for estimating how much power I'd be drawing at any particular percentage of my computer's rough maximum TDP/W? Is there a better way of estimating that? (I understand that TDP/W is generally a measurement of heat output rather than power consumption...)
• If I'm going directly from my battery system's 12v to a computer's 12v "DC-ATX" power supply, will I end up having a more energy efficient system? I'm assuming that my components will consume roughly the same amount of power, but I might be saving some power for not having to convert from AC to DC. Do you have any advice for connecting my computer's power supply to a RV's DC power?
• I've found converters for Watts to Amps, but I have no idea if am getting a sensible estimate by going from ~TDP to Amps and multiplying that by hours to get "Amp Hours..."

And here are some more pointed questions:

• How much power consumption would you expect a system to have under the following loads? (You can frame this however you want, honestly. Whether you're assuming a certain hardware setup, or you want to give percentages of max power consumption, or whatever else.)
  • Idle on the desktop, possibly running some background processes like a torrent client.
  • Random web browsing and YouTube videos/streams.
  • 2D "indie game" able to be run off of the iGPU. (Likely run by the GPU, though, unless I am able to set up some sort of DIY "Nvidia Optimus" like situation.)
  • Old school, easy-to-run 3D game that is able to be frame limited to ~30, 60, or 120fps.
  • Emulated game (~PS2/Gamecube era) with easy to render frame limited 3D graphics.
  • AAA game that is a load for both CPU and GPU.
  • Streaming to a service like twitch while gaming. (Ambitious assumption that a mobile network could and would sustain this.)
  • Rendering a 4K video.
• What kind of hardware would you be looking at if you needed to run a system off of battery power?
  • Do you think you could build a system that would be more efficient than a similar laptop? (I.e. a desktop with an undervolted 1070 vs a laptop with a Max-Q 1070.)
  • How would you build your system if you knew it was going to be in a RV that might roll over potholes or face other road turbulence like sudden stops?
  • Do you believe there is a certain sweet spot between performance and efficiency?
  • Is more powerful hardware, when undervolted, more efficient than less powerful hardware? (I.e. a 1080 undervolted vs a 1060.)

This is very much a theoretical pipe dream for me, still, but it's one that I keep coming back to. So, if you have any useful knowledge or just feel like humoring me, it'd be appreciated. Ideally, I could set up a system that has satisfactory power while still being able to be run for several hours a day just off of solar power. It sounds lofty, but I think within a few years it could be sensible as all of these technologies evolve independently. ('These technologies' being solar power, batteries, DC-DC PC power supplies, efficient PC hardware, and random odds and ends in between.)

 

[Kmpkt]

I don't know much of anything about how battery systems work, but I've been trying to do some homework and it seems like a lot of things aren't compatible with the numbers I'm used to. (RV) Batteries seem to be rated in amps and amp hours, where as computers are typically rated in watts and there's a general assumption that you're converting AC power to DC power.

Really most of these units are interchangeable using the basic formula [Power (W) = Voltage (V) x Current (A)] . Watt hours are simply (Voltage x Amp) hours. What I mean by that is if you have a 12V system at 6Ah then you have 72 Watt Hours. Conversely if you know you have a 12V battery system in your automobile and it is rated at 12 amp hours, then the total capacity would be 144 watt hours. As another example, I am doing a battery system right now and it uses three 6A, 18V power tool batteries. Each battery produces 108 Watt hours of power for a total of 324 watt hours.

What should I be looking at to figure out how much power a computer would draw from a RV's batteries? Do you have any suggestions for estimating how much power I'd be drawing at any particular percentage of my computer's rough maximum TDP/W? Is there a better way of estimating that? (I understand that TDP/W is generally a measurement of heat output rather than power consumption...)

Once you know the watt hours available to you then you can do one of a couple things. One would be to calculate the consumption of the unit based on rated wattage per component (GPU/CPU/Chipset/RAM). I personally think the better way to go would be to buy a Kill-A-Watt unit and monitor total watts consumed during a standard hour of computer use.

