Hacker News Reader

Comment by genewitch | Hacker News Reader

as a data point youtube is unusable on raspberry pi 3. This happened within the last year, because prior to that you could "watch" videos at about 10-15FPS which is enough, for instance, to get repair videos in a shop setting (ask me how i know). When the raspberry pi model B - the first one released - came out, you could play 1080p video from storage, watch youtube, play games.

I'm not sure what youtube is doing (or everyone else for that matter.)

If we're serious about this climate crisis/change business, someone needs to cast a very hard look at google and meta for these sorts of shenanigans. eating CPU cycles for profit (ad-tech would be my off the cuff guess for why youtube sucks on these low power devices) should be loudly derided in the media and people should use more efficient services, even if the overall UX is worse.

View on HN

Could it just be due to lack of hardware video decoding? The Pi3 has x264 HW acceleration and youtube started using other codecs a while ago.

I have no idea if it still works, but the "h264ify" browser extension used to be great for working around this issue (by forcing youtube to serve h264) https://github.com/erkserkserks/h264ify

i did a full apt dist-upgrade to try and get the h264ify plugin to install and if i remember correctly i never was able to get it to install. I upgraded from "chromium" to "chromium-browser" and set all the compositing and other settings recommended for the RPI.

and to reply to another sibling, "yt-dlp" isn't workable, this is for a senior citizen that does small motor repairs.

I got an HP elitedesk that's a few years old coming in monday to replace the RPI; hopefully that will last another 3 years before google et al decide to "optimize" again.

RPI 3 for a senior citizen seems like a poor solution in the first place.

I would have opted for a small business-pc that is x86 based and 3-4 years old.

a used laptop that can play youtube videos can be had for about the same money

ytdl-format=best[vcodec!*=vp9]

Probably. I remember when YouTube switched to H.264 (it might have been some Flash-based video before that). I had an older Mac mini hooked up to my TV at the time and suddenly video framerates dropped to an unwatchable level because they saved their bandwidth (and mine but I didn't have to care about my Internet service was not metered) at the expense of client-side processing.

Is YT so impoverished they can't manage some sort of negotiation mechanism that includes x264 and makes it work?

They encode videos ahead of time and they likely decided that whatever hardware you’re judging them by is only .9% of the market so fuck those guys.

Big companies use percentages in places they shouldn’t and it gets them in trouble. .1% when you have a billion users is a million people you’re shitting on.

For me that might be a dozen people. Very different.

Encoding and storing billions of videos in a format used by 0.1% of users feels like a waste though

robustness is only wasted if you're lucky

The context above was exclusion of people based on income level.

Supposedly, the whole point of Google financing “open codecs” was for them to break free from MPEG codec licensing. I imagine the total amount of fees had a lot of zeros. So, yes, each time they don't serve H.264 (unless absolutely required) results in saving a lot of money.

That might be more on the browser that you’re using. It might be saying “yes I can play this format” to a format it can barely play.

every yt video is available as x264 but vp9 is cheaper (smaller) and has better quality

YouTube is definitely getting heavier. My early 2021 MacBook Air (Intel) now gets random video pauses under moderate load, something that never used to happen.

Could be just ads that adblocker tries to block. Google is trying new ways all the time to bypass adblockers.

>If we're serious about this climate crisis/change business, someone needs to cast a very hard look at google and meta for these sorts of shenanigans

By all accounts client devices' energy consumption is a rounding error in terms of contribution to climate change. Going after them to solve climate change makes as much sense as plastic straw or bag bans.

IT is emitting around as much as aviation, and that was a surprise to me, most of it are due to client devices. Don’t have the source at hand at the moment though. And of that, most emissions are upfront until you buy it. Buying a new device because it’s not fast anymore causes emissions, not running it. Think about e-waste as well.

>IT is emitting around as much as aviation

What counts as "IT"? It's most certainly a superset of "client devices", which is what my and the parent comment was talking about.

It has a cumulative effect and drives the continual "upgrade" cycle. When you consider the life-time of an average mobile device, and the resources required to manufacture and ship them, it's a not insignificant problem.

Random source from google[1]:

>Berners-Lee writes that in 2020, there were 7.7 billion mobile phones in use, with a footprint of roughly 580 million tonnes of CO2e. This equates to approximately 1% of all global emissions

Of course, not everyone is replacing their phones yearly. Another source[2] says the average consumer phone is 3 years old. That works out to 0.33% of global emissions, assuming the phones aren't recycled/reused to developing countries. Even if assume people are upgrading their phones for app/web performance reasons, the impact is far less than 1%.

[1] https://reboxed.co/blogs/outsidethebox/the-carbon-footprint-...

[2] https://www.statista.com/statistics/619788/average-smartphon...

To be clear, these emissions include the manufacturing cost, which for reasonable users seems to make up ~80-90% of the carbon footprint. The power usage of the phone itself and associated data centres etc is only a small portion.

It's still somewhat surprising that one could attribute 0.2% of global emissions solely to phone power consumption... I would have expected it to be lower.

Isn’t that quite huge number to be fair?

Compared to a single person's emissions? Yeah sure, but that's because anything multiplied by 8 billion people is going to be huge. The same could be said for plastic bags and/or straws. In relative terms it's absolutely minuscule, and in terms of low hanging fruit it's definitely not the top. You'd be far better off figuring out ways to decarbonize the electricity grid (40%) or the transport system (20%)

I would imagine for phones and laptops the extraction of materials (rare earth metals to make fancy new chips, lithium for batteries,etc) is probably the bigger issue.

Having gotten away from 500+ watt desktops as the standard for light non-gaming computing has been a win in the energy consumption court.

I think there are lots of good reasons to avoid the upgrade cycle but energy consumption of the end device itself probably isn't it. (Embodied energy of the devices, environmental impacts of mining, no good EOL story for ewaste, etc)

> By all accounts client devices' energy consumption is a rounding error in terms of contribution to climate change.

It adds up? How many devices are there? Tens of billions?

Web 345 devs just don't care because the costs are borne by the customer.

The customer doesn't care either because a page that takes 5s longer to load on a 1W TDP SoC costs them around one-millionth of a penny. Even if you're refreshing 100 times per day it's only around 0.05 kWh per year, which at any reasonable electricity prices is a sum that's simply not worth worrying about. You'd get more savings from getting people to turn off their led light bulbs for a few minutes.

US centric electricity prices view :)

Also, it's not just your site. It's every site. And the customer pays all those millionths of a penny added up out of their pocket. And all those 5 second delays out of their lifetime.

Edit: btw at a quick glance you underestimated cell phone soc TDP by a 2-4 factor.

A single use of an electric kettle sounds like it would completely dominate this consumption.

The time cost is certainly the greatest expense here, power is cheap in consumer computing contexts, generally speaking (at least nowadays with most things racing to sleep), and is mostly relevant because of battery life, not power cost.

> A single use of an electric kettle sounds like it would completely dominate this consumption.

But at least that gets you tea, instead of engagement.

You've got to put X joules in to boil Y liters of water. No choice there, except giving up on the tea.

You can greatly reduce the joules necessary to see cat photos though. And you don't have to give up on seeing the cat photos.

I use Invidious for browsing the site, and watch the actual videos via a script that deobfuscates and gets the actual stream URL and then passes that to VLC.

As another data point, YouTube a decade ago would've been perfectly fine on that hardware too. The culprit is web bloat in general, and more specifically the monstrosities of abstraction that have become common in JS.

Even for those who don't believe at all in "climate crisis", there is something to be said for the loss of craftsmanship and quality over time that's caused this mess, so I think it's something everyone across the whole political spectrum can agree with.

'apt install yt-dlp mpv'

then put this in '.config/mpv/mpv.conf' to twart hw requirements

ytdl-format=best[height<=?720][vcodec!=vp9]/bestvideo[height<=?720][vcodec!=vp9]+bestaudio/best[vcodec!*=vp9]/best

and pass url's to it (i use 'play-with' ff extension)

Can you share that script? Also using invidious, but passing to vlc sounds good for saving cpu cycles.

Just use something like this:

    mpv --demuxer-max-bytes=1024MiB --vo=gpu --opengl-es=yes --ytdl --ytdl-format="best[height<=800]" "$url"

We need some watchdog group that watches page weight across sites and users and names and shames them.

Maybe they could do that Consumer Reports style, or maybe it’s an add on the works a bit like Nielsen ratings.

Its worth trying out different browsers. In my experience Chromium based browsers are a bit faster than Firefox on really low end devices (Pinephone, ...) as long as you have enough ram (>1Gb?).

E.g. On the OG Pinephone a 720p video on Youtube is running smoothly in Chromium, but not Firefox.

I've got an old Roku box that has started rebooting after a few minutes of playing youtube videos.

In your case, maybe pulling the video with yt-dlp then playing it works...

YouTube was not tested because monitors can't handle CMYK, and we need a lot of that extra coal black to color the results.

I had to upgrade my 12" Macbook because Youtube Music brought it to a crawl. I could play music or work, but not both.

I used a 12" Macbook as my main development machine. It ran IntelliJ with Python/Django applications, Postgres & Redis running in parallel (along with Safari, Mail, etc) around 2018-2020 just fine.

Tried it somewhat recently around Ventura and the machine clearly appeared to be struggling with the OS alone. So we had a machine that used to be capable of actual, productive work, and is now seemingly struggling at idle? It doesn't look like the new OS brought anything new or useful to the table (besides copious amounts of whitespace) either.

I got mine in 2022-2023, so I couldn't tell how it changed. However I wouldn't be surprised. Operating systems used to be much snappier.

That's absurd. I remember using winamp (and the skin compatible Linux clone, I forgot it's name) streaming internet radios while programing a toy OS in 2004. I could listen to music while compiling and running the BOSHS emulator on my AMD Atlon CPU with a whooping 256MiB of RAM.

Reply on news.ycombinator.com