This subject is very high in my interest and activity recently. I am
trying to find more good models, data sources, so do please add
anything of quality. Vanderbauwhede's is a good account of device
composition, but as you can see it's a hugely more complex set of
inter-linked issues,
Here are some pop/accessible articles I wrote last year:
So this is not just about reducing cycles and network calls, but also about increasing the longevity of software and hardware. From a software point of view this means core quality attributes: performance, security, maintainability.
As a web developer I feel we _want_ to focus more on these things, as they are interesting and rewarding, but from a business perspective we typically see them as costly, secondary thresholds that we need to achieve, rather than primary targets.
Mechanical sympathy needs time to design, think, learn, measure. Stable, maintainable software needs testing, analysis, refactoring. Again, these are all things pretty much every intrinsically motivated programmer likes to do. But with everything related to resource distribution: People only start to pay for things when the external pressures are already there, or when its too late.
This is fascinating. We are barely keeping up with demand for computing with all those inefficient languages, architecture and difficulties of scale we face now. We waste cycles, bandwidth, memory and still do not have enough developer time to build or simply keep up with demand.
Just imagine how much more work is to be done when we face additional constraints of reuse of old devices, conserving energy, and device utilisation.
Additionally, as we slow down development of software to accommodate new constraints or simply cut usage of compute, we delay/hinder deployment of software which can help save our biosphere elsewhere. Or lift people from poverty. So this might not be justified after all.
This is complicated, we must work harder and save the planet. One thing is 8K tvs and datacenters to stream content, other is socially and environmentally important compute.
We aren't slowing down, we are going back to normal levels of development. For the last 20 years we've been rushing and agiling everything for the sake of money, sacrificing backwards compatibility and maintainability.
Reusing old devices is a given. A program works until an update breaks it. The hardware is perfectly suited for this - x86 has ridiculous back compat. The challenges you point out are completely self made and it's time we repay our technical debt.
As far as biosphere goes, it's like saying a slow dentist harms surgeons. What even is this mythical software which protects the biosphere and lifts people from poverty? SV is not contributing to such efforts, nor is any general SE corporation. We will just slow down and consume less software. Less, but better maintained.
If this follows the same trend as energy, I'm curious about what will happen when the energy and environmental impact has to be factored into the price of internet consumption.
At some points, bandwith will stop being "free", the price of electronics will grow, etc... Then it will reach a tipping point where the masses can not afford anymore what had become "essential".
Millenials and GenZ made lots of fun of boomers in yellow yest complaining about gas price.
I wonder what demonstrations will look like when we can't afford Netflix and Tiktok and Zoom any more.
For me, I'm old enough to remember a time where switching the modem on was a thoughfull decision, because of the phone bill.
But I'm afraid it's not going to make things any easier than it was for the yellow vests who had gone through one or two gas shocks in the 70s...
"...there isn’t any interest in this while it’s cheap"
I agree. This topic was discussed recently on a HN thread and I'll repeat what I said then.
Some of the most popular languages are the least performative. But if a program runs slowly, many developers will simply throw more hardware at the problem. The "hardware is cheap" attitude is widespread among developers and is frankly embarrassing.
Compare that to other industries that strive to improve energy efficiency. Imagine if a manufacturer said that they were going to make fridges/washing machines or other appliances without regard to energy-efficiency because "hardware is cheap". It sounds ridiculous.
A typed, compiled language can give you good performance and smaller memory footprint for free with the reduced computing resources that implies - no premature optimisation required.
Among developers, there is also a double-standard in attitudes and practice. Developers complain loudly when they are at the receiving end of slow or memory-hungry apps (or websites). But when it's their turn to build apps or sites, they will prioritise languages or technologies that suit their own ease and comfort above all else. The user-experience, performance, memory footprint - all take a back seat.
I find 'frugal' align very well with 'convival'. It stresses the relation of creator and tool and demands the tool to serve and not dominate or hide the task at hand but foster freedom and happiness.
The internet should operate like a sunflower, servers should follow the sun. Data centers should operate on solar energy, deploying themselves at the sunrise and shutting down at sunset in given location.
I like this a lot. Or even without a distributed system, just having a server deliver varying levels of compute/bandwidth based on the position of the Sun (a forum goes text-only at sunset, starts delivering images again at sunrise, does batch processing at high noon).
> As a society we need to start treating computational resources as finite and precious, to be utilised only when necessary, and as effectively as possible.
Considering that on AWS potentially every millisecond of compute time and every database write is billed separately, I think the financial incentives are already there to consider computational resources precious.
> Extending the life time is also the key action for datacentres
Currently e.g. AWS still is actively selling m4 type ec2 instances, which were introduced in 2015 and use Haswell generation CPUs from 2013. Heck, they seem to be even selling the ancient m1 type instances through their "previous generation instance" program[1].
I think it is safe to assume that hyperscalers are squeezing every bit of use they can from old hardware. Afaik their main motivation for new hardware is to optimize energy efficiency, so that sounds exactly what the article wants.
I think people underestimate the economic mechanisms needed to solve an issue like this. changing personal individual behavior and spouting platitudes works in disney movies, not in a society with 8-billion members
> I think the financial incentives are already there to consider computational resources precious.
I'd argue we ARE already working with computational resources as precious commodities. If you are in an area where it matters, you consider the logistics of acquiring, retaining, and utilizing all of your CPU and memory. It's precisely this that fed the rise of AWS.
So, do we need a new economy putting pricing on these resources? No! We already have it. We're just shocked at how cheap it is. But why shouldn't it be? You're messing with virtual realities where it's possible to teleport objects from one side of the universe to the other at almost no cost. Compare that with the friction of moving a real world object to the other side of your house.
Thermodynamics apply. I'm pretty tired of navel gazing about the efficiency of computing when it was invented as a mechanism to increase the efficiency of the CRITICALLY inefficient systems.
Your smartphone and its code are not the bottleneck in the world system. Stop optimizing the wrong parts.
AWS has incentives to reduce their footprint to a minimum, but they have an even stronger incentive to increase their consumers' resource consumption (because at the end of the day, that's what they are billing!)
> To address these challenges, action is needed on many fronts. What will you do to make frugal computing a reality?
Frustrating that the authors ends with such an "open note", without any practical "first steps". (This [1] at least _tries_ to give some options, as arguable as they are.)
There are obvious contenders:
* keep your phone longer. Brag about it. Find a persuasive way to _demand_ that apps, updates, etc... keep working. I don't know what difference 5 years vs 3 years do, though.
* make fun of people who change their phone all the time. (Or for that matter, make fun of consumerism _in general_.) Assuming peer pressure stands a chance against advertising industry.
* write efficient software ? (The joke is that anything that can be rewritten in Rust will be rewritten in Rust; but the jury is still out on whether the energy saved _running_ Rust code will offset the energy consumed _compiling_ it.)
* write no software ? (especially if you're in the cat picture and fake news sharing business ?)
> the jury is still out on whether the energy saved _running_ Rust code will offset the energy consumed _compiling_ it
Obviously depends on how much the compiled code is run. If it's a one-off script, no. If it's a server that will serve a million requests, or a browser that is shipped to a million users that will use the same version for a few months, a thousand times yes. And the vast majority of energy use that can be attributed to programs is in the second category.
You know, it feels like some of that advice isn't entirely viable, though i support it on some level.