If I'm going directly from my battery system's 12v to a computer's 12v "DC-ATX" power supply, will I end up having a more energy efficient system? I'm assuming that my components will consume roughly the same amount of power, but I might be saving some power for not having to convert from AC to DC. Do you have any advice for connecting my computer's power supply to a RV's DC power?

Yes. For reference looking at the HDPlex DC-DC units you would lose about 6% of your power as heat going from 16-24V down to 12V. Using something like a G-Unique that is natively 12V is almost lossless as you'll only have to do a small amount of conversion to 3.3V and 5V.

How much power consumption would you expect a system to have under the following loads? (You can frame this however you want, honestly. Whether you're assuming a certain hardware setup, or you want to give percentages of max power consumption, or whatever else.)

This is a pretty tough question to answer. How much power you consume will directly vary with not only the choice of components (2 core vs 4 core vs 8 core chip, T series vs K series intel chip, etc.) but how you tune them. Overclocking the CPU/GPU/Memory will increase overhead while underclocking and undervolting will make things more efficient, stretching out battery life. Furthermore on batteries, you'll want to be much more cautious about how many background operations you're running. I am planning a battery powered unit and plan to do a number of benchmarks and stress tests with various configurations for undervolting, under clocking and software optimization. Again I would make use of a Kill-A-Watt unit to measure overall power consumption in an hour of each your the aforementioned scenarios.

What kind of hardware would you be looking at if you needed to run a system off of battery power?

This really comes down to what you need. The easy way to do this would be to choose a 35 or 45W T-class intel chip and RAM that has lower clocks (2133 DDR4). Additionally choosing an nVidia GPU that really uses no more power than you absolutely need would be advisable. If the 1050ti would suffice it would be my first choice while the 1060 would be my second. 1080 could be good as it is a more efficient chip overall, but you'll need to perform a lot of tuning to get the power consumption down.

How would you build your system if you knew it was going to be in a RV that might roll over potholes or face other road turbulence like sudden stops?

I'd seriously consider a wall mounted case myself.

Is more powerful hardware, when undervolted, more efficient than less powerful hardware?

I will be checking this in the not too distant future hopefully. I actually purchased an Intel 7700K, 7700 and 7700T to test exactly this with CPUs. Since both the 7700 and 7700T are locked in the BIOS, what I plan to do is do progressive under clocks with the 7700K to see if I can stably match both processors, how the under clocked performance compares, and if the T chip is even worth having ever. Given the binning findings on the Risen 7 chips, I am really hopeful that a better chip will run at a lower voltage/wattage across the entire spectrum.

 

[Choidebu]

..and I'm trying to figure out if it'd be feasible to account for a lot of computer usage off of mostly something like solar power.

Interesting idea, but if you plan to install solar panels in your RV, I think it's more cost effective to have them charge your main battery instead of just powering your pc (since you'll still use a battery - solar isn't reliable)
Panels are rated in (maximum!) watts, so you can actually plan how many you'll need for your system. I reckon a sub 200W total system, no more or you'll need big multiple panels (if you really want them tonjustbrun your system)

(For what it's worth, I've given some consideration to getting a very energy efficient laptop/NUC-style machine

NUCs are in sub 100Ws, 50-60W is common. Ones with older Amd APUs can game light. A gaming laptop can be anywhere from 120-200W.

• If I'm going directly from my battery system's 12v to a computer's 12v "DC-ATX" power supply, will I end up having a more energy efficient system? I'm assuming that my components will consume roughly the same amount of power, but I might be saving some power for not having to convert from AC to DC. Do you have any advice for connecting my computer's power supply to a RV's DC power?

Yes. Ac-Dc bricks have efficiencies around 90-95%. Dc-atx range 95%++.
If you use both, you're looking at around 80% ish.
BUT if you plan to power the system directly from the RV battery which is connected to the engine, see below..

• How would you build your system if you knew it was going to be in a RV that might roll over potholes or face other road turbulence like sudden stops?

Back in the days of carputers, there are few sorts of dc-atx modules that can compensate for spikes or lows that can happen in car environment. It can also be wired to automatically turn on and gracefully turn off your system when the ignition is off.
I don't know much about RVs to tell you more, but it's worth researching about IMO.
Other things to consider is fan and heatsink size, no hdds of course..

