If you're happy with OpenBSD, probably any computer is good enough
muezza.ca
muezza.ca
Here is what no one is mentioning. Currently, 90% of the world's most advanaced chips are made by one manufacturer, TSMC. If something hapens to TSMC, then the ability to utilise older chips and older computers is going to become relevant.
Personally, I think "modern" web browsers suck for using the web recreationally, e.g., visiting random websites, i.e., usage that is not banking, shopping or other commercially-related activity. However, when applied to local files, these "browsers" can be useful as PDF readers, audio and video players. Whether older computers can handle today's media files is left as a question for the reader.
Reading websites is easy without a big, bloated "tech" company browser. Using a text-only browser on a daily basis, I have been doing this for decades.
Just like the lifting of Nazi rocket scientists, it's a technology that is directly intertwined with the existence of contemporary advanced economies/societies, if TSMC ceases to exist there will be a concerted effort from major economies to re-gain that knowledge. Right now it's just too expensive to do it instead of outsourcing it, the moment it becomes an existential threat it won't be relatively that expensive anymore.
I don't think old computers as such will ever be relevant as you envision, if it happens it will be a blip during a gap of 3-5 years until production restarts somewhere else.
If you're talking about production of modern chips here, it will take a lot longer than 3-5 years.
Later selling to private enterprises, etc. is another matter to be discussed but governments would very very likely be directly involved.
As another comment said, the Manhattan Project took 5 years, it's not impossible.
But if it's a national priority, it could certainly be done. The MP is an example, and so is the moon landing.
For most common people, is there really?
It's usually browsers and web pages that brings these older PCs down.
Better yet, are there any ways to *not even load the other junk*, just load the text only for the reader mode.
As far as I understand, Firefox still loads everything, does the parsing for everything, loads all those ads etc (and no, uBO does not get all of them), and then - *if you click the reader button in the URL box* - it will switch to reader mode. Seems quite backwards to the point of not really a workable solution for the automation junkie programmers of HN.
Would be nice to have a solution with better defaults to give reader mode automatically, default use uMatrix (much better than noscript btw) for control native to the browser, etc for a more snappy and controllable (with paths to private and secure) browsing experience. Having a tui would be icing on the cake.
Lynx just works though, and sidesteps all the js junk.
You can use uMatrix to default block all stylesheets and JS. Then apply your own user stylesheet.
It's not about hating new things. It's about things making tradeoffs that don't make sense to some.
Context: I was on a 3kbps connection and basically everything was unusable, so I ran a textmode browser on a VPS over Mosh.
Also, gopher sites thru gopher://hngopher.com and gopher://magical.fish.
I could even play some streaming radio with OPUS @16KBPS from http://dir.xiph.org once I set mplayer to cache the 90% of the stream.
Some way to tell the server to render text-only, gopher, or gemtext (from gemini protocol)?
what wildly multithreaded memory hogs are you running? I love my x230s, and I have dozens and dozens of windows open all the time. probably I benefit a good bit from running ublock Origin and uMatrix, but I have gigabytes of RAM free. I don't think running a newer machine would give me any benefits at all.
(oh and yeah I have no intention of using a newer system, I bought four X230's and an X220 for dirt cheap lol)
What changes when the program closes and restarts other than memory refresh would be the obvious starting point. I like hunting these things down, may take a look!
if I don't do this when i restart, HN tells me it can't serve pages that fast :)
Also, if you dare,
sysctl hw.smt=1
Also, install and enable obsdfreqd and read the docs at /usr/local/share/doc/pkg-readmes/.
As for login.conf, yeah, I already raised those limits (especially to debug firefox with gdb lol), haven't enabled SMT though. Not sure it will help much with pages bogging down the CPU though, considering, waves broadly at the state of the web today >_>
The Web stack probably isn’t what you’d come up with if you were starting from scratch. But are C, x86, or a lot of other standards we are living with? It seems like the ultimate demonstration of “worse is better:” people had the option of sitting around waiting for some nicely-designed, clean option that was cross-platform across all desktop devices and mobile devices, or they could piggyback on the Web. And with all the investment that’s gone into it as a result of that popularity, it is now fairly good anyway.
By "only adequate" I mean full-of and/or based-on bad design patterns and techniques: eg. usual for active areas (buttons) to be not distinct from other areas; normal for pages to be based-on inefficient/heavy-weight toolkits (base gmail page is 10MB of download)...
10MB might be a lot relative to what's going on (hard to think of many comparable applications that are as old as Gmail to me) but it's still a blip for most people's connections and certainly less trouble than using an installer.
The web has gotten exponentially more useful. The fact that my computer needs to be trivially more modern to use it seems like a completely reasonable trade-off.
You couldn't buy a fully functioning computer for $35 in 2023 dollars back before the web "went astray." That wouldn't even buy you a modem!
When a chat and video/audio call client take 4GB all by itself...
If people even used a framework like Qt (a really mature and well-maintained desktop framework the favored framework of KDE Plasma, etc.), their memory usage would be so much lower. (I'm not counting the memory used by Qt's shared .so / dynamic library files.)
I find it far easier to believe that Qt and GTK have fallen behind rather than the whole world is under a grand delusion and didn't know Qt exists.
Whether this is "right" or "wrong" is a difficult question to answer.
Many people on HN either don't understand or don't care for trading off runtime performance and other technical qualities for other things like e.g. development speed, DX, easier hiring, etc.
No matter how much we deny it! :D
Electron would be okay for them and their users on account of that.
Most people on HN understand those tradeoff very well but also understand that those tradeoffs are being optimized for the company's profit and not for the good of society overall. Without meaningful competition (and for Discord that would mean interopable clients, not just alternate chat applicaitons with their own servers) the incentive for making more efficient software is ~0.
