The Electrical Usage So Far This Summer For Linux Benchmarking

[caligula]

I had the impression that not all benchmarks require huge servers. I have one 160W (200W peak) PicoPSU which should be sufficient for a non-overclocked system with ~70W GPU (from the bus) and ~70W CPU. Currently running a system with 250W GPU (geforce 980 ti) and 45W skylake CPU. It's powered by a Meanwell 12W power supply. You could add 3.5" HDDs as many as you like since they also use 12V directly.

Also all fans and pumps for liquid cooling use 12V directly..

[Fry-kun]

Having just installed a solar system, I have to mention it's a lot more involved that most people think.
If you're going the DIY route & installing on your roof, keep in mind that you may need a permit from your city & need an inspector to come check that your install is up to building & fire safety codes.
DIY or not, with this amount of usage it's unlikely you'll build a system to drop your electrical bill to $0* -- but it'll definitely provide a good offset. Ideally, you'll drop into a lower usage tier and reduce your electrical bill by a good amount -- that should help pay back the PV system quite a bit faster.
In CA, the "come out even" period seems to be about 5-7 years now (prices for both panels & inverters had dropped quite a bit in the past few years, and price of electricity went up)


* or go off-grid, altogether.. Not sure how it is in your state, but the best way to pay back for the PV install in California is to keep it tied to the grid -- you sell more power to the power company in summer, and buy it back in winter; at the end of the year, they even out the bill. If you want to go completely off the grid, you have to build a system that generates at least your full day's usage in winter (or actually even more power, in case of outages!). Going off-grid and having PV system "pay back for itself" are competing goals and can't both be met at this point. Note, it's not even "pays back for itself _quickly_" -- PV panels degrade every year (down to ~80% efficiency after 20 years) and inverters break down ~10yrs on average.

HTH

[DDF420]

What factors influence solar system energy yields?

The orientation and tilt of your installation
Whether there are shadows cast over your cells
The number of daylight hours (governed by where you live/season son on)
The intensity of the sunlight with differences between a 4KW system producing from 20KWh down to 14KWh a day governed by what part of the country you live in
The number of hours of full sun vs cloudy days


A wall-mounted, 7kWh lithium-ion battery storage system that comes with a 10 year warranty. The battery has a daily cycle, meaning it's designed to charge and discharge each day. The efficiency of the battery is 92%, so although it has a 7kWh capacity, it is more like 6.4kWh. Tesla also has a 10kWh weekly cycle version intended for back-up applications, but it's the 7kWh version you'll see in most home installations.

How much does it cost? a 4kW solar panel system and Powerwall battery with a cheaper Fronius hybrid inverter goes for around $13,990 fully installed.

So say you are living in the desert, with full sun intensity, and nothing but clear days all year round and created/saved/ 20KWh a day or 600x 26cents x 12months only $1872 a year saving . Thats sounds impressive? well that depends. Although id be paying 26 cents per KWh,rather high to most parts of the world, being such a high power user, id also be receiving a 15% discount from my supplier, so that now becomes $1592 or about 9 years to break even. By which time the battery warranty has expired and you likely need fork out again to buy another battery. Just not worth it.

[Falcon1]

Solar panels produce low voltage direct current. Creating alternating current high voltage and converting it back is just wasting power to heat which requires extra power to dissipate.

Why not remove all individual fans and go for large cooling block and mount large fans on the rack. Like this guy: https://simbimbo.wordpress.com/.

Perhaps switch to picopsu's. That might generate some savings already as well as reducing noise. Adding a additional pico PSU just for the GPU's might be an option.

[grigi]

Capital outlay is a big issue, changing power supplies to DC or installing batteries may be nice in an ideal world, but Michael here needs to keep capital outlay to a minimum.

I'd suggest a simple grid-connected inverter (they typically run at ~98% efficiency), and connect several panels to it. It doesn't try to act as a backup power supply at all, so your regular power supply still needs to be able to cope with your load. it just generates power when the sun is shining, and therefore reduce the amount you have to import from the grid.

No expensive batteries that need replacement. No expensive secondary power grid. No expensive project to replace all existing power supplies with DC-DC ones.

It really makes sense for Michael at a small scale, if you can offset up-to 2kwh for 5-9 hours a day (say, average of 6hr) for a fixed, once-off cost of a few thousand dollars, it could pay back within 2 years if you size it right. (note, don't go too big here, you want to guarantee that all the power it generates will get used locally for maximum TCO)

Depending on the legislation of your local municipality, you may have the option to sell back a little bit, on low-load cases, in which case you can size it slightly larger (but try to keep less than 15% of your power feeding back into the grid, else the TCO starts falling off a cliff again).