> * keep your phone longer. Brag about it. Find a persuasive way to _demand_ that apps, updates, etc... keep working. I don't know what difference 5 years vs 3 years do, though.
> * make fun of people who change their phone all the time. (Or for that matter, make fun of consumerism _in general_.) Assuming peer pressure stands a chance against advertising industry.
Currently i have an ulePhone Armor X7 Pro, a phone that i got because it was both affordable and relatively rugged (i've dropped it once on concrete while doing pull-ups, it didn't even have a scratch on it) as well as water/dust proof. Despite the fact that the battery is not removable (something that's an entirely different can of worms) or in a form factor that's hard to get aftermarket replacements for, it's suitable for most of my needs - probably one of the least bad choices i could have made with these constraints in mind.
That said, the kernel is from 2020, the Android security patches are from 2020 and Android 10 will hit its own EOL in a few years anyways. Not only does the software have planned obsolescence baked into it by the OS going out of support eventually, not only is it abandoned by the developers who aren't actively shipping patches, but it will also not be possible to install custom ROMs because of either the bootloaders being locked by default (which sometimes don't even have ways around this) or the drivers being oddly specific and half of the phone's functionality breaking with those, which i've experienced even back in Android 2.X days.
Of course, phones aren't the only problematic bits of software here: old hardware is basically killed due to insecure OSes like Windows XP, Windows 7 and others for certain goals like gaming (and no, Linux is not and will not be suitable for that on older hardware for numerous titles) or other pieces of software that have become increasingly bloated nowadays and require better/modern hardware to run.
Who knows, maybe we'll eventually even see x86 be retired completely, it's not like all of our software is backwards compatible even as we speak, many focusing on just x64 and not even testing their software on older architectures. Though perhaps that only applies to decades, not a few years, as far as time scales are concerned.
> * write efficient software ? (The joke is that anything that can be rewritten in Rust will be rewritten in Rust; but the jury is still out on whether the energy saved _running_ Rust code will offset the energy consumed _compiling_ it.)
> * write no software ? (especially if you're in the cat picture and fake news sharing business ?)
Maybe we should also think about how to limit what our current software can do, to prevent it from eating all of the system resources for whatever reason (it being written badly, having unreasonable demands for hardware because "RAM that's not used is wasted" and so on)?
For example, currently i run Docker Swarm on my homelab with containers, each of which has a resource limitation in place: none can use up more than 1 core of my CPU which consequently means that nothing can typically bring down the entire system by misbehaving.
I wish we'd easily be able to do the same for desktop software on all OSes, just set affinity to N core(s) by default for software that we don't want to give too much processing power and allow it to request more. Same for memory. Same for disk space. I am okay with offending packages working slowly, hitting the swap space or being OOM killed (or even them requesting more resources), i'm not okay with some software package thinking its owns my machine and slowing everything down.
In summary, however, i'll try to make my hardware and software last: Debian (community LTS)/Ubuntu LTS/Rocky Linux are all suitable for now, even on my homelab boxes with used 200GEs wit...
I've found Nabil Hassein's "Computing, Climate Change, and all our Relationships" to be a very insightful take on this topic, much more so than the linked article:
Well, but making fun of consumerism usually seems to involve making fun of the forms of consumerism affordable by those with less money, e.g. fancy coffee from Starbucks.
I don't know if I'm unusual, or not... but I've never voluntarily upgraded my phone. I've always lost it or broke it, and been forced to buy a new one. My longest surviving phone was maybe three years.
I could certainly believe that I'm just more violent and forgetful than most...
I’ve found that writing mobile apps (especially native apps) is basically de facto “frugal computing.”
Power consumption is a big part of mobile app development.
I haven’t had the opportunity to write stuff that really sucks down a power load.
I have a pretty cool ML app that really sucks down the battery. It can drain 3% of my battery in about 5mins. The phone gets noticeably hot in my hand. I guess they are using their own ML engine, as opposed to the hardware-assisted stuff in the phone.
On the other hand, most games on my phone seem to run cool. I suspect that’s because they leverage Metal.
Exactly… the projection that datacenters' demand for computing power won't be offset by silicon efficiency improvements is somewhat speculative; the projection that Bitcoin's demand for computing power won't be offset by silicon efficiency improvements is obvious and is by design of the fucking system.
bitcoin mining provides an opportunity for anyone, anywhere to convert their computing resources into value. the competitive pressure is to have more efficient energy usage than your competitors
so in reality, bitcoin provides an direct incentive for more efficient computing (the kind that can't be handle with by building smaller, more efficient processors). it might do more to boost computational efficiency (and reduce emissions) than all the millions of pages of blogger platitude-ing flooding the internet
Absolute bullshit. Bitcoin does not even use processors for mining, it uses massively specialised hardware that is absolutely useless for anything else. No advances in bitcoin mining will translate to gain in anything else.
It also massively increases emissions and does nothing to decrease them. The difficulty adjustment algorithm of bitcoin will ensure that you always have to keep wasting more resources to mine, and will cancel out any gains you ever make. By design. Bitcoin is explicitly built to waste resources.
by specialized hardware, you mean GPUs? which are extremely useful for graphics, machine learning, and large-scale mathematical calculations? the gains to society that GPUs brought can't be understated. how do you think they run the climate-projection models? [1]
and my argument is about second-order effects. even if bitcoin mining remains inefficient, the technological gains of directly incentivising lower energy usage than your competitors carries over to other markets / enterprises
plus, bitcoin is scheduled to run out. so eventually the mining issue is moot
a second-order prediction is obviously more noisy and holds less weight without any direct evidence, but 'lack of direct evidence' characterizes much of this debate
That hasn't been the case for most of the decade. I already told you what does: Completely custom-built hardware that can do nothing but bitcoin mining, such as Antminers.
alright, in that case you're right, the processors seem wasteful
though interestingly, the specialized processors alkso seem like an example of a second order effect of bitcoin: driving innovation towards more energy efficiency. (though the tradeoff here seems to be wasting materials versus wasting energy)
I'm sorry, what? Bitcoin mining has spawned an entire new industry focused solely on converting a scarce shared resource (electrical energy) into perceived personal profits. It is wasteful pollution on a global scale: a single bitcoin transaction consumes as much power as an average U.S. household does in 77 days (https://digiconomist.net/bitcoin-energy-consumption)
It's a perfect example of the tragedy of the commons. The energy already wasted on bitcoin mining will never be recuperated by the efficiency gains you so eagerly (and without merit) ascribe to bitcoin.
So when bitcoin hits it's schedule max limit and can't be mined anymore, is the point moot?
I guess the question is: how much of a total energy bill will it have racked up before we get to that point, and should we consider the likely improvements to computing efficiency when we start speculating about the long term impact
"converting a scarce shared resource (electrical energy) into perceived personal profits" (1) is this entirely new?, and (2) it doesn't just exist for increasing profits, it provides a service: a digital cash infrastructure
The good thing about Bitcoin is that we can just ban it, since nobody actually uses it for anything important. It could be turned off tomorrow and nothing of value would be lost.
except all the goods and services that it currently stands in as a reference for (given people had to buy it with their own money, which they provided some service to obtain)
but hey,
just because bitcoin wasn't directly involved in a transaction doesn't mean it isn't standing in for economic value
if I steal an ancient coin worth $10,000 dollars from you, and you're upset about it, am I making a good rebuttal by saying: "it's not like you could buy anything at the grocery store with it anyways"
except all the goods and services that it currently stands in as a reference for (given people had to buy it with their own money, which they provided some service to obtain)
Harping on about how individuals can reduce energy usage in computing but not talking about Bitcoin is directly comparable to pushing lame individual recycling efforts but ignoring coal power plants and industrial scale methane leaks.
