Metaverse Vision Requires 1000x More Computational Power, Says Intel (intel.com) 79
Leading chip-maker Intel has stressed that building Metaverse -- at scale and accessible by billions of humans in real time -- will require a 1,000-times increase in computational efficiency from what we have today. Insider reports: Raja Koduri, a senior vice president and head of Intel's Accelerated Computing Systems and Graphics Group, said that our computing, storage and networking infrastructure today is simply not enough to enable this Metaverse vision, being popularized by Meta (formerly Facebook) and other companies. "We need several orders of magnitude more powerful computing capability, accessible at much lower latencies across a multitude of device form factors," Koduri said in a blog post. To enable these capabilities at scale, the entire plumbing of the internet will need major upgrades, he added.
Of course they would say that. (Score:2)
Then fudge the benchmarks.
Re: (Score:2)
Then fudge the benchmarks.
The elephant in the room in the "metaverse". Is anybody even asking for it?
Re: (Score:2)
Metapurse, that's one of those European man-purses, right? Is that the one that's like a sporran made from Metasequoia bark?
Everybody is going to want a Metapurse.
AKA spendmore (Score:2)
Re: (Score:1)
They all do it. Replacement Replacement Replacement! [slashdot.org] (cue thrown chair)
Re: (Score:1)
They all do it. Replacement Replacement Replacement! [slashdot.org] (cue thrown chair)
Yeah, nobody on Slashdot ever gets rid of a CPU because a 1.5% faster one came out. Nobody here, in Nerd Mecca, ever replaces perfectly good equipment in a quest to squeeze out a bit more performance. It's all a conspiracy by corporations..
Re: (Score:2)
Re: (Score:1)
Gaming may be different; speed matters here and now. But Intel is scaring people to be "ready" for future AI/AR that's allegedly a CPU hog sight unseen. "It's the new thing that everybody will want" is different than a clear existing need. A variation is getting apps and OS's ready for fingers and tablets, yet cubicle people mostly still use the mouse. Or being ready for "web scale" via microservice and other bloatifiers even though you still work for "Jake's Discount HVAC and Tires".
Re: (Score:1)
Gaming may be different; speed matters here and now. But Intel is scaring people to be "ready" for future AI/AR that's allegedly a CPU hog sight unseen. "It's the new thing that everybody will want" is different than a clear existing need. A variation is getting apps and OS's ready for fingers and tablets, yet cubicle people mostly still use the mouse. Or being ready for "web scale" via microservice and other bloatifiers even though you still work for "Jake's Discount HVAC and Tires".
So let me get this straight.. You don't think VR worlds, with possibly billions of people using them, will require more CPU power than we currently have?
Re: (Score:1)
What I really question is that everybody will want it badly. Maybe someday they will finally catch on when enough aspects fall into place, but it's hard to say when that will be. It's a higher-res take on MS-Bob.
Re: (Score:1)
Fair enough.
I certainly won't be participating in it. I enjoy real-life too much.
Re: (Score:2)
A struggling tech company says we must replace everything.
I wonder if they might have a vested financial interest in selling us stuff.
Of course, they are going to be the part.
Yet this is not crucially essential. Mankind can continue its (and world's) digital conversion, this is pretty straightforward. Albeit, both producing, and also keeping in use that, will end in further batches of pollution and heat waste. Progressing into collapse.
Re: (Score:1)
Y'all have strange definitions of "struggling"
Intel revenue for the quarter ending September 30, 2021 was $19.192B, a 4.69% increase year-over-year.
Intel revenue for the twelve months ending September 30, 2021 was $78.474B, a 0.48% increase year-over-year.
Intel annual revenue for 2020 was $77.867B, a 8.2% increase from 2019.
Intel annual revenue for 2019 was $71.965B, a 1.58% increase from 2018.
Intel annual revenue for 2018 was $70.848B, a 12.89% increase from 2017.
I'm struggling to understand how anyo
Re: (Score:2)
True, but that doesn't mean they're not right.
Just Paint it Green! (Score:5, Insightful)
Re:Just Paint it Green! (Score:5, Funny)
Fascinating that the current sacred cows of tech hype all are based on vast amounts of CO2 producing energy consumption: Growing marijuana hydroponically when most places in the world it literally grows like a weed, crypto-currency mining, and the Metaverse. If only we had a proper economic model ... maybe 'Stupidity Offsets'.
Dude, I hope you are selling, because after a week of log4j wrangling, I am buying.
Either the 'Stupidity Offsets' or the weed, I don't care, both will work. Hit me up.
Re: (Score:2)
Re: (Score:2)
What do they think will happen when one of the real agriculture states in the midwest opens the flood gates and a few hundred thousand acres gets shifted back to hemp prod
Re: (Score:2)
Yeah... shitty weed that is full of seeds. Hydroponics REDUCES the amount of fresh water required to grow, reduces the amount of material needed by producing far higher potency output and reduces the carbon footprint. If you are going to complain about something complain about the growlights and odor control filtration systems.
Cryptocurrency has power requirements which are about 1/8th the traditional banking and
And for what? (Score:2)
Re: (Score:2)
It's better than the real world. I only stay in that because nobody has any worthwhile alternatives on offer.
Re: (Score:2)
Then you have a world where you can buy that crappy run down couch at the yard sale and apply a skin that you and others see through your augmented view. You solid 4 girlfriend becomes a solid 10 as viewed by you through your glasses. You have virtual laser lights project on things you are cutting and demonstrate where to cut when you are learning to filet a fish or
If Moore's Law were still applicable (Score:2, Insightful)
Maybe my count's off, but that'd be about 8-10 years worth of doubling every 18-24 months...so...about as far off as cold fusion, right?
Re: (Score:2)
That may be just long enough for Zuckeface's battery to die and the "metavision" hopefully along with that.
Correction (Score:3)
Re: (Score:2)
What has Moore's law got to do with it? he is talking about computational power not transistor density.
Generally, the computational power does scale with density. So for these kind of back of the envelope calculations, it is a very good startingpoint.
Re: (Score:2)
No. Fusion, PERIOD, is PERPETUALLY 10 years away.
COLD fusion is perpetually more than 50.
Re: (Score:2)
Fusion is stuck at 10 years away because the funding for it is being cut in real terms at the same rate as technology is improving. If governments actually invested seriously in fusion research, instead of joke token payments, we'd have it by now.
Cold fusion is a fantasy.
Re: (Score:2)
No. Fusion, PERIOD, is PERPETUALLY 10 years away.
COLD fusion is perpetually more than 50.
Shit, you're right! My bad.
Metaverse aint gonna happen (Score:4, Insightful)
Start planting trees, instead of planning for a hellish future FFS
Re: Until you get rid of the Democrats... (Score:2)
Re: (Score:2)
He doesn't think, that's why he blames covid on the one group of people actually dealing with the problem rather than on those making it thousands of times worse than it need be.
accessible by billions (Score:5, Insightful)
Would it help if some of us opt out? I'm willing to do my part. I've been reasonably happy with the physical world and the company of live people. Not sure I'll be needing an artificial world or fake people.
Re: (Score:2)
Why would you want to miss out on snow-rash-19
Ahhh, Good Ole Facebook... (Score:3)
If Bitcoin can use that much power, why not Facebook. Right?
One thing that Mark Z is always really good at is being relative to the rest of the world. He dosesn't lead anything. He just looks at the world and says, "Well, if I do just 1% less worse than the worst offender, then I'm not terrible, right?"
--
"A squirrel dying in front of your house may be more relevant to your interests right now than people dying in Africa." - Mark Zuckerberg