For colour, I'll add: it was a fairly old application first developed before QML/Qt Quick. It was also a multiplatform desktop application; then and now it's unclear to me as to whether or not Qt Quick is especially suitable for desktop applications.
We'd been in the process of moving chunks of it into an HTML/JS web view for several years - I think when that decision was first made it was well before QtQuick would've been sufficient for desktop usage. Other reasons we were doing it, besides the aforementioned bugs/quirks/limitations were that it's easier (and cheaper) to hire web devs than native devs, and that it allowed for code/infrastructure/design reuse with our web frontend. QML doesn't address either of those issues.
Admittedly, I haven't tried QML in anger, but I also feel like...given the quality of the rest of the framework, which is comparatively mature, it'd be pretty shocking to me if it didn't have its own slew of bugs/quirks/limitations.
Virtually every company and developer opts for simpler development and worse runtimes. Is that "wrong"? I'm not sure how to answer that. You could probably make a case either way.
Not to mention half the users here rally against EVs, public transport, and probably own an F150. The environment only matters when we are talking about a 1w savings with election vs qt.
Both of these would be using less power with more efficient software.
But no, lets excuse shitty practices that could save a lot (over the entire userbase) for relatively little effort because it doesn't solve everything.
That, imo, is totally unacceptable (also, financially not optimal).
I've had people tell me that the "developer productivity gains" of using Python totally justifies the circa 10 X electricity usage + 10 X hosting/cloud costs. (Yes, developers cost a lot, but the hypothesis that Python/other slow languages result in so much higher developer productivity is a questionable assertion -- especially, in contrast to languages like Kotlin, Go, etc.)
At least with backend software there is a financial incentive for efficient software. For client software there is no incentive unless there is actual competition, of which there is often effectively none as in the user cannot choose better software because interopability is not a thing.
Would it be surprising to find that they were? It's not like developers have a lot of credibility.
I use an old 2008 iMac running Linux frequently and it runs remarkably well. Firefox and MS Edge are certainly the apps that bring it down from a memory perspective ( it only has 6 Gigs ). Electron apps like MS Teams are right up there though and make it chug even more than the browsers do.
To be fair, I have dozens of tabs open including beasts like GMail but I certainly have to restart the browsers from time to time.
Compiling huge C++ code bases is a memory nightmare as well.
I started with C++, did that for about 5 years, then got sucked into web where I've been for about 20 years. In a funny turn of events, my company has adopted C++ and now I'm writing C++ again. I don't hate it, but 15 minutes to compile/boot a server or 5 minutes to run some tests is kind of unpleasant. And the occasionally UB bug or dangling pointer. It's just really outdated. Enough so that I don't want to relearn how to build a GUI in C++. I'll wait a few more years for Zig or Carbon to mature.
Yes, but things they're counting as "best" include considerations other than performance and resource usage. Not that they're wrong in doing so -- everything's a tradeoff, after all, and engineering is mostly about selecting tradeoffs.
But as a user, Electron apps tend to be pretty bad, and so from my point of view, Electron is as well. I tend to avoid them as a result. But my point of view as a user is very narrow.
Use ccache and clang. Also, with Linux you can set ZRAM and convert these 6GB in something like 9-10GB.
But then I'd be reaching way less people than with the web. My web apps can be tried without installing anything, work on mobile and tablets and on any desktop (except may be cool weirdos using iMac g4s).
The web allows me to reach more people; if I'm working on apps that help make the world a better place (or so I think), that feels more important than the RAM they use (even though using more RAM might be driving more e-waste, thus also making the world a bit worse on that count.)
Not unless the company is penny-pinching when it comes to developer PCs.
That said, the cost really is extreme. So much so that there is an ongoing effort to remove Qt from our products entirely. There is general agreement among the devs that the cost is not justified by the benefits, and nobody can remember why it was chosen in the first place.
All that said, I like Qt. It's quite capable and you can produce decent, portable UIs with it. But I agree it's too expensive to justify.
Looking at the rest of the comments I see that Qt core is free, but the rest of it isn't.
While the "web everything" trend is definitely responsible for a lot of the bloat, it is upon closer inspection not the whole story. For example, I remember various "AJAX chat" clients were quite popular around the turn of the century, worked in a browser like IE5 on a system with probably 64-128MB of RAM and a single core Pentium II/III, and you could also do other things like listen to music, play flash games, or read other sites at the same time. Those clients loaded relatively quickly over connections with bandwidth measured in KBps, and I don't recall much in the way of complaints about their resource usage.
I still use it though, and the graphical client running on my system (which is also an RSS reader) uses 23 MB of RAM at the moment, displaying 1375 messages.
Seen too many (including myself at some point) have all their data in one external HDD and pass it around friends with a caution to be safe with their data.
I've lost most of my photo and video collection due to two HDD failures between 2017-2020.
It /is/ amazing the progress we've seen in our lifetimes. I remember life before the web (http) was invented, before mobile phones (much less smartphones) became commonplace.
Some of the supposed 'progress' is really regression---like the number of people who economically could productively use computers but think a mainstream smartphone can let them do the same things. Sure, they can web browse, access online banking, ticketing, commerce, and other services through apps, but they usually miss out on the entire creative and productive side of computing.
However, I don't just blame technology. At least in Occidental countries, we tell ourselves we're consumers, vocabulary that hides the fact we're also producers/creators in our work. But it fits a mechanistic view of society, and works well with the central planners Utopian pretensions.
I love hearing of projects like the OPs---breathing new life into unused machines.
Projects like this bring it all back.
Also that CPU is ultra low power, like those on laptops, and it doesn't age nearly as well as server stuff.
I thought the same (A500, new enough to have the Agnus variant and jumper to make it 1MB chip total). It happens a lot when it's about memory.