One aspect I didn't see much discussion on is about tradeoffs. Yes, computing demand is increasing, but does it all have a negative environmental balance?
Example - modern cars. They are full of electronics (there are around 80 ECUs in a modern car), but these electronics make the cars safer and lower the fuel consumption. Just look at how much the CO2 output for the same engine power decreased over time. Some of it comes from better mechanics like turbos or filters, but a lot of it comes from better electronics in the engine ECU.
Another example - logistics. Delivery companies are able to much better plan demand, routes and inventory thanks to big data processing in data centers. Yes, a data center consumes a lot of energy, but it also helps reduce emissions from trucks driving around less.
I don't have exact numbers, but I think everybody can imagine how increased computing in some areas (not all) improves the environmental impact. These are things that need to be considered in order to avoid unintended consequences.
In the case of cars, according to this [1] (seemingly based on US Federal data), the improvements in car efficiency since the 60s, is:
> Noteworthy fuel-economy trends, taking into account the length of time represented:
> A minor decrease between 1966 and 1973 (from 13.5 mpg to 12.9 mpg).
> A modest increase between 1973 and 1991 (from 12.9 mpg to 19.6 mpg).
> No change between 1991 and 2004 (19.6 mpg for both years).
> A modest increase between 2004 and 2008 (from 19.6 mpg to 21.8 mpg).
> A minor increase between 2008 and 2017 (from 21.8 mpg to 22.3 mpg).
The term "modest" is probabling misleading, but looking at the graph in the linked site is more telling: basically, efficienty increased after the first gas shock, and "more or less" stalled in the 2000s.
So maybe the addition of electronics in the 70/80s had an impact in consumption, but I would argue it's less obvious for the additions in the 90s/2000s/2010s - although it would take a bit longer to show as the older cars are still on the road.
Impact on safety, though, I don't know.
As for the CO2 emission, I have a hard time taking any number at face value, knowing that, ironically, some of the electronings on board can be designed to cheat emision tests [2] - but that's probably unfair.
Electronic fuel injection definitely made cars more frugal back in the 90s. Also improved aerodynamics helped a bit. On the other hand, due to crash test requirements cars are now significantly heavier (I guess around 20%), people are also driving faster and enjoy more power and dynamics from the vehicle, which not only wastes fuel, but also leads to the use of wider tires, which are more expensive to produce and impact fuel economy negatively.
this why I don't trust platitude-ing about the environment that isn't back by economics. real solutions require thinking of the world as a complex, dynamic system, where influencing one variable can have unintended consequences on another.
> To address these challenges, action is needed on many fronts. What will you do to make frugal computing a reality?
I dunno, isn't this the same thing as blaming climate change on the people who take a single vacation by plane per year, no matter what else they're doing the other 51 weeks to try to live a non-wasteful life?
I like my home server and having my data here, on LAN speed. Of course that's probably not the ideal thing from a power usage perspective. Does it offset the fact that I am hardly using a car? Am I allowed to use up the old lightbulbs I still have or is it really better to go out and buy LEDs for everything? Should I replace my hardware all the time because it uses less power or keep the old stuff that is just fine? What can I do if the damn phone manufacturers won't ship updated after 2 years when I'd gladly use it for 4 years? Must I buy Apple now?
Sure, everyone has their own 'I will not give in' topics, but more and more I am getting disillusioned that I personally can make any impact.
There is no point in trying to fight climate change as an individual. The scale of pollution by industry is so mind-bogglingly massive that there is nothing that we can do to make a difference. We could all turn off every device, never buy anything new, and go back to eating foraged berries while living in a cave and climate change would still happen.
To give recent example of how industry works with no regard whatsoever to the climate, the German airline Lufthansa flew 21,000 entirely empty flights during periods of lockdown so that airports didn't change their schedules[1].
We are at the point now that unless industries change we should really start thinking about how we live with climate change, not how we stop it.
> We could all turn off every device, never buy anything new, and go back to eating foraged berries while living in a cave and climate change would still happen.
If consumers disappeared then its not like industry would keep on churning pollution to no end.
Temporarily. Because they were expecting the consumers to return, which was in this case reasonable expectation. Do you really think they would continue to do so indefinitely?
It'll always be a series of temporary things though. Companies will always use whatever makes a convenient, reasonable-sounding excuse to carry on doing massively stupid stuff that kills the planet if they can. They expend huge amounts of energy thinking up reasons to not do good things. It's not even for the sake of profit, because a lot of the time these things don't make them any money. They just lurch from one thing to another in order to put off a dip in the share price because shareholders are terrified of losing any money (plus people working in the business have taken a big chunk of their salary in shares, which incentiveizes maintaining a higher stock price...)
So sure, they're not doing whatever they're doing today permanently. They're doing a series of temporary things forever. Because that's different!
> I dunno, isn't this the same thing as blaming climate change on the people who take a single vacation by plane per year
It could be, yep. And it could be accurate. Although that behaviour is probably what a lot of us consider normal, an estimated[2] 80% of the world's population have not travelled by plane at all (not once per year -- ever!)
(travel and travel by air is not intrinsically the problem - it's the negative externalities of doing so. but it's interesting how normalized a damaging behaviour can become simply based on the behaviours of people around us)
> I dunno, isn't this the same thing as blaming climate change on the people who take a single vacation by plane per year, no matter what else they're doing the other 51 weeks to try to live a non-wasteful life?
It’s the opposite. The article is targeting people in the software industry. Their actions are multiplied by every instance of their software being used.
So it’s more like asking people who design aircraft what they will do about the climate crisis — not the individual consumer who buys a plane ticket.
Abstinence-only climate education is such a depressing notion. There is plenty of clean energy. It is not running out. The idea that we should impoverish ourselves to protect against a problem we can solve without impoverishing ourselves is unjustifiably self-destructive.
It's not necessarily impoverishing and depressing. It can be, especially if enforced from outside, but many people find satisfaction in minimalism and frugality.
Potential, yes. We just don't have the infrastructure to harness it at the moment, nor will we have it in the close future.
In the meantime we're running on dirty energy. (Source: IEA)
Yes, it's depressing.
> It is not running out.
Unless you have a technological breakthrough to announce, the problem is not so much "clean energy running out" as "clean energy not being there yet, and dirty energy either running out or our time with a bearable climate running out."
Yes, it's depressing.
> Abstinence-only climate education is such a depressing notion.
On board with that. But this is about moderation, not abstinence.
"Wear condoms, drink sensibly, and don't change your phone every year."
Yep, it's depressing.
Except the alternative is more depressing to me. But sure, if we cure aids, vaccinate against hangovers, and perfect fusion...
It would be better if people with broken crystal balls stopped insisting on business as usual on the middle of revolutions.
We are not on the "there's no reason renewables couldn't replace fossil fuels, we just have to invest a little in R&D" anymore. We are at the "yep, we replaced all of those on your left yesterday, today we are doing the next line" stage, and we are going through it fast.