 

[CaneyJ]

I run a gaming computer from my boat batteries as well as directly from a portable 6S (19V) LIFEP04 pack.

My specs are:
- i5-4590S @ 3Ghz (65W TDP but I disabled the 3.7Ghz turbo boost saving 15-20 Watts consumption)
- Basic Asrock H81 ITX motherboard
- 2x4GB 1.35V Low-Profile Samsung 1600 DIMMs
- Palit GTX1050 Ti 4GB
- HD-PLEX 160W fed from a DC-DC Boost converter from the boat’s 12V lead acid batteries to get 19V or directly from a LIFEPO4 pack with a Low-Voltage disconnect + Ammeter to monitor.
- 24” LG Monitor converted to run off the same 19V boosted input as I chose a model with an external 19V power brick which I then bypassed.

The combined setup including the monitor under average gaming load uses around 100 watts or 80W for the computer alone.

Converting that to batteries means I use draw around 9 amps per hour when accounting for a little conversion efficiency loss and pheukert effect.

When applied to a small 75AH leisure battery, I’d estimate that gives your around 4 hours gaming time with no other external draws (e.g. lights, fridge). You should be able to scale up batteries for a longer time.

You can’t run a computer from 12V directly. If you have solar or an alternator feeding the battery then the voltage could spike upto 14-15V. If the charger has an maintenance cycle this could go up to 16V which would deffo fry some things.

My computer idles at around 23 Watts and runs at 35 watts average when watching a full-screen youtube video. I’ve also measured a peak of 112 Watts. All these readings are taken from the DC side, not from the AC wall side so you might want to add 10% to those figures when using an AC adapter but since it’s from the batteries directly this efficiency doesn’t need to be considered.

80-90 watts gaming load was taken from playing Overwatch.

If I could do it gain I might use a GTX1060 and power limit it to maybe 70-80% to benefit from the CUDA cores. Stick with any i5 quad Core and just disable turbo boost. Not sure how efficient Ryzen 3 is with under-volts and disabled boost. If I want maximum gaming longevity then go with a G4560 and a GT1030 with reduced setting and you could game for maybe 6 Hours off a 75AH, 12Vbattery. It’s all a question of what you need.

I put more details of my gaming setup in this thread:

https://smallformfactor.net/forum/t...-decent-enough-pc-experience.3049/#post-60285

 

[zovc]

Interesting idea, but if you plan to install solar panels in your RV, I think it's more cost effective to have them charge your main battery instead of just powering your pc (since you'll still use a battery - solar isn't reliable)
Panels are rated in (maximum!) watts, so you can actually plan how many you'll need for your system. I reckon a sub 200W total system, no more or you'll need big multiple panels (if you really want them tonjustbrun your system)

Thanks for all the feedback, folks! I'm digesting it all, just a little distracted by work.

I wanted to clarify, though. When I mentioned powering my computer 'by solar,' I meant charging batteries and using those to power my computer. Though, if there was some way to use 'excess' power after my batteries were charged to power electronics that would obviously be ideal? (I think?)

Anyways, I'll give a better reply some time soon, hopefully this evening.

 

[Choidebu]

well that's makes things simpler.
Get a deep cycle battery, some solar panel, solar charging equipments. Research (I recommend to start at mini-box) about dc-atx PSUs that is designed for harsh environment that can accomodate voltage spikes and lows.

About 'excess' power, there's a whole range of electrical appliances designed to be powered off of car cigarette plug like small fridges, electric kettle and mobile chargers, as long as you limit yourself to batteries that outputs 12V. Other household appliances will need an inverter to convert dc to ac.

 

[zovc]

I personally think the better way to go would be to buy a Kill-A-Watt unit and monitor total watts consumed during a standard hour of computer use.

About using a Kill-A-Watt, do you have any suggestions for 'scientifically' recording its readings? I have one and I've used it to monitor and get a rough idea of things but I feel like it's very imprecise to count on myself to note the highs and lows and whatnot when it changes pretty rapidly, especially over the course of a long benchmark.