1MB was so much RAM, relative to just 512KB. It could suddenly run Monkey Island and Sherman M4, and it was suddenly much easier to work with Deluxe Paint and Prowrite.
There was also the IDE hard drive interface that I thought was ingenious; it was a daughter board in the 68000 socket, under the CPU, with a ribbon cable connector beside the CPU socket, and the main chips between the legs of the CPU socket.
I think we need to come back to gopher and ftp because http is a rolling target.
But we are https (s comes from encription).
Linux users can compile gcc 4.x and pkgsrc just fine on any > 2001 distro.
It works in virtually all "old" browsers and can be very useful for older computers.
https://github.com/ttalvitie/browservice
For very old web browsers, WRP works similarly but replaces the javascript with an imagemap.
But it looks like this particular project is more focused on getting retro hardware usable on today's internet. Which is not a bad goal either.
I fully disassembled it and cleaned out the dust. I spent an evening lapping the CPU and heat sink down to 2000 grit. I put it in a new case, as the old one's front panel fell off. I installed my old GTX 970 GPU, which seems like a perfect pairing.
Maybe I'll gift it to a young nephew, or maybe I'll set it to easily boot up into a mentally stimulating game for my 4 year old daughter. Her knowledge of computer games right now is beamng.drive and tux racer!
Things like Spectre created a lot of additional obsolescence.
Personal systems are running a lot more untrusted code than, for example, servers.
Mitigations are more necessary on personal computers than on servers.
As for "it's mitigated in the browser", the issue is that there's no "one" mitigation for anything, you have to mitigate it on the OS, in the browser, in the microcode for the CPU. Everything. You can't just apply one mitigation and be done. FWIW the "mitigation" in browsers was to disable high precision timers; which seems to be making its way back, and was never necessary, it just made the exploit easier.
I highly recommend keeping mitigations enabled for personal machines; when it comes to servers with a decent WAF that are not shared amongst users that can run code: disabling mitigations makes more sense. (dedicated databases, for example).
Someone described the mitigations as: "Surrounding a drunk person with trampolines in the hope that when they collapse they will be corrected into the right position", I think this is the most apt example I can think of.
For personal systems that run a lot of untrusted code (due mostly to the web): I wouldn't personally advocate for removing a trampoline.
1. Physical space. Most of the older tech takes too much space. I don't live in a mansion and I have no dedicated rooms just for big bulky computers.
2. Power consumption, heat, noise. I would love to be less consumerist and use stuff from 15 years ago and stop feeding greedy corporations but... the old stuff consumes 90 to 200 watts, heats up easily and noticeably, and can get loud.
Both parts of this problem are trivially solved by modern tech. I got a 13" Chinese fanless laptop second hand for 160 euro; it has a Celeron J CPU with 12GB RAM and 256GB SATA SSD. It can do anything I need that's not programming or rendering -- including playing 60FPS video -- and last time I checked it, it consumed 30W under full load (all CPU cores at 100%) and idled at 11W or so.
Modern tech is not only about its buyers being mindless consumerist drones. It offers tangible benefits, including long-term environmental sustainability.
Yes a lot of CO2 was likely released due to its production but I'd wager that tech that heats up less and uses less power can last 10-13 years and will pay off its environmental footprint with a generous interest on top.
So unless you're a retro collector or want to write compilers for older CPUs, there are zero reasons to be excited about older computers.
(That being said, if somebody finally creates a modern 6502 CPU computer e.g. in the shape of a 10" netbook I'll buy it immediately, provided I can also program it through USB from my workstation.)
One notable exception is the PC Engines APU 2, though that’s stretching the definition of modern.
Seriously: Pick a laptop SKU that is still being manufactured and has similar ergonomics to a MacBook Pro or classic thinkpad (centered keyboard + trackpad, 4K screen, all the other bits in the right spots, reasonably quiet, 8 hour real world battery), and mark it up by $200. Sell it from a site that OpenBSD links to.
Make sure the manufacturer agrees to build that exact laptop for at least 5 years. (Moore’s law is dead, after all.)
Have the project test 100% of the hardware devices in that laptop on each point release, and donate the $200 markup to the project.
For me the thing that keeps me from running OpenBSD 100% of the time is my Steam library, but I'm considering demoting my PC to a living room gaming device and handling my modest desktop computing needs via something small and fanless running OpenBSD.
Most stuff I can buy is overpriced branded crap
The guy bought it from AliExpress. Since then I bought a bunch of electronics from there and was always satisfied.
And Chuwi have official store there btw.
The second-hand market is filled with material that was built at same point; you having access to your nice second hand laptop is due to newer laptops being built all the time. It is still sending a signal upstream that such a computer can be sold.
I do agree that buying second hand is much better than buying new, but not even buying is still better here
Economically the incentives are not aligned this way though, if you pay for grid power... you still have to pay for the electricity, and there's also the opportunity cost of that payment (e.g. you could spend the money you save on your electric bill on carbon credits or green charities or whatever).
Economic sense is not always relevant. I might even argue that economic sense is at odds with environmental sense
I don't really think it does, at least not some of them.
A lot of those older machines are just running really, really high idle power draw. Ex: Pentium 4/5, Core 2 Duo, Athlon64, etc. Running from 40 to 100 watts idle.
For comparison, a new RPI will go as low as ~1-3 watt idle, tops out at ~5-7 watts under full load, and will generally have better performance characteristics.
I run a lot of old machines in my basement because it can make sense (I throw them in a k8s farm), but I also generally have them spend a week on an electricity meter first (honestly, just a function of the UPS I run down there) and anything drawing more than 20 watts idle gets sent to the recycling instead.
The bad ones tend to run high + hot + loud, and they absolutely aren't worth it.