The rate at which electricity costs are decreasing from solar is pretty damn staggering. So much so that government estimates have been laughably wrong for decades now (they consistently over-estimate future costs and under-estimate utilization), without much sign of change (ex: https://maartensteinbuch.com/2017/06/12/photovoltaic-growth-...)
Which doesn't in any way change the fact that we're still nowhere close to having a comprehensive solution to climate change, but it does mean that I find the "energy austerity" side misguided at best. In the same way that economic austerity is also usually misguided.
I do not read it as "we replaced all of those on your left yesterday" ; but rather as "we have moslty kept those on your left, but some of the new ones we're building are a not as dirty, and anyway last year was weird."
If you look at the whole "electricy mix vs energy mix", I read it as, yes, sorry, even more of a challenge.
Now, "depressing" is not in anyway "desperate". Of course the grids are improving in rich countries, the mixes are improving in rich countries, the batteries are popping here and there that will make intermittency more manageable in rich countries, etc... And that's a challenge !
But I insist on "in the close future".
There's more than 60% of our electricity, and more than 80% or our energy that we need to transition.
I understand we have 10/20 years to do so, not 50.
Then again, I'm bad at predictions. Maybe we're at the start of an exponential curve - I've spent the last two years missing them.
Do you believe coal plants are going to close "viraly" in the next few years ? (Not only new hydro/wind/solar popping up)
Do you see a wave of gasoline cars getting out of the road ? (Not only new EVs poping up ?)
A patient goes to the doctor.
"You drink two bottles of scotch per week, that's too much !"
A month later, the patient comes back.
"I made progress doctor ! Now I drink half a glass of orange juice, and three bottles of scotch."
> "we have moslty kept those on your left, but some of the new ones we're building are a not as dirty, and anyway last year was weird."
Where do you see that? Coal is falling about as quickly as the fastest rate it ever grew, natural gas is falling (I didn't expect that one 2 years ago). The only fossil fuel growing is oil, very slowly and from a low initial level.
And the growth on renewables continues on an exponential curve. At some point the exponential must stop, but here we are, on the double digits of market share and still exponential. (Click on the "Relative Change" box if you can't see it on the main one.)
Anyway, we have more than 10 years if we start strong and right away. The more we do right now, the more time we have to the hardest problems. And once we have electricity surplus (solar will almost certainly give us that), we can extend it indefinitely.
> Do you believe coal plants are going to close "viraly" in the next few years ?
They have been doing that for the last 3 years.
> Do you see a wave of gasoline cars getting out of the road?
That will take time. I see a wave of diesel trucks getting out of the road, but not cars.
About your edit, yes, if you restrict your time frame and geographical area, your data gets noisier.
From the analysis there, it looks like the US suffered an intense drought last year and replaced the lost hydropower with coal. That probably pushed the global total up, but it's not a given that it has increased overall.
But anyway, yes, let's keep replacing them. The faster the better.
This take seems to assume that clean electricity is the main issue here. But the opening of the article states that the production of computing devices is the main issue rather than their operation. Manufacturing is not just about clean electricity.
No energy is clean. Some sources are dirtier than others, but they all have an enormous footprint.
The only clean energy is the one you don't consume.
Also, there's no reason why frugality should lead to impoverishment, in fact what drives energy consumption the most isn't by far what benefit the user the most: tracking, ads, video as the default medium (because ads!), spam, planned hardware obsolescence, etc.
I made the most efficient and scalable HTTP app server with integrated distributed JSON database to replace all serverside software I was depending on:
The incentives are still there for churning out more and more code, compile it on more and more CI/CD pipelines, run it on clusters of more and more powerful machines, and interface it to more and more powerful computers in our pockets which have gigabytes of RAM, terabytes of storage, mad gigahertz multi-core CPUs but somehow still amount to being slow dumb terminals.
Damn, even open source world favors constant churn. New libraries come with breakage that demands every application out there to adapt with little to no reward, those applications aren’t becoming any better, they just keep chasing new languages, new frameworks, new windowing systems, re-solve problems solved long ago, and all of it is just to keep up.
Gosh, there was this comparison on how long would it take to cold boot a machine, open a document in a text processor, and print it out. The contenders were a 8-bit Commodore with floppies and a modern MacBook Pro. Guess what, while Mac won, it wasn’t by a wide margin at all. The Mac could likely easily emulate a dozen Commodores. Where are all the gains?
If I were to become a dictator of a large part of the world, I’d tax heavily all the extra cycles, all extra watts of power wasted, all extra RAM, and all excess bytes sent over the wire, and tax it enough to eat into profits of big tech. Incentives done right would make the developers think of efficiency from day one.
There’s also this thing that while we rely on tech more and more, we are just as far from actually automating the boring stuff away as we have been for the last 40 years. The stories about a plucky worker automating their job and not telling anyone are just as appealing today as they were two decades ago. And all this tech has bugs, is fragile, sucks big time and brings in a lot of frustration.
(Steps away from the soapbox and walks into the sunset.)
"If I were to become a dictator of a large part of the world, I’d tax heavily all the extra cycles, all extra watts of power wasted, all extra RAM, and all excess bytes sent over the wire, and tax it enough to eat into profits of big tech. Incentives done right would make the developers think of efficiency from day one"
The best part about this is all the current first world countries could deal with that tax increase, and all the developing countries can't. It's a great way to ensure the technology gains of the world don't go to any of the countries that need them
"we are just as far from actually automating the boring stuff away as we have been for the last 40 years."
that's quite a statement. millions of white papers on new technology and billions (probably trillions) of dollars spent on automation moved us zero steps in the direction we've been aiming for?
> The best part about this is all the current first world countries could deal with that tax increase, and all the developing countries can't.
Well, if your way of dealing with such tax would be to outsource all the "dirty" bits to developing countries where they just don't care, or be like Germany, claiming of being oh so green while buying dirty electricity made from coal from Poland and Czechia, sorry, can't stand behind this.
> dollars spent on automation moved us zero steps in the direction we've been aiming for?
Things do get automated in, say, factories, but not in the everyday life.
When it comes to dealing with some governments who jumped onto this newfangled digitalization bandwagon, it looks like they have taken all the shit parts from the online processes, all the shit parts from the old paper-based processes and jumbled them together into the most abominable, inefficient form ever, while pretending to be so 21st century.
Another thing is the increasing siloization of platforms and APIs moving like quicksand, so if you ever try to integrate stuff to automate some of your workflows, you get into this place where there's constant churn.
And don't get me started on home automation, the only things the existing home automation solutions really automate are frustration and despair.
>When it comes to dealing with some governments who jumped onto this newfangled digitalization bandwagon, it looks like they have taken all the shit parts from the online processes, all the shit parts from the old paper-based processes and jumbled them together into the most abominable, inefficient form ever, while pretending to be so 21st century.
This isn't the case in many European countries. So many things are super easy to do in Sweden digitally. Denmark was also pretty good, one thing I liked about Denmark's system was how easy it was to get a hold of a real human.
>And don't get me started on home automation, the only things the existing home automation solutions really automate are frustration and despair.
The real home automation solutions are the washer, drier, and dish washer. Maybe also the robot vacuum (I've heard good things but never used one myself). These are all excellent, though I don't use my drier in favour of a rack.
Mostly computing has become a "thing" and not a tool.