This really comes down to what you need. The easy way to do this would be to choose a 35 or 45W T-class intel chip and RAM that has lower clocks (2133 DDR4). Additionally choosing an nVidia GPU that really uses no more power than you absolutely need would be advisable. If the 1050ti would suffice it would be my first choice while the 1060 would be my second. 1080 could be good as it is a more efficient chip overall, but you'll need to perform a lot of tuning to get the power consumption down.

I will be checking this in the not too distant future hopefully. I actually purchased an Intel 7700K, 7700 and 7700T to test exactly this with CPUs. Since both the 7700 and 7700T are locked in the BIOS, what I plan to do is do progressive under clocks with the 7700K to see if I can stably match both processors, how the under clocked performance compares, and if the T chip is even worth having ever. Given the binning findings on the Risen 7 chips, I am really hopeful that a better chip will run at a lower voltage/wattage across the entire spectrum.

I really need to do more looking into the T-series chips (and I am excited to see what you find when comparing the undervolted K and regular CPUs to them), but the 'enthusiast' in me kept telling me it was bad value when I'm currently living in a conventional situation with access to wall power.

You can’t run a computer from 12V directly. If you have solar or an alternator feeding the battery then the voltage could spike upto 14-15V. If the charger has an maintenance cycle this could go up to 16V which would deffo fry some things.

IIRC most of the "DC-ATX" power supplies I've looked at accept a range of voltages, I think HDPLEX is ~16-24V? That said, we have talked a little about this since the post I'm quoting--I'd be running my system off of batteries that are charged by solar. I don't think I did a good job of making that clear in the original post. Batteries should stay pretty consistently around 12V, right?

My computer idles at around 23 Watts and runs at 35 watts average when watching a full-screen youtube video. I’ve also measured a peak of 112 Watts. All these readings are taken from the DC side, not from the AC wall side so you might want to add 10% to those figures when using an AC adapter but since it’s from the batteries directly this efficiency doesn’t need to be considered.

80-90 watts gaming load was taken from playing Overwatch.

Are these numbers including your monitor's power draw?

Also, when I was asking if I could run my system off of "excess" solar power I think there might have been another misunderstanding. I definitely intend to run other things off of my batteries besides my computer. At least lights, but probably other appliances like a fridge and whatnot. What I was trying to ask was "Assuming my batteries are at 100% charge, could my system draw power directly from the solar panels rather than depleting my batteries?"

 

[Kmpkt]

About using a Kill-A-Watt, do you have any suggestions for 'scientifically' recording its readings? I have one and I've used it to monitor and get a rough idea of things but I feel like it's very imprecise to count on myself to note the highs and lows and whatnot when it changes pretty rapidly, especially over the course of a long benchmark.

I don't think the Kill-A-Watt is going to tell you anything useful about power spiking. I was suggesting its use more for monitoring consumption over time. I tried using one when I was stress testing the HDPlex 300W AC-DC. I was regularly triggering OCP at 400W but the Kill-A-Watt never registered consumption above 310W.

 

[CaneyJ]

A lead acid battery can range in voltage from around 9V when pretty much flat to nearly 15V during a solar controllers “absorption” stage of charging. If you are running a computer during the end of a batteries charging cycle it will run on that input voltage.

At rest, a fully charged ‘brand new’ lead acid battery will sit at around 12.8-13V with no load or charge being applied. When receiving a trickle charge at 100% state-of-charge during a solar charge controller’s float/maintenance cycle it will hover at around 13.2V-13.6V depending on the controller’s battery setting.

For example; say you go mental and run an off-grid computer directly from a battery to the 12V ATX rails and it’s rocking a Threadripper and SLI GTX1080’s. At full-load this will draw down the batteries voltage even in a full state of charge to maybe 12.5V under load. This is called battery “sag” and when the computer comes off load this will rebound back to 13V+ and pop the motherboards VRMs.

I would never run a computers 12V rails from a battery directly as it easily strays outside the 10% tolerance I’d be happy with. The best wide-range power supply optimised for 12V I’ve seen is the 250W Mini-box M4-ATX. This can supply 12-Amps [144-Watts+] on its ATX 12V rail with a 16A Peak [192-Watts+]. This should in theory be enough to run a Mid-TDP i5 and maybe a GTX1060?