Usually I just leave the new machine as the only thing idling on the sockets without battery backup and compare the two.
I also have one of these (https://www.amazon.com/gp/product/B08LD53F3P/ref=ppx_yo_dt_b...) floating around, but you have to be careful with it since it's not a surge protector. That said, it's cheap and will monitor each socket individually.
My server box is also a 2012 era machine, it's a 3770k build. It's perfectly fine for what I use it for. Yeah it takes a bit more power than a modern intel chip, but it sure beats the alternative of throwing it in a landfill to save a few watts at idle.
But comparing spending 40 bucks on craigslist or ebay for an old pentium/core 2 duo, vs getting an rpi... You're probably better off with the Pi (assuming you can source it, which is not a given atm).
I don't even buy most of mine used, I just get hand-me-downs from friends/family because they want to get rid of the machine and ask me to wipe the drive in exchange for the hardware, but even at an upfront cost of free, a machine using 100 watts over a year is 80 bucks in my region, and my power is right around the US average.
Plus power in my region is tiered, so the first 650kwh per month are cheap, the next 350 are avg, and the rest are expensive. So adding machines at this point is actually closer to ~$130 a year at 100 watts idle (paying for t3 rates).
Personally, it's just not worth it to me to take a freebie that runs me $100+ in costs a year. Not even accounting for the knock-on costs, like the extra AC needed because 100 watts is a small heater.
----
So sure - lots of older machines do just fine (the end of the core 2 duo line is quite reasonable, and most laptops are actually fine) but it's really probably worth measuring the power draw.
The five year cost of operating that rpi is ~$30, the five year cost of operating a 100 watt machine are ~$450. And I can run a LOT more low power machines and stay in my t1/t2 pricing for power.
Do you have any pointers what sort of equipment to use to measure the power draw? Are you simply measuring it from a UPS?
Some figures from around my house.
Lenovo Tkinkpad T400 - 2.4Ghz Core 2 Duo Laptop idles at about 10 Watts and peaks at just over 30 watts.
Lenovo Thinkpad T420 - 2nd Gen i5 is about the same.
Lenovo Mini 10 - 1.66Ghz Atom doesn't seem to know how to idle its power regardless of OS. Runs at a constant 10.5 Watts - 15 watts if using a Hard Drive installed instead of SSD.
Core 2 Duo Mac Mini - 8 watts idle - 24 watts peak.
Sony Bluray player - 0.5 Watts idle. 5 Watts during Bluray. 3 Watts during DVD.
Desktop PC - 2nd Gen i7 + Geforce 1650. Idle 61 Watts - Peak in game 95-101 Watts.
Sony AM/FM Radio - 0.5 Watts. That one was actually really neat to know just how lower power it takes to run.
Also why would you just keep running extra machines all the time? I understand a personal machine and a server but more than that.. for what?
Add an old house on top and throw in working from home, and usage can creep up on you very quickly. We average around 1000kwh a month, but summer months where it's 95+ and near 100% humidity means the AC reasonably needs to be on.
Jun/Jul are the worst, we hit 2200kwh over those months last year.
As for the machines - like I mentioned, they run a k8s cluster I host numerous services on. I run seafile/bookstack/jellyfin/keycloak/homeassistant/email/dns/mealie/snipeit/etc on them. Most idle around 5watts and aren't a big deal. I recycle the machines that would end up costing me real money in power usage.
---
I'll add - my next personal project is going to be moving the entire cluster over to a dedicated solar system. As long as I don't do any grid-tie, I can do it without permits. Just waiting to find a decent deal on inverters at this point. Since I already have a pallet of used panels.
I have an i7 3770 and it's still ok for gaming even.
However, I agree with with the OP comment that computer should not be replaced simply because newer, more energy efficient alternatives exist: at least not for anything made after 2010.
When comparing whether it is worse it or not to extend the life of a device that still does the job its intended to do, it's almost always worth it to extend the device's life.
As an exemple, if we go by Apple's data: a 2016 13' Intel MacBook Pro create a yearly impact of 18 kg of CO2eq through usage (electricity).
The 2021 13' MacBook M1 is a better machine environmentally by all account: manufacturing it polluted less than the 2016 machine (168kg CO2 vs 386kg), and it is more energy efficient (13 kg / CO2 per year).
Still, you would need a whooping 33.6 years for the use of the more energy efficient laptop to cover the CO2 from manufacturing it. And that's using global carbon data, if you live in a low carbon grid country, it could take a century to recover the greenhouse gas emitted during manufacturing of the new laptop.
(Not saying I would not still buy a M1 over keeping the slow Mac Intel but please don't claim your are helping the environment when doing so).
Source: https://www.apple.com/environment/pdf/products/notebooks/13i... https://www.apple.com/environment/pdf/products/notebooks/13-...
You do need storage, but you need it for all of them anyways. I certainly won't trust the original spinning disk HDD that comes with hand me down machines with anything more than an OS image. Everything else goes on the NAS.
> However, I agree with with the OP comment that computer should not be replaced simply because newer, more energy efficient alternatives exist: at least not for anything made after 2010.
This - this is where the answer is to measure it. Sometimes the result will surprise you quite a bit. Most machines made after 2010 will be fine, but trust me, it's really worth plugging it into a watt meter and making sure.
Turns out all sorts of things adjust power draw, and one of the least maintained parts of old machines is the PSU - usually it's filled with dust/dirt, and as long as the machine still runs no one has paid it any mind in the last 10 years.
> Not saying I would not still buy a M1 over keeping the slow Mac Intel but please don't claim your are helping the environment when doing so
Who the fuck is talking about replacing an intel macbook with an m1? I'm talking about getting old tower machines from office/school closeouts or as hand-me-downs and repurposing them. An intel mac (even a 2010-2013 era one) is still worth WAY more than an RPI (~$200 vs $40), and laptops in general use a very different power curve (batteries matter - who knew...?).