I've noticed this at "work", network failed one day, no more printer; elders told me they had paper templates to fill in case computers don't work. It took 30s to fill it while listening to the person[0]. It takes 3 good minutes on a computer.
I guess it's how cycle goes.. things go obese until they have to crash and be redesigned.
ps: Not only this but the process is hyper redundant, and humanly heavier .. on paper I listen to the person, fill as data comes, that person doesn't wait alone, while I wait for the app to respond and have to input the same data 3 times in different forms (data that can be inferred from that person's case id). We all make more efforts, waste more time.
[0] similar to the abacus vs pocket calculator contest done long ago
Are you serious? You're using opening a text document as the criteria here? So the fact that the MacBook Pro can run code and simulations that would've taken an unimaginable amount of energy and effort in the commodore days is moot to you?
Barring Bitcoin, pretty much every technology has made processing of information and data far, far more efficient. At the end of the day, it doesn't matter if the United Airlines app is 200 MB larger than it should be, if it's saving the company millions of dollars and saving everybody else energy.
I have at least totally given up on high-end personal computing (think your top-notch Alienware computers) and have been living with a 470S that I purchased from the company as X-hand a while ago.
I'll probably stay frugal on computing power for the rest of my life, unless forced to change.
Was you reason for this change purely because of the environmental impacts? I know a decent amount of people who are concerned for the environment, but tend to buy high end PCs, mainly for gaming.
I'd imagine corporations and smartphones are mostly responsible. Phones seem to last only about 3 years. My company has probably upgraded my workstation/laptop 5 times in the 10 years I've worked there. I guess a pi and audrino are the newest computers I have. But my newest typical/traditional computer is 10 years old, and my personal laptop is 14.
What is SpaceX's starship good for? Well, what does space have? Solar power in gobs.
A lot of the economic barriers for space based stuff is just moving stuff around and getting it back to earth. Well, for computation you don't have to do that. You just beam the answers back.
Starlink has the networking substrate already. With the end of exponential churn in hardware, soon orbiting computing can have a longer lifecycle.
Starship is targetting 10$/kilogram. You can fit a computer in a kg easily. The launch cost will be a fraction of the procurement and other lifetime costs of the computer.
I guess cooling is the other issue. I don't know much about cooling in space. I know it is REALLY HOT on the sun side, and REALLY COLD on the other, but radiating heat usually involves other molecules touching the hot thing. Maybe you put reflective paint on the computing unit to keep the sun out, and the cold side chills the rest of the computer to near-absolute zero? Again, I am clueless about space heat dissipation.
But if you can ALSO get really cold temperatures, you can also overclock the CPUs to run faster and get more compute.
Apparently the space telescopes have liquid helium cooling, and there are .1 Kelvin pools in space for astronomical detectors.
I guess you can use pools of onboard liquid as heat sinks if the radiative surface is insufficient at peak, and then radiate the rest from the heat sink pool when it is, say, behind the earth or not working as hard.
98 comments
[ 4.8 ms ] story [ 159 ms ] threadHere are some pop/accessible articles I wrote last year:
http://techrights.org/2022/01/02/article-on-digital-rights-i...
https://cheapskatesguide.org/articles/techno-clutter-farnell...
https://cheapskatesguide.org/articles/uncharitable.html
See also Gerry McGovern's "World Wide Waste 2021" which links to useful papers and sources of analysis
Katie Singer's article:
https://www.ourwebofinconvenienttruths.com/letter-3/
This by Alba Ardura Gutiérrez:
https://cvbj.biz/we-eat-fish-contaminated-by-the-electronic-...
As a web developer I feel we _want_ to focus more on these things, as they are interesting and rewarding, but from a business perspective we typically see them as costly, secondary thresholds that we need to achieve, rather than primary targets.
Mechanical sympathy needs time to design, think, learn, measure. Stable, maintainable software needs testing, analysis, refactoring. Again, these are all things pretty much every intrinsically motivated programmer likes to do. But with everything related to resource distribution: People only start to pay for things when the external pressures are already there, or when its too late.
Just imagine how much more work is to be done when we face additional constraints of reuse of old devices, conserving energy, and device utilisation.
Additionally, as we slow down development of software to accommodate new constraints or simply cut usage of compute, we delay/hinder deployment of software which can help save our biosphere elsewhere. Or lift people from poverty. So this might not be justified after all.
This is complicated, we must work harder and save the planet. One thing is 8K tvs and datacenters to stream content, other is socially and environmentally important compute.
2.) Make it right.
3.) Make it fast.
While some individual software is at step 3, as a society overall we have much work ahead of us at step 1.
Reusing old devices is a given. A program works until an update breaks it. The hardware is perfectly suited for this - x86 has ridiculous back compat. The challenges you point out are completely self made and it's time we repay our technical debt.
As far as biosphere goes, it's like saying a slow dentist harms surgeons. What even is this mythical software which protects the biosphere and lifts people from poverty? SV is not contributing to such efforts, nor is any general SE corporation. We will just slow down and consume less software. Less, but better maintained.
Software that helps people to organise for a political cause mainly. So, communication is the first example.
Software that helps scientific research of ecology also directly impacts our ability to fix things.
More software is impactful in indirect ways.
At some points, bandwith will stop being "free", the price of electronics will grow, etc... Then it will reach a tipping point where the masses can not afford anymore what had become "essential".
Millenials and GenZ made lots of fun of boomers in yellow yest complaining about gas price.
I wonder what demonstrations will look like when we can't afford Netflix and Tiktok and Zoom any more.
For me, I'm old enough to remember a time where switching the modem on was a thoughfull decision, because of the phone bill.
But I'm afraid it's not going to make things any easier than it was for the yellow vests who had gone through one or two gas shocks in the 70s...
I agree. This topic was discussed recently on a HN thread and I'll repeat what I said then.
Some of the most popular languages are the least performative. But if a program runs slowly, many developers will simply throw more hardware at the problem. The "hardware is cheap" attitude is widespread among developers and is frankly embarrassing.
Compare that to other industries that strive to improve energy efficiency. Imagine if a manufacturer said that they were going to make fridges/washing machines or other appliances without regard to energy-efficiency because "hardware is cheap". It sounds ridiculous.
A typed, compiled language can give you good performance and smaller memory footprint for free with the reduced computing resources that implies - no premature optimisation required.
Among developers, there is also a double-standard in attitudes and practice. Developers complain loudly when they are at the receiving end of slow or memory-hungry apps (or websites). But when it's their turn to build apps or sites, they will prioritise languages or technologies that suit their own ease and comfort above all else. The user-experience, performance, memory footprint - all take a back seat.
Also of interest: Permacomputing https://news.ycombinator.com/item?id=27199225
Know https://solar.lowtechmagazine.com/2020/01/how-sustainable-is...?
Considering that on AWS potentially every millisecond of compute time and every database write is billed separately, I think the financial incentives are already there to consider computational resources precious.
> Extending the life time is also the key action for datacentres
Currently e.g. AWS still is actively selling m4 type ec2 instances, which were introduced in 2015 and use Haswell generation CPUs from 2013. Heck, they seem to be even selling the ancient m1 type instances through their "previous generation instance" program[1].
I think it is safe to assume that hyperscalers are squeezing every bit of use they can from old hardware. Afaik their main motivation for new hardware is to optimize energy efficiency, so that sounds exactly what the article wants.
https://aws.amazon.com/ec2/previous-generation/
I'd argue we ARE already working with computational resources as precious commodities. If you are in an area where it matters, you consider the logistics of acquiring, retaining, and utilizing all of your CPU and memory. It's precisely this that fed the rise of AWS.
