h a l f b a k e r yPlease listen carefully, as our opinions have changed.
add, search, annotate, link, view, overview, recent, by name, best, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Lots of useful computer benchmarks out there, but none that apply to overall computer usage as a function of energy i.e. money.
Tweak the existing CPU, GPU, HD, IO and application benchmarks to accept a power factor, starting with an exponent of zero and rising as the sea levels do.
(Or in my case
just trying to get rid of the fans.)
((Will sign autograph books, but please no vocalising.))
Recognising artificial intelligence
Recognising_20artificial_20intelligence
shameless self plug.
My take on using the efficiency of a computer to develop AI [xaviergisz, Jul 10 2009]
1Gflop/W
http://www.thefreel...LOP+...-a0168820012
no exponent necessary [csea, Jul 12 2009]
[link]
|
| |
You do realise that gigaflopS per second is redundant, right? Unless you want a measure of Moore's Law against power consumption. |
|
| |
I do and I do. And no tricking me into marriage again ! |
|
| |
Sporadically, computer magazines have made comparisons of different CPUs by benchmarking, and then charted the benchmark results against the wattage of the CPU. |
|
| |
Problematic:
A) : different power-saving schemes in the CPUs mean that a CPU that uses very few power in normal use will max out during benchmarking, while an E-guzzler just maxes all the time.
B) : The 'real' speed of any processor depends heavily on the intended use, so the vector describing the powerfullness of a processor will have more dimensions than can be intuitivelky displayed on a screen/paper (but that's a general problem of benchmarking, not specific to your idea) |
|
| |
Point in case: The new Ion Nvidia chipset with the Atom 330 is a very considerate user of electrical energy, but i could not find a benchmarking that put it up against 'normal' CPU-motherboard combos, it was always compared to peers (nettops), with different benchmarks. |
|
| |
What's all this nonsense about exponents? Your basic idea: //computer usage as a function of energy// isn't overly flawed, once you define "useage"; why not start there?
Try Gflops/Joule. |
|
| |
I think what you're after is something like:
Bits per second per joule per unit area. |
|
| |
It's a very topical area in wireless (and wired) communication. |
|
| |
Nonsense about exponents - |
|
| |
B = GFlops / (Energy ^ Exponent) |
|
| |
Benchmarks could be calculated for exponents 0,1,2 etc. The average usage benchmarks would then be biassed towards different priorities of energy. |
|
| |
In other words, the amount of processing you get for the electricity you put in. Three flaws: |
|
| |
1) Most CPUs are nearly idle most of the time, but still draw significant power. The power drain of a netbook CPU running flat out is unimportant since it won't be used like that. |
|
| |
2) You can lower the clock frequency of a processor which will reduce the power requirements more than the loss of speed. Eg a half clocked CPU might need a quarter of the power. In general you do not actually get a fixed ratio of electrical power to computing power. You can choose between running efficiently and running fast. |
|
| |
3) It's just one component. I suspect AMD and Intel wil ultimately reduce idle power consumption to almost zero but the screen and RAM will still be constantly draining power. If you are playing games then the graphics card is probably drawing more power than the CPU although it could be subject to this benchmark. |
|
| |
Could be a useful measure for servers and number crunchers though. Power costs are starting to exceed hardware costs and it's no small matter for companies like Google. |
|
| |
I think this could also be applied to computer games. Game publishers could be required to put the computational requirements for their games on the back of the box at low and max settings, so if you knew how efficient your computer was you could figure out how much an hour playing your new game was costing. |
|
| |
It would be really nice if they started making games so you could run them on laptops and play for three hours without charging. Which would be optional so you could whack the detail up when you got home. |
|
| |
Good comments [Bad_Jim]. I completely overlooked the idle power, although I have seen it mentioned in some MB reviews. My box has mobile tech, but it still swallows 35 watts at idle. This doubles when thrashing the machine. It also makes the monitor at 80 watts the dominant energy hog, and that has a small annoying fan in it too. |
|
| |
Does this idea rise to an invention? |
|
| |
//Does this idea rise to an invention?// |
|
| |
Which version would you like - Eco Benchmarks 2009 for Linux or PC ? |
|
| |
//The new Ion Nvidia chipset with the Atom 330// |
|
| |
Those Zotac boards are a bit pricy for an Atom, and still 25W unloaded. Thats almost the same as my stock Intel board with not too shabby graphics and plenty of grunt to spare. One watt just the Atom CPU, what gives ? |
|
| |
[edit: units tidied as per annos. Thanks. Or maybe I should go with Gigaflopawatt ?] |
|
| |
Are you rating the running consumption of, eg a Zotac mini-
iTX board, or are you also including the manufacturing hit on
the environment too (including the toxicity and
unsustainability of the creation of the board-level SMD
components, etc)? |
|
| |
Not to mention the environmental impact of all the computers to crunch numbers and make artwork to flog the things. |
|
| |
Given that some products just fall flat, it would have a much lower environmental impact to stack the investment into a big pile of $100 bills and give it to me. I would donate 50% to charitable barbeques, 40% for my own heating and become a black market in origami with the rest. |
|
| |