Not to mention you can't even run a decent OS on the modern intel macs after 2016... (although at least Asahi is making progress these days) so trust me - I'm hardly suggesting a macbook update (I would strongly suggest purchasing from a different company, though - but that's a different conversation related to a different set of morals).
> It still makes sense to use a decade-old computer. >> I don't really think it does, at least not some of them.
I totally agree with you regarding the homelab topic (although personally, I started using low-powered refurbished x64 thin-clients (using about 10W) in place of a RPI cluster, mainly to get real sata ports on them.)
My point is that, environmentally, continuing to use a 10y computer rather than buying a new one is generally a sound advice (assuming it does still what you want it to do).
Since I didn't have data for a 10y old computer, I used what I could find: the 2016 vs the 2021 Macbooks Lifecycle assessments from Apple. The same point would probably stand between a 2010 computer and a 2023 computer. Even if the old one is less energy efficient, the 150 to 300kg+ of CO2 you save not building a new device goes a long way.
And instead of consumer devices or RPi, if you are talking about datacenter-class servers, then you save even more CO2 by prolonging your device life since a new server's manufacturing footprint is more like 1500 - 3000 kg CO2eq.
But I believe the fact that this laptop can last at least 10 years + it will consume a single order of magnitude less power over these 10 years adds up to more environment care.
The production is already completed. The damage has been done. Let's use the machines at least because they are 10x more efficient and we can slow down the environmental damage.
Also to many of us outright not buying machines means we gotta switch industry. And a lot of us don't want to.
> It is still sending a signal upstream that such a computer can be sold.
That's a bit out there though; it only signals the vendor that people hold on to their older machines for longer time. If anything, the laptop I got second-hand is no longer even listed on the vendor's AliExpress store. It only has a marketing page on the website and you can only find it second-hand or here and there on the internet.
If it does then the people using them don't need to buy another one, new or refurbished. If they don't need to buy, other people can't sell as easily. If they can't sell as easily, they might use it longer, and not need/want to buy another one, old or refurbished, etc...
It's a whole market and if constructors see that the overall tendency is of using longer, they will get the signal that there can be less sales overall. And thus build less.
Everything we do is part of a larger system, and I believe we should understand what the system is like and how it works if we want to have a real impact.
The process you describe is yet to happen e.g. the computers I have I don't want to replace unless they crash and burn. So in 5-ish years I believe vendors will slow down manufacturing indeed because many people do like I do: they get the newer more power-efficient stuff and then hold on to it for longer periods.
Again, I'm not saying what you're doing is bad, if everyone could fulfill their digital needs and wants out of second hand, that'd be awesome. I only wanted to point out that using less electricity isn't automatically better.
It's a balancing act between what we need to get our jobs done, what are our ergonomic preferences -- I have the so-called "lap desk" (basically two parallelepiped-shaped pillows and a well-designed plate with USB extensions to put on top of them) where I like to work on the couch or the bed -- and what is kinda sorta good for the planet.
I can't hyper-optimize only for the planet. I am doing my best but obviously everyone else has to help a little bit as well.
Thanks for entertaining this discussion, it was interesting.
To first order the upper bound for carbon cost of production/recycling is going to be a multiple of the energy required to melt the components. Call that multiple four: production of raw materials + reshaping in factory + recycling of materials at EOL + energy cost for assembly. So a rough upper limit of the energy to switch a computers would be four times the energy to heat the computer's mass in water (4 J/gC) up to the melting point of silica (1900 C), and a rough lower bound would be the same using the heat capcity and melting point of iron (10x lower heat capacity, 1500 C), and neglecting assembly. For 1 kg of material, that gives a range of 1.8 and 32 MJ. Now you can compare to the energy used. For myself, I'm on a 2 kg laptop that draws up to 250W, but conservatively 50W. That adds up to 32 MJ/kg in 400 h. If I can reduce the energy use by 5W with a new model, that will pay itself back in 2 years of workday computation. Crazy.
Caveat: there will be additional non-CO2 environmental costs. Most components are not recycled, and resource extraction is also environmentally damaging.
The power issue is more with large desktop systems. I was setting up a moderately old but fairly hefty system as a file server. Couldn't get it working--which I discovered belatedly was an issue related to my network switch--so I bought an unpopulated Synology box. But, given power consumption, I also realized that was probably the more cost-effective thing to do anyway.
Appreciate the dedication!
My desktop PC, which is over a decade old, performs fine under current releases of Linux and Windows. My laptop PC is a bit over two years old. Performance is lacklustre under the current release of Windows. Newer isn't always better, particularly when you're comparing laptops to desktops. I also doubt that many laptops would have an equivalent lifespan to a desktop. They have more perilous lives due to their portability and portability means they are frequently built with more compact and less robust components. That is especially true when you consider that laptops are more difficult to repair (due to more integration and more difficult to obtain parts).
Granted, the iMac G4 is in a different league since it is a 20 year old machine that was produced when performance jumps were more meaningful. I can see it being a useful machine for someone who has limited interest and use for computers or simply wants something disconnected to get work done. I took the latter approach in the early 2000's and it worked quite well. (The fastest machine I had was a 68040, and I did run NetBSD on it.)
There were likely several versions of lcd panels with it - the t420s's I got were way better than the t480 and e580 I picked up recently, even though there's 6 years between them. I believe the ones I have are either 1920 x 1080 or 1600 x 900 (whereas the t480 is inexplicably only 1366 x 768 ; and e580 is 1080p but awful contrast or gamma or something)
Is there a good site for recommended upgrades by model? I'm pretty pleased with this as is for Linux use already. (8 GB RAM, 250 GB SSD).