So, do we need a new economy putting pricing on these resources? No! We already have it. We're just shocked at how cheap it is. But why shouldn't it be? You're messing with virtual realities where it's possible to teleport objects from one side of the universe to the other at almost no cost. Compare that with the friction of moving a real world object to the other side of your house.
Thermodynamics apply. I'm pretty tired of navel gazing about the efficiency of computing when it was invented as a mechanism to increase the efficiency of the CRITICALLY inefficient systems.
Your smartphone and its code are not the bottleneck in the world system. Stop optimizing the wrong parts.
Frustrating that the authors ends with such an "open note", without any practical "first steps". (This [1] at least _tries_ to give some options, as arguable as they are.)
There are obvious contenders:
* keep your phone longer. Brag about it. Find a persuasive way to _demand_ that apps, updates, etc... keep working. I don't know what difference 5 years vs 3 years do, though.
* make fun of people who change their phone all the time. (Or for that matter, make fun of consumerism _in general_.) Assuming peer pressure stands a chance against advertising industry.
* write efficient software ? (The joke is that anything that can be rewritten in Rust will be rewritten in Rust; but the jury is still out on whether the energy saved _running_ Rust code will offset the energy consumed _compiling_ it.)
* write no software ? (especially if you're in the cat picture and fake news sharing business ?)
[1] http://worrydream.com/ClimateChange/
Obviously depends on how much the compiled code is run. If it's a one-off script, no. If it's a server that will serve a million requests, or a browser that is shipped to a million users that will use the same version for a few months, a thousand times yes. And the vast majority of energy use that can be attributed to programs is in the second category.
> * keep your phone longer. Brag about it. Find a persuasive way to _demand_ that apps, updates, etc... keep working. I don't know what difference 5 years vs 3 years do, though.
> * make fun of people who change their phone all the time. (Or for that matter, make fun of consumerism _in general_.) Assuming peer pressure stands a chance against advertising industry.
Currently i have an ulePhone Armor X7 Pro, a phone that i got because it was both affordable and relatively rugged (i've dropped it once on concrete while doing pull-ups, it didn't even have a scratch on it) as well as water/dust proof. Despite the fact that the battery is not removable (something that's an entirely different can of worms) or in a form factor that's hard to get aftermarket replacements for, it's suitable for most of my needs - probably one of the least bad choices i could have made with these constraints in mind.
That said, the kernel is from 2020, the Android security patches are from 2020 and Android 10 will hit its own EOL in a few years anyways. Not only does the software have planned obsolescence baked into it by the OS going out of support eventually, not only is it abandoned by the developers who aren't actively shipping patches, but it will also not be possible to install custom ROMs because of either the bootloaders being locked by default (which sometimes don't even have ways around this) or the drivers being oddly specific and half of the phone's functionality breaking with those, which i've experienced even back in Android 2.X days.
Of course, phones aren't the only problematic bits of software here: old hardware is basically killed due to insecure OSes like Windows XP, Windows 7 and others for certain goals like gaming (and no, Linux is not and will not be suitable for that on older hardware for numerous titles) or other pieces of software that have become increasingly bloated nowadays and require better/modern hardware to run.
Who knows, maybe we'll eventually even see x86 be retired completely, it's not like all of our software is backwards compatible even as we speak, many focusing on just x64 and not even testing their software on older architectures. Though perhaps that only applies to decades, not a few years, as far as time scales are concerned.
> * write efficient software ? (The joke is that anything that can be rewritten in Rust will be rewritten in Rust; but the jury is still out on whether the energy saved _running_ Rust code will offset the energy consumed _compiling_ it.)
> * write no software ? (especially if you're in the cat picture and fake news sharing business ?)
Maybe we should also think about how to limit what our current software can do, to prevent it from eating all of the system resources for whatever reason (it being written badly, having unreasonable demands for hardware because "RAM that's not used is wasted" and so on)?
For example, currently i run Docker Swarm on my homelab with containers, each of which has a resource limitation in place: none can use up more than 1 core of my CPU which consequently means that nothing can typically bring down the entire system by misbehaving.
I wish we'd easily be able to do the same for desktop software on all OSes, just set affinity to N core(s) by default for software that we don't want to give too much processing power and allow it to request more. Same for memory. Same for disk space. I am okay with offending packages working slowly, hitting the swap space or being OOM killed (or even them requesting more resources), i'm not okay with some software package thinking its owns my machine and slowing everything down.
In summary, however, i'll try to make my hardware and software last: Debian (community LTS)/Ubuntu LTS/Rocky Linux are all suitable for now, even on my homelab boxes with used 200GEs wit...
I've found Nabil Hassein's "Computing, Climate Change, and all our Relationships" to be a very insightful take on this topic, much more so than the linked article:
https://www.deconstructconf.com/2018/nabil-hassein-computing...
I could certainly believe that I'm just more violent and forgetful than most...
Power consumption is a big part of mobile app development.
I haven’t had the opportunity to write stuff that really sucks down a power load.
I have a pretty cool ML app that really sucks down the battery. It can drain 3% of my battery in about 5mins. The phone gets noticeably hot in my hand. I guess they are using their own ML engine, as opposed to the hardware-assisted stuff in the phone.
On the other hand, most games on my phone seem to run cool. I suspect that’s because they leverage Metal.
so in reality, bitcoin provides an direct incentive for more efficient computing (the kind that can't be handle with by building smaller, more efficient processors). it might do more to boost computational efficiency (and reduce emissions) than all the millions of pages of blogger platitude-ing flooding the internet
It also massively increases emissions and does nothing to decrease them. The difficulty adjustment algorithm of bitcoin will ensure that you always have to keep wasting more resources to mine, and will cancel out any gains you ever make. By design. Bitcoin is explicitly built to waste resources.
and my argument is about second-order effects. even if bitcoin mining remains inefficient, the technological gains of directly incentivising lower energy usage than your competitors carries over to other markets / enterprises
plus, bitcoin is scheduled to run out. so eventually the mining issue is moot
a second-order prediction is obviously more noisy and holds less weight without any direct evidence, but 'lack of direct evidence' characterizes much of this debate
[1]: https://blogs.nvidia.com/blog/2019/07/17/clima-climate-model...
For instance, here is a shipping container that burns 1 megawatt of power in a pure and complete waste of energy and carbon emissions: https://shop.bitmain.com/product/detail?pid=0002021123118285...
though interestingly, the specialized processors alkso seem like an example of a second order effect of bitcoin: driving innovation towards more energy efficiency. (though the tradeoff here seems to be wasting materials versus wasting energy)
It's a perfect example of the tragedy of the commons. The energy already wasted on bitcoin mining will never be recuperated by the efficiency gains you so eagerly (and without merit) ascribe to bitcoin.
I guess the question is: how much of a total energy bill will it have racked up before we get to that point, and should we consider the likely improvements to computing efficiency when we start speculating about the long term impact
"converting a scarce shared resource (electrical energy) into perceived personal profits" (1) is this entirely new?, and (2) it doesn't just exist for increasing profits, it provides a service: a digital cash infrastructure
if I steal an ancient coin worth $10,000 dollars from you, and you're upset about it, am I making a good rebuttal by saying: "it's not like you could buy anything at the grocery store with it anyways"
That is bringing actual value to society. We can get rid of it and it will be a net positive.