[alpha_one_x86]

For me you need reduce the power consumption, do continus benchmarking with very few x86, more with ARM. At least if you want that phoronix is sustainable.

[caligula]

No expensive batteries that need replacement. No expensive secondary power grid. No expensive project to replace all existing power supplies with DC-DC ones.

Um I think you have no idea what you're talking about. DC-DC converters are really cheap. For example a 95% efficient 1,5 kW converter from 8-60V to 12V is $16, including shipping.

It really makes sense for Michael at a small scale, if you can offset up-to 2kwh for 5-9 hours a day (say, average of 6hr) for a fixed, once-off cost of a few thousand dollars, it could pay back within 2 years if you size it right. (note, don't go too big here, you want to guarantee that all the power it generates will get used locally for maximum TCO)

2kWh of produced net power for few thousand dollars (apparently including all accessories). Ok, I see. 1 kW DIY panels cost $500 + taxes + customs duties. In reality they produce closer to 100W of stable net power. So a cheap 2kW DIY solution would cost at least $10 000, but apparently you can find one including all inverters for few thousand dollars.

[torsionbar28]

What factors influence solar system energy yields?

The orientation and tilt of your installation
Whether there are shadows cast over your cells
The number of daylight hours (governed by where you live/season son on)
The intensity of the sunlight with differences between a 4KW system producing from 20KWh down to 14KWh a day governed by what part of the country you live in
The number of hours of full sun vs cloudy days


A wall-mounted, 7kWh lithium-ion battery storage system that comes with a 10 year warranty. The battery has a daily cycle, meaning it's designed to charge and discharge each day. The efficiency of the battery is 92%, so although it has a 7kWh capacity, it is more like 6.4kWh. Tesla also has a 10kWh weekly cycle version intended for back-up applications, but it's the 7kWh version you'll see in most home installations.

How much does it cost? a 4kW solar panel system and Powerwall battery with a cheaper Fronius hybrid inverter goes for around $13,990 fully installed.

So say you are living in the desert, with full sun intensity, and nothing but clear days all year round and created/saved/ 20KWh a day or 600x 26cents x 12months only $1872 a year saving . Thats sounds impressive? well that depends. Although id be paying 26 cents per KWh,rather high to most parts of the world, being such a high power user, id also be receiving a 15% discount from my supplier, so that now becomes $1592 or about 9 years to break even. By which time the battery warranty has expired and you likely need fork out again to buy another battery. Just not worth it.

x2, solar doesn't make economic sense for a homeowner in the majority of cases. Not to mention the regular cleaning of the panels to maintain their output - most homes in the US have high angled roofs not easily (safely) accessible by the average homeowner, so that means paying someone to come clean them regularly.

The few people I know who have them, are all wealthy or upper upper middle class families who have large disposable income, and are doing it mainly as an enviro fashion statement, not because it makes any sort of economic sense.

The only places where solar is truly practical, are remote deserts and mountaintops where utility electric service simply isn't available, the cost of making it available far exceeds the costs of a solar installation.

[caligula]

x2, solar doesn't make economic sense for a homeowner in the majority of cases. Not to mention the regular cleaning of the panels to maintain their output - most homes in the US have high angled roofs not easily (safely) accessible by the average homeowner, so that means paying someone to come clean them regularly.

The few people I know who have them, are all wealthy or upper upper middle class families who have large disposable income, and are doing it mainly as an enviro fashion statement, not because it makes any sort of economic sense.

The only places where solar is truly practical, are remote deserts and mountaintops where utility electric service simply isn't available, the cost of making it available far exceeds the costs of a solar installation.

The funniest thing is, I know some 'greens' in Alaska and northern Europe advocate solar panels. Their power output is at least 66% less than the optimum due to the geographical location and also the snow and ice is bad for the panels. Apparently three times the cost isn't a problem if you believe in the right thing.

[starshipeleven]

I had the impression that not all benchmarks require huge servers. I have one 160W (200W peak) PicoPSU which should be sufficient for a non-overclocked system with ~70W GPU (from the bus) and ~70W CPU. Currently running a system with 250W GPU (geforce 980 ti) and 45W skylake CPU. It's powered by a Meanwell 12W power supply. You could add 3.5" HDDs as many as you like since they also use 12V directly.

I wanted to convert my own workstation actually, so yeah, I've got around 500 Watts of stuff at peak consumption that I'd like to be covered, possibly moar.
I was wondering about simply chaining multiple picoPSUs+bricks but something kept ringing alarms in my head.

The issue with disks is that they require a significant amount of watts on spinup (mostly on 12V as that's what the motor uses), like 15-20 watts per disk, and this limits a bit the number, unless your board features staggered spinup option (server motherboards do), or you do some hardware gimmick that keeps power cut off to groups of disks and then powers them up after a set time.