But again, power consumption.
And I'll grand you that it's possible that this Chinese laptop I got will have its logic board burn long before the 10 year mark. Sadly that's likely, yes. But as mentioned, when I am not programming or doing anything that requires more CPU power, it performs perfectly for what I need, and I feel better knowing it consumes ~17W on average.
There are likely sweet spots e.g. get a Ryzen R1000 or V1000 mini PC and stuff it with 2x NVMe and 2x SATA SSDs and have it drive 2-4 displays at home. The displays will likely consume 20x the power of the machine itself (lol), and that machine is very likely to last for a long, long time.
For my needs however a Chinese laptop that costed me 160 EUR and has proven it can last anywhere from 7 to 11 hours, is perfect. (Though I'll get pissed if it lasts less than 5-6 years for sure.)
Which is great when it's -30 outside, like today. :)
Yes, you listed them and they are valid, except
> including long-term environmental sustainability
Err... no not really ! The environmental impact of pretty much all consumer electronics is by (very) far dominated by the impact of its manufacturing (mining, water for manufacturing the chips, oil for transport, robots for assembly in countries where electricity is made with coal, etc), therefore anyone trying to be serious about "environmental sustainability" understands that it mostly implies to extend as much as possible existing equipment and avoid buying new ones (even if the old ones are bulky, noisy and consume a lot of electricity compared to the new)
And in the end I was left with a machine that ran well over 300 watts to do what a Raspberry Pi4 can easy do in well under 5 watts. I should have known, I was there when the 1st Intel Mac Mini came along running as fast as these things on launch - I knew they were a slouch even then. I should have known better.
Eventually something went wrong with the power supply and I ended up scrapping everything but the case which now has my PC mounted in it.
I use old laptops. My daily driver until a year ago was a Dell e7470 which worked well enough for browsing and coding in VSCode.. until the keyboard broke, twice. Now it's gathering dust because I'm too lazy to change the keyboard again but it should work fine on linux atleast.
My current daily driver is a thinkpad t490 with an i7-8565U and 16 GB of ram that was provided for me at work and which I can buy used for 500$ atm.
But if what you want is fanless and decent hardware I'd recommend a NUC.
I'm not sure which Celeron J you're referring to, but I know some models were affected by the infamous LPC clock bug --- which might explain why there seems to be a surplus of machines with those CPUs available for amazingly low prices.
I like computers the same way I like my cars, loud and hot ;-)
Eco-virtue-signaling, the other modern plague...
But I also don't like paying a lot for electricity. :)
Virtue signalling is dumb and I hate it. In this case I happen to also have a personal interest.
Feels like we jumped 10 years forward in a single iteration.
However I quickly figured out there is not much to actually do there. The way modern web works is horrible (I don’t want to click just another cookie banner thanks), but at least it’s easy for all people to contribute.
(a few months later)
"Why is everything such unmitigated crap?"
https://github.com/w00fpack/dilloNG
Mpv it's recommended, at it has a context menu to play videos with it and the help of youtube-dl or yt-dlp. A config for it:
~/.config/mpv/config
ytdl-format="bestvideo[height<=?420]+bestaudio[height<=?420]"
vo=xv,drm
audio-pitch-correction=no
quiet=yes
pause=no
vd-lavc-skiploopfilter=all
/usr/share/X11/xorg.conf.d/10-ati.conf
Section "Device"
Identifier "ati"
Driver "ati"
Option "AccelMethod" "EXA"
Option "DRI" "2"If any, you can set
machdep.allowaperture=2
Also, /etc/fbtab needs some settings for MESA, too. Basically you need to set read write the files under /dev/dri* with fbtab from /dev/ttyC5 if I can recall it well.
EDIT: typo fix in /etc/sysctl.conf
Unfortunately, the problem of proprietary software and APIs is always around the corner. I can use a TUI client for Telegram, but I can't if I want to message someone on Facebook. I can use a simple newsreader for blogs, but I can't for my friends' Instagram posts.
I could honestly scrap by with a 20yo computer if the world was more open sourced. Such a shame
Not "usable (but read the fine print)". Comfortable to use (running Linux at least).
IMac G5 from 2004 on the contrary?
Too heavy, too hot, two slow (single thread cpu doesn't cut it)
Late 2007 Intel core 2 duo? RAM maxes out at 4Go but surprisingly usable. Not comfortable but usable.
That should give you an idea of where to set your expectations. I do recommend buying second hand but I wouldn't go under 2012 for a daily driver personally, for peace of mind.
Can it run OS 9 natively? The earliest iMac G4 could, I've heard.
I bought a Mac G4 a couple years ago so I could have the most powerful computer that runs Mac OS 9 natively which then enables me to run Symantec's MORE.
MORE is this ancient, super dope outliner editor, which is just so fast and ergonomic at the keyboard. It's go all these features that have never really been replicated since, like cascading styles and presentation mode.
https://tidbits.com/1993/10/18/more-more-dinosaur/
I use Workflowy, which I love, but I kind of want to play with porting MORE's UI to the web and see if it could perform.
I've been trying to fight that a bit myself: https://github.com/cellularmitosis/leopard.sh
You got 64-bit CPU, SATA HDD, PCIe graphics with OpenGL 2, USB 2.0 connectors, and DVI flat panel display. Its something I'd expect you can just throw Debian Stable in and be perfectly usable with lightweight WM, or maybe even something like XFCE for full graphical desktop. Of course as others have mentioned, browser is the real problem, but 1GB might be enough to browse HN at least?
It feels like a base install leaves you in several hours of configuration debt, and I apparently don't have the concentration span to dig my way out to somewhere interesting and not just half way back to productive on the systems I already use.