Example - modern cars. They are full of electronics (there are around 80 ECUs in a modern car), but these electronics make the cars safer and lower the fuel consumption. Just look at how much the CO2 output for the same engine power decreased over time. Some of it comes from better mechanics like turbos or filters, but a lot of it comes from better electronics in the engine ECU.
Another example - logistics. Delivery companies are able to much better plan demand, routes and inventory thanks to big data processing in data centers. Yes, a data center consumes a lot of energy, but it also helps reduce emissions from trucks driving around less.
I don't have exact numbers, but I think everybody can imagine how increased computing in some areas (not all) improves the environmental impact. These are things that need to be considered in order to avoid unintended consequences.
> Noteworthy fuel-economy trends, taking into account the length of time represented:
> A minor decrease between 1966 and 1973 (from 13.5 mpg to 12.9 mpg).
> A modest increase between 1973 and 1991 (from 12.9 mpg to 19.6 mpg).
> No change between 1991 and 2004 (19.6 mpg for both years).
> A modest increase between 2004 and 2008 (from 19.6 mpg to 21.8 mpg).
> A minor increase between 2008 and 2017 (from 21.8 mpg to 22.3 mpg).
The term "modest" is probabling misleading, but looking at the graph in the linked site is more telling: basically, efficienty increased after the first gas shock, and "more or less" stalled in the 2000s.
So maybe the addition of electronics in the 70/80s had an impact in consumption, but I would argue it's less obvious for the additions in the 90s/2000s/2010s - although it would take a bit longer to show as the older cars are still on the road.
Impact on safety, though, I don't know.
As for the CO2 emission, I have a hard time taking any number at face value, knowing that, ironically, some of the electronings on board can be designed to cheat emision tests [2] - but that's probably unfair.
[1] https://www.greencarcongress.com/2019/09/20190930-sivak.html
[2] https://en.wikipedia.org/wiki/Volkswagen_emissions_scandal
https://m.youtube.com/watch?v=0PclgBd6_Zs
[1] Forth programming language: https://en.wikipedia.org/wiki/Forth_(programming_language)
I dunno, isn't this the same thing as blaming climate change on the people who take a single vacation by plane per year, no matter what else they're doing the other 51 weeks to try to live a non-wasteful life?
I like my home server and having my data here, on LAN speed. Of course that's probably not the ideal thing from a power usage perspective. Does it offset the fact that I am hardly using a car? Am I allowed to use up the old lightbulbs I still have or is it really better to go out and buy LEDs for everything? Should I replace my hardware all the time because it uses less power or keep the old stuff that is just fine? What can I do if the damn phone manufacturers won't ship updated after 2 years when I'd gladly use it for 4 years? Must I buy Apple now?
Sure, everyone has their own 'I will not give in' topics, but more and more I am getting disillusioned that I personally can make any impact.
To give recent example of how industry works with no regard whatsoever to the climate, the German airline Lufthansa flew 21,000 entirely empty flights during periods of lockdown so that airports didn't change their schedules[1].
We are at the point now that unless industries change we should really start thinking about how we live with climate change, not how we stop it.
[1] https://www.wired.co.uk/article/airplanes-empty-slots-covid
> We could all turn off every device, never buy anything new, and go back to eating foraged berries while living in a cave and climate change would still happen.
If consumers disappeared then its not like industry would keep on churning pollution to no end.
Airlines carried on flying without passengers.
So sure, they're not doing whatever they're doing today permanently. They're doing a series of temporary things forever. Because that's different!
It could be, yep. And it could be accurate. Although that behaviour is probably what a lot of us consider normal, an estimated[2] 80% of the world's population have not travelled by plane at all (not once per year -- ever!)
(travel and travel by air is not intrinsically the problem - it's the negative externalities of doing so. but it's interesting how normalized a damaging behaviour can become simply based on the behaviours of people around us)
[1] - https://ourworldindata.org/transport
It’s the opposite. The article is targeting people in the software industry. Their actions are multiplied by every instance of their software being used.
So it’s more like asking people who design aircraft what they will do about the climate crisis — not the individual consumer who buys a plane ticket.
How about not letting the state and alarmist NGOs make that decision for me?
- caloric restriction is very often cited as global health improvement
- I personally biked everywhere in the last months and I don't miss my car, my muscle mass has been restored and cardiac health too
- less performant computing also means less addictive behavior, deeper use potentially
Potential, yes. We just don't have the infrastructure to harness it at the moment, nor will we have it in the close future.
In the meantime we're running on dirty energy. (Source: IEA)
Yes, it's depressing.
> It is not running out.
Unless you have a technological breakthrough to announce, the problem is not so much "clean energy running out" as "clean energy not being there yet, and dirty energy either running out or our time with a bearable climate running out."
Yes, it's depressing.
> Abstinence-only climate education is such a depressing notion.
On board with that. But this is about moderation, not abstinence.
"Wear condoms, drink sensibly, and don't change your phone every year."
Yep, it's depressing.
Except the alternative is more depressing to me. But sure, if we cure aids, vaccinate against hangovers, and perfect fusion...
It would be better if people with broken crystal balls stopped insisting on business as usual on the middle of revolutions.
We are not on the "there's no reason renewables couldn't replace fossil fuels, we just have to invest a little in R&D" anymore. We are at the "yep, we replaced all of those on your left yesterday, today we are doing the next line" stage, and we are going through it fast.
You may like this graph:
https://ourworldindata.org/grapher/electricity-production-by...
The rate at which electricity costs are decreasing from solar is pretty damn staggering. So much so that government estimates have been laughably wrong for decades now (they consistently over-estimate future costs and under-estimate utilization), without much sign of change (ex: https://maartensteinbuch.com/2017/06/12/photovoltaic-growth-...)
Which doesn't in any way change the fact that we're still nowhere close to having a comprehensive solution to climate change, but it does mean that I find the "energy austerity" side misguided at best. In the same way that economic austerity is also usually misguided.
But looking at this:
https://ourworldindata.org/grapher/electricity-prod-source-s...
I do not read it as "we replaced all of those on your left yesterday" ; but rather as "we have moslty kept those on your left, but some of the new ones we're building are a not as dirty, and anyway last year was weird."
If you look at the whole "electricy mix vs energy mix", I read it as, yes, sorry, even more of a challenge.
https://ourworldindata.org/grapher/global-energy-substitutio...
Now, "depressing" is not in anyway "desperate". Of course the grids are improving in rich countries, the mixes are improving in rich countries, the batteries are popping here and there that will make intermittency more manageable in rich countries, etc... And that's a challenge !
But I insist on "in the close future".
There's more than 60% of our electricity, and more than 80% or our energy that we need to transition.
I understand we have 10/20 years to do so, not 50.
Then again, I'm bad at predictions. Maybe we're at the start of an exponential curve - I've spent the last two years missing them.
Do you believe coal plants are going to close "viraly" in the next few years ? (Not only new hydro/wind/solar popping up)
Do you see a wave of gasoline cars getting out of the road ? (Not only new EVs poping up ?)
[Edit]: Regarding the whole "we've replaced those on your left..." https://www.instituteforenergyresearch.org/fossil-fuels/coal...
Where do you see that? Coal is falling about as quickly as the fastest rate it ever grew, natural gas is falling (I didn't expect that one 2 years ago). The only fossil fuel growing is oil, very slowly and from a low initial level.
And the growth on renewables continues on an exponential curve. At some point the exponential must stop, but here we are, on the double digits of market share and still exponential. (Click on the "Relative Change" box if you can't see it on the main one.)
Anyway, we have more than 10 years if we start strong and right away. The more we do right now, the more time we have to the hardest problems. And once we have electricity surplus (solar will almost certainly give us that), we can extend it indefinitely.
> Do you believe coal plants are going to close "viraly" in the next few years ?
They have been doing that for the last 3 years.
> Do you see a wave of gasoline cars getting out of the road?
That will take time. I see a wave of diesel trucks getting out of the road, but not cars.
About your edit, yes, if you restrict your time frame and geographical area, your data gets noisier.
We'll have to wait for data about the global trend in 2021- I suspect we'll have a rebound effect after the pandemic.
That still leaves plenty to replace "on your left"... Let's do that, shall we ?
But anyway, yes, let's keep replacing them. The faster the better.
If that was true the problem of climate change wouldn't exist.
No energy is clean. Some sources are dirtier than others, but they all have an enormous footprint.
The only clean energy is the one you don't consume.
Also, there's no reason why frugality should lead to impoverishment, in fact what drives energy consumption the most isn't by far what benefit the user the most: tracking, ads, video as the default medium (because ads!), spam, planned hardware obsolescence, etc.
http://github.com/tinspin/rupy
If you want to be able to run production servers on Raspberry 2 & 4 this is today your only real alternative.
Now I'm building a 3D action MMO that runs on the Raspberry 4, good luck getting Unity or Unreal to work on that device!
The biggest waste of electricity is AAA gaming, both because it uses alot but also because it creates nothing of value.
Follow Minecraft (btw also a waste of energy) modding and Roblox and extrapolate, useful gaming allows you to OWN your work.
Damn, even open source world favors constant churn. New libraries come with breakage that demands every application out there to adapt with little to no reward, those applications aren’t becoming any better, they just keep chasing new languages, new frameworks, new windowing systems, re-solve problems solved long ago, and all of it is just to keep up.
Gosh, there was this comparison on how long would it take to cold boot a machine, open a document in a text processor, and print it out. The contenders were a 8-bit Commodore with floppies and a modern MacBook Pro. Guess what, while Mac won, it wasn’t by a wide margin at all. The Mac could likely easily emulate a dozen Commodores. Where are all the gains?
If I were to become a dictator of a large part of the world, I’d tax heavily all the extra cycles, all extra watts of power wasted, all extra RAM, and all excess bytes sent over the wire, and tax it enough to eat into profits of big tech. Incentives done right would make the developers think of efficiency from day one.
There’s also this thing that while we rely on tech more and more, we are just as far from actually automating the boring stuff away as we have been for the last 40 years. The stories about a plucky worker automating their job and not telling anyone are just as appealing today as they were two decades ago. And all this tech has bugs, is fragile, sucks big time and brings in a lot of frustration.
(Steps away from the soapbox and walks into the sunset.)
The best part about this is all the current first world countries could deal with that tax increase, and all the developing countries can't. It's a great way to ensure the technology gains of the world don't go to any of the countries that need them
"we are just as far from actually automating the boring stuff away as we have been for the last 40 years."
that's quite a statement. millions of white papers on new technology and billions (probably trillions) of dollars spent on automation moved us zero steps in the direction we've been aiming for?
Well, if your way of dealing with such tax would be to outsource all the "dirty" bits to developing countries where they just don't care, or be like Germany, claiming of being oh so green while buying dirty electricity made from coal from Poland and Czechia, sorry, can't stand behind this.
> dollars spent on automation moved us zero steps in the direction we've been aiming for?
Things do get automated in, say, factories, but not in the everyday life.
When it comes to dealing with some governments who jumped onto this newfangled digitalization bandwagon, it looks like they have taken all the shit parts from the online processes, all the shit parts from the old paper-based processes and jumbled them together into the most abominable, inefficient form ever, while pretending to be so 21st century.
Another thing is the increasing siloization of platforms and APIs moving like quicksand, so if you ever try to integrate stuff to automate some of your workflows, you get into this place where there's constant churn.
And don't get me started on home automation, the only things the existing home automation solutions really automate are frustration and despair.
This isn't the case in many European countries. So many things are super easy to do in Sweden digitally. Denmark was also pretty good, one thing I liked about Denmark's system was how easy it was to get a hold of a real human.
>And don't get me started on home automation, the only things the existing home automation solutions really automate are frustration and despair.
The real home automation solutions are the washer, drier, and dish washer. Maybe also the robot vacuum (I've heard good things but never used one myself). These are all excellent, though I don't use my drier in favour of a rack.
I've noticed this at "work", network failed one day, no more printer; elders told me they had paper templates to fill in case computers don't work. It took 30s to fill it while listening to the person[0]. It takes 3 good minutes on a computer.
I guess it's how cycle goes.. things go obese until they have to crash and be redesigned.
ps: Not only this but the process is hyper redundant, and humanly heavier .. on paper I listen to the person, fill as data comes, that person doesn't wait alone, while I wait for the app to respond and have to input the same data 3 times in different forms (data that can be inferred from that person's case id). We all make more efforts, waste more time.
[0] similar to the abacus vs pocket calculator contest done long ago
Are you serious? You're using opening a text document as the criteria here? So the fact that the MacBook Pro can run code and simulations that would've taken an unimaginable amount of energy and effort in the commodore days is moot to you?
Barring Bitcoin, pretty much every technology has made processing of information and data far, far more efficient. At the end of the day, it doesn't matter if the United Airlines app is 200 MB larger than it should be, if it's saving the company millions of dollars and saving everybody else energy.
I'll probably stay frugal on computing power for the rest of my life, unless forced to change.
What is SpaceX's starship good for? Well, what does space have? Solar power in gobs.
A lot of the economic barriers for space based stuff is just moving stuff around and getting it back to earth. Well, for computation you don't have to do that. You just beam the answers back.
Starlink has the networking substrate already. With the end of exponential churn in hardware, soon orbiting computing can have a longer lifecycle.
Starship is targetting 10$/kilogram. You can fit a computer in a kg easily. The launch cost will be a fraction of the procurement and other lifetime costs of the computer.
I guess cooling is the other issue. I don't know much about cooling in space. I know it is REALLY HOT on the sun side, and REALLY COLD on the other, but radiating heat usually involves other molecules touching the hot thing. Maybe you put reflective paint on the computing unit to keep the sun out, and the cold side chills the rest of the computer to near-absolute zero? Again, I am clueless about space heat dissipation.
But if you can ALSO get really cold temperatures, you can also overclock the CPUs to run faster and get more compute.
I guess you can use pools of onboard liquid as heat sinks if the radiative surface is insufficient at peak, and then radiate the rest from the heat sink pool when it is, say, behind the earth or not working as hard.
I hate Twitch streams, but Twitch assuming I am watching nets items in a game I play, and that Skinner box is strong.
At least I put the streams on extra low quality. But I recon Twitch streams running only for "drops" are about as bad for the climate as Bitcoin.
Yay 2022. Yay misguided advertisement or whatever the reason they do silly stuff like this.