However; the M4-ATX is a pretty old power supply nowadays. I would use a safer method which is using a DC-DC step-up converter to convert the batteries variable 12V up to a rock-solid 19V which is much less inefficient than using an AC Inverter, Laptop Adapter only for it to go back to 12V or 19V DC. The fixed 19V can then be safely fed into my modern HD-Plex 160W PSU as well as the monitor directly.

Kill-a-Watt meters are also not an accurate measure of current draw from a battery directly as they measure the AC draw from the AC adapter side which could be reading 10-20% more than the computer really uses and are poor at tracking momentary power peaks.

The power numbers I quoted are for my computer alone. The DC draw of my 24” monitor is around 15-20 watts extra on top depending on its brightness setting. For example, my computer will game at 80-90 watts average + 20 Watts from the monitor. At idle the draw is 25-35 Watts from the computer and another 20 Watts from the monitor. If you use a 15.6” USB powered HDMI monitor like a GECHIC then you can save a bit of power as it might only need 5 Watts to run it.

A small 75AH leisure battery can be drained down to 50% giving us 37.5 useable amp-hours. You should recharge a battery once its voltage has dropped to around 11.8V. If we take the average of a fully charged battery minus its end voltage we get an average of 12.3V. So we take the 37.5AH available X 12.3V giving us 461 Watt-Hours.

My computer therefore when gaming on a combined monitor and PC load of 100-110 Watts every hour means I can get 4-Hours of solid gaming done before I need to recharge. If I use a pair of 225AH golf cart batteries commonly used in RV’s with no other DC loads I can get 13-hours of gaming done before recharging and that doesn’t account for the constant input from a solar setup.

Say you have 300 Watts of solar on your RV. You will need a 20 – 30 Amp MPPT solar charge controller to take advantage of that. Solar charge controllers often have a load output but this is unregulated and often fed through from the battery terminals with a relay to disconnect the load output if the batteries go below a certain voltage due to insufficient sun. If you solar panels produce more watts than your computer draws then you won’t draw any from your batteries.

 

[Overtek]

To me in my ideal world head something like a double online conversion UPS where you could drop the DC straigt in to charge batteries and also still be able to hook up mains power when available without causing any interruptions or disturbance on the output AC side. But these things are not cheap, that said, get lucky and you could pick one up on the cheap. I managed to get a 3000VA one for 50quid some years ago straight off ebay that was out someone's office. Refreshed the batteries and resold for a tidy profit. £800-900 new they were over 10 years ago.
It was a ermm damn what was the company argh my brain, even having visited them many moons ago in Taiwan I can't remember. The model was a DS3000 Durable series. Ahh Opti UPS.

Anyway just a thought to throw at you for consideration.

 

[zovc]

I will be checking this in the not too distant future hopefully. I actually purchased an Intel 7700K, 7700 and 7700T to test exactly this with CPUs. Since both the 7700 and 7700T are locked in the BIOS, what I plan to do is do progressive under clocks with the 7700K to see if I can stably match both processors, how the under clocked performance compares, and if the T chip is even worth having ever. Given the binning findings on the Risen 7 chips, I am really hopeful that a better chip will run at a lower voltage/wattage across the entire spectrum.

Hey @Kmpkt,

I get the impression you've been pretty busy, but did you ever get to compare the 7700K, 7700, and 7700T? I'm again falling into the hole of looking at doing a camper conversion, and am trying to make sense of whether a DIY 'desktop' computer is worth it or if I should just get a laptop that'll last hours under most loads besides gaming, and can be charged at cafes and the like.

There's a few other options I've been considering, like a laptop + eGPU dock, a laptop for general use and a more beefy desktop for more intensive use, or even possibly two laptops where one's super light and one's able to handle heftier tasks.

That refurbished GS63VR is pretty impressive with the mobile 1060, but I honestly can't see its 6700HQ lasting in heavier use cases... I got into using VCV rack and even that can bog the CPU down. Its battery life also is pretty weak, even with a fair undervolt. Makes me skeptical of trusting a mobile CPU to age well with gaming or light production tasks.

 

[Kmpkt]

Honestly that project is something I had completely forgotten about. I'll see if I can find the time over the holidays.