My newest computer is from 2014 for reference, oldest from around 2008 and I emulate older stuff.
I know I'm "doing it wrong" and perhaps it isn't a right fit for everyone but ... That's how it is.
For this same reason, it's simply a poor desktop OS. A modern UNIX-like desktop is a very complex system, with so many moving parts held together with hacks and duct tape. As OpenBSD effectively forces you to familiarise yourself with such duct tape solutions, it also forces you to do more work than you'd want - or to stick with environments that feel somewhat basic.
And I doubt it could handle more than a few open tabs with 512MB of RAM.
Most everything else I use computers for have been fast enough for 20 years or so.
So 2014 Chrome with webapps like 2014 Gmail wasn't really practical on 1GB. So I very much doubt that 2023 Chrome with 2023 webapps would be practical on 512MB RAM.
I always take time to consider where the balance point is between replacing working, but less efficient equipment with newer, more efficient equipment.
It took materials and energy to produce the old one. I know electricity is expensive some places, but even then, sometimes the break even point is years in the future.
Despite all of this, I only realize its more complicated than I hope.
You know that producing that PC used way more energy than it would consume if you left it running?
> taking into account how much electricity rose in Europe
This seems like your main reason actually. Your wallet, not the environment. If the energy prices went down, you would have kept the old PC around rather than having it end up in a landfill or shipped to a poor country where people literally burn the PCBs to recover tiny bits of rare metals.
At the end of the day if all you need is a computer, it’s nice to know that basically any computer can still be serviceable. In a world where you can get 4k series i5s in the trash though, I don’t actually think anyone will be using such an old Mac seriously.
What an incredible waste.
Still, my intent is moreso, well, companies are literally throwing these systems into the trash, so paying for them feels like I am getting ripped off. I'm always hoping for the opportunity to snag a system or two that is being tossed. It has happened before, or I've paid like $5-10 for something that is destined for the dumpster. It definitely seems like a "have to be in the right place at the right time" kind of thing, though.
I think this statement may be wrong.
A good example in a similar fashion is the Super Nintendo. The power supply is rated to 7.85 Watts but it only uses roughly half of that when running. I believe the Neo Geo used 5 Watts peak. And these aren't physically tiny machines like a Raspberry Pi.
Honestly after seeing the absolute state of Linux on the desktop over the last decade, there is no reason for me to even consider it any more.
What is wrong with Linux on the desktop, and how is OpenBSD better?
Now compare that with ALSA / PulseAudio / JACK / PipeWire... The audio system in Linux is fragmented and incomprehensible.
I try to use w3m a lot for that stuff. Many sites still have a cheeseburger menu made out of a tree of <ul> lists up the top of the page, and require CSS indentation or borders to render discussions nicely. Are there terminal browsers that support these features better?
I guess this is a kind of Computational Maslow's Hierarchy of Needs.
Basic needs can be covered by the vast majority of computer hardware. But once you get to large scale web browsing and gaming - more complex hardware is required.
The difference between a hermit living content in a cave and a billionaire in a 42 bed room mansion. Both could be happy but it really depends on their perspective.
We have been using those systems since then and they are really amazing performers (linux and bsd, many java apps, jetbrains IDEs, etc). This is totally because of how efficient and minimal modern Linux and BSD distros can be, by the way - even with modern desktop environment like Gnome 43!
Only downside is the powerdraw. They idle at 180 watts, s3 sleep is still 150 watts and on BSD idle is 240 watts! This wasn't too much of a problem before the whole energy crisis thing, because electricity was quite cheap (around 0.19cts per kwh). But with prices touching 0.80cts pet kwh, it becomes untenable quickly.
We currently Wake-on-LAN these boxes only when we need them and they auto turn off after 1 hour of idle.
Haven't tried OpenBSD but Windows 11 works quite well Samsung Galaxy Book 2 (ARM):
https://www.theverge.com/2018/10/26/18024696/samsung-galaxy-...
It has superb monitor, touchpad, touchscreen, pen and detachable keyboard.
I am running Visual Studio 2022 on it to build Sciter - quite convenient to test and debug handling of those input devices.
It is really quite good machine for mobile use.
You can buy second hand Samsung Galaxy Book 2 for ~200-300 USD
ifconfig umb0 apn hutch up
The ultimate test: can you develop Java on it? (In practice, not in principle.)
This, of course, is becoming less and less true.
For the rest, UBlock Origin it's your friend and git://bitreich.org/privacy-haters to cut in half the realistical Chromium requeriments.
Switch Chrome by Brave. Add a tab suspender. If you have 4GB or more, set the browser cache to RAM (a tmpfs directory).
You'll be surprised of how speedy and usable is your old Celeron computer.
And this CPU is very power efficient, so no heat and no fan. Silence.
My 2016 Lenovo laptop was almost unusable with windows 10 and its myriad of processes hogging the CPU.
Manjaro Linux brought this little laptop to life and now it's my daily driver, more than enough for my needs (browsing, coding).
git clone git://bitreich.org/privacy-haters
That and UBlock Origin with the country specific ad blocking rules will cut down a good bunch of crap.
Whatever.
The GitHub repo I posted has sane defaults for both Chromium based browsers (adapt the environment variable for the custom fork, you have it at /usr/bin/chrom file) and Mozilla under *Nix. If something "clean" as Mozilla has a good chunk to be tweaked, just imagine the amount of telemetry Brave has.
I lived the 90's and 00's in computing. I pretty surely know the spyware and adware terms. And Brave it's spyware, period.
One of the systems on which I run NetBSD is a 33 MHz m68030 Mac LC III+ (http://elsie.zia.io - it's hosting a site about an LC II, which I'm still working on). It's quite useful to see how assumptions people make about "acceptable" performance regressions bear out in the real world. Sure, not much can be done about taking six or seven minutes to ssh, but when bad coding causes a shell script to take twice as long, it's much more obvious on hardware like this than on a modern Ryzen system.
Nobody is telling others to forego your modern computer and your Windows needs. It's just interesting to some of us that we can still run modern things on very modest and, in some cases, very non-mainstream hardware in 2023.
Obviously I would never suggest someone replace their modern daily driver computer with one of this age.
This works.
-Emily
There are some video links in the (german) article https://heise.de/-7448263. Mind-blowing.
The latest minimal kernel will still boot on 4MB of memory on 32 bits machines, although it will take more than that to be usable.
Unfortunately NetBSD can only do so much considering the sad state of the Linux ecosystem. GCC has become so big it won't self host on most platform, making cross compilation mandatory for them. Switching to PCC would be a solution, but a lot of software won't compile with PCC and obviously their maintainers couldn't care less.
Then you have the newer behemoths like Rust and Go. Once they start creeping in some majors projects like GCC or OpenSSL, a lot of machines will turn into e-waste, and the resource requirements will climb even further.
If you care about making software fast, you invest time in the measurements-ideas-measurements loop, where having fast hardware helps a lot (it allows you get measurements faster, and try out more ideas in a given time frame). Since any sane measurement is based on benchmarking and not eyeballing, slower CPUs don't help you. And you certainly can care about fast software without having a slow computer personally.
I recently changed 1 byte in Chrome.dll and decreased the lag for copying large images by a factor of 2-3. (The UI freezes for 5-10 seconds. Turns out it's re-encoding a 2MB JPG into a 20MB PNG for the clipboard, at compression level 6 no less!)
I put up with it for years thinking there was nothing I could do, but one day I decided to ask some people smarter than me, and they gave me enough info to track down the code in IDA (I had never used it before).
My point is, motivation is a powerful thing!
(Offtopic but turns out Firefox copies the original file, so you can paste it right into Windows explorer—blew my mind!)
https://source.chromium.org/chromium/chromium/src/+/main:ui/...
An improvement would be to replace EncodeBGRASkBitmap with FastEncodeBGRASkBitmap so that it uses the fastest zlib level instead of the default (6). This is what the Linux X11 (ozone) implementation uses. This would be an easy fix to upstream if someone wants to do it.
https://chromium-review.googlesource.com/c/chromium/src/+/42...
CPUs have changed a lot, in that caches, speculative execution, and parallelism are the way to get performance, and that was not true in the heyday of a classic Mac.
I'm still in favor of porting to old platforms for various reasons. But we need to keep in mind that performance optimization is not one size fits all.
In fact, I found an over-dimensioned workstation to be one of the best tools improving performance of non-trivial programs. There's a lot of performance issues that don't stand out sufficiently on today's common-place enduser hardware, that will be very visible on a current two socket system. And fixing the issues benefits the lower end hardware of today. More often than not, the large issue one can see in large machines today, are issues seen on lower-end hardware in a few years.
It's easy to see using too many cycles on enduser hardware. The effects of memory latency, various forms of cache contention, etc however, not so much.
> Since any sane measurement is based on benchmarking and not eyeballing, slower CPUs don't help you
I do find that somehow, to a non-rational degree, I have a considerably easier time getting a "feeling" for the performance bottlenecks when benchmarking on my workstation, than when benchmarking on a server. Which I guess kinda falls out of proper benchmarking.
Also, ironically, people elsewhere in this article thread keep mentioning that older hardware is too power intensive. But the reason the APU2 is still around is because neither Intel nor AMD offer modern chips with a maximum TDP as low as the AMD G series. At maximum load, with all 4 cores, DRAM, and NICs at full throttle, the APU2 only dissipates something like 10W of heat. Modern x86 chips are more power efficient, but their maximum TDPs are just too high. Some of the laptop and embedded series get close at ~15W (not sure if that includes DRAM and NICs), but that's too hot. The APU2 is designed to only require passive cooling, without finned cases, without conditioned air, and still provide years of reliable always-on service.
I see the same issue with higher-end AMD and Intel chips. Under light load their power draw is very low, but under load their chips are often more power hungry than comparable models from 10 years ago. Yes, they get more work done per watt, but there are many situations where I'd rather know that a server will never draw more than, say, 50-60W, even if that means requests end up being throttled. But I don't want to use their crippled low-end chips, either, especially when I can just continue using my 10+ year-old servers that provide better performance. (Theoretically you could manually throttle requests to maintain the power envelope, but I don't want to be burdened managing software configurations.) I prefer the "edge computing" series of chips from Intel and AMD, but over the years their max TDP has slowly crept up. Thankfully there are still a few options that work well, but it's definitely a niche that doesn't see much attention.
The closest thing we have today is getting a quad core Ryzen and running sysctl -w machdep.cpu.frequency.target=1600. Personally, I'd rather have more of those wonderful 2014 motherboards and CPUs.
And Linux' packet processing now pushes 100Gig and beyond, which is not motivated by low-end hardware but certainly helps it. :-)
> But the reason the APU2 is still around is because neither Intel nor AMD offer modern chips with a maximum TDP as low as the AMD G series.
You can get Alder Lake-N at 6 W TDP, which is the same as on AMD GX-412TC (the CPU used on the APU2).
It takes too long, but much less than that, on Netbsd @ A1200 with 030, albeit 50MHz.
I'd expect a 50 MHz '030 to be faster than a 33 MHz one ;) But running newer ssh and sshd means more expensive default ciphers. I can speed it up a bit by being more deliberate about the choice of cipher, but I want to run things by default, generally speaking.
The CPU fan died at some point, and I didn't notice until I opened it up to change a hard-drive. If it's not still running it's probably due to an exploded capacitor, because that thing was indestructible.