I had a look at the implications of the Facebook Metaverse on emissions from computing. It is scary.
For the ideal VR experience, you need a network bandwidth of 1.6Gbit/s. To watch an HD video at 4K currently, you need 25Mbits/s; 8K would be 100Mbits/s.
So VR requires 64x more bandwidth than 4K video!
Ignoring the infrastructure emissions, purely running this from a cloud data centre effectively means a 64x increase in energy consumption and therefore in carbon emissions.
When I researched my talk about Frugal Computing, I did not discuss VR, as none of the studies I referenced considered it. But with a giant like Facebook behind it, VR might become a very considerable part of our lives.
That would be a disaster: already, emissions from computing are dominated by video. VR will make this many, many times worse.
The only bright side is that with current technologies, there is not enough electricity in the world to power this ideal-experience VR revolution.
@wim_v12e In thinking through possible validations for the metaverse, an earlier realisation emerges again.
I've previously argued that large, bloated, slow websites requiring highly-provisioned systems and high bandwidth are a market-segmentation technique, adopted either intentionally or accidentally.
In a world in which anyone anywhere can access online services, and the Web and Internet are increasingly used for commercial activity and as a power- and wealth-signalling mechanism, such sites serve as high-value shopping and entertainment districts in the physical world.
There's also the fact that Facebook originated as a socially-appealing cohort of highly-selective university students, a cachet it's long since lost.
The Metaverse reimposes those costs on participants, requiring vastly greater end-user resources. It also creates a capabilities moat among system providers which only a very small set of present tech giants could possibly hope to meet.
In short: it's an attempt to re-create a highly-selective, highly-limited, members-only club of the well-heeled and powerful.
The though first occurred to me in the context of Ling's Cars and its ... exhuberant ... design ethic:
@dredmorbius So the poor users will get crappy VR, the rich ones high-quality.
My concern is that even the poor quality VR will cause a considerable rise in emissions, something we really can't afford.
@wim_v12e I hear you, but I think you're looking in the wrong places.
Think advertising, market segementation, data gathering, and chasing the upmarket cohort.
The fact that the initial rollout is apparently a virtual-office offering is ... all the more pathetic.
My sense is that this is an effort to break or disrupt the Facebook Files news cycle. So someone's weekend-project cartoon-workplace MVP is getting broadcast on NBC specials with exclusive interviews.
@wim_v12e It's always somethin".Always has been,till the novelty wares off and move it on to something else. I stopped paying attention decades ago. Repetitious pattern of fear agenda all my life.
@jfred Yeah, I was thinking the same thing. Video has to send you all the encoded frames and audio while VR is just... Textures that doesn't really change when loaded, models that in general are quite small and doesn't really change when loaded... Then it's positional data and audio. So in my understanding, for example. The VR meetingrooms use less bandwidth and has less latency than say... Google Meet. Where you actually stream video and audio all the time. @wim_v12e
I'm guessing they're imagining that they're going to be streaming an entire AAA gameworld every time you walk through a door... I don't think people need or especially want that, and there's a bunch of new compression tech they could adopt if they were serious.
@faun What I describe is what is called the "ideal VR experience" in a technical report from Huawei. All I'm saying is that the bandwidth requirements for this "ideal experience" are entirely unsustainable. This is already taking into account the best possible compression. Otherwise it would be an order of magnitude worse.
For what they call the "fair VR experience" you only need 60Mbit/s, but even that is more than twice that needed for 4K video.
My point about FB is only that if they push it, VR might really take of, whereas previously I didn't think it would.
Hmm Isn't everyone currently making retinal VR screens committed to not doing things that way, including facebook?
I could see it happening later on... But OTOH it represents a tradeoff between having rendering done in the datacenter and having it done on the user's PC/phone, usually the datacenter way, minus transmission costs, would be more efficient?
@faun I personally think local rendering could be more efficient, esp. if the data centre is far away. The transmission cost is likely to dominate. The worrying part is that these figures (64x in crease in BW) are introduced without any consideration for their impact on emissions. As I've shown, computing emissions are already growing totally unsustainably, and VR will certainly make this worse. And these figures show that it could be *a lot* worse.
I hope it doesn't happen, but hope is not enough.
As someone with 0 expertise I'd think actually desirable quality VR will be energy intensive no matter whether rendered locally or centrally. It's either bandwidth way above high quality video or a lot of local graphics processing. Which I doubt will be possible on today's relatively efficient smartphones. Maybe if one is happy with cartoonish graphics. But most likely technical possibilities will be pushed to woo customers, so high energy use. Centralized rendering could be chosen to get everyone in, even with slow/cheap client hardware. But in the end I have no idea. 🤷🏿♀️
@wim_v12e I've tried various VR systems over many years, including very expensive industrial ones, but I doubt that even Facebook will be able to make it go mainstream. The last time I tried a VR headset after a couple of minutes of novelty factor but feeling slightly stressed and dizzy I put it down and thought "this is something I actively don't want".
Augmented reality however has a much better chance, provided that they can get the hardware right.
@wim_v12e whither multicast?
maybe whereas video streaming is incredibly inefficiently done and yet not inefficient enough to do something about it, simultaneously wandering around 3D pictures might be enough to tip the balance?
@ve0hak I'm not sure about how much difference multicast can make as I have not looked into it. What could make a big difference is local rendering using accelerators. Nevertheless, I fear none of the tech improvements will be enough to offset the increase in emissions resulting from the projected growth.
@wim_v12e I'm not sure either. Doing the rendering centrally probably means that multicast for transmission wouldn't help much. Multicast would be efficient for sending geometries and textures that everyone needs to know about.
I'm also not sure about the relative cost of rendering vs transmission. Even transmission is as efficiently as possible, GPUs use a ton of energy, and that would basically be constant whether done centrally or at the edge.
So yeah, either way, I'm not optimistic either.
@Anke@social.scribblers.club @email@example.com @firstname.lastname@example.org @email@example.com This post and the replies to it show a clear lack of any experience even close to the field of VR. Of course Huawei would vastly overestimate bandwidth consumption, they manufacture network equipment after all.
So far, Facebook's Metaverse (which is incredibly awful for various reasons, but power consumption is not one of them) is intended to be used with their Quest headset which does all the rendering on-device and only needs short spikes of high bandwidth to load worlds).
Also @firstname.lastname@example.org standalone VR headsets like the Oculus Quest are literally cheaper than gaming PCs and consoles.
the mastodon instance at cybre.space is retired
see the end-of-life plan for details: https://cybre.space/~chr/cybre-space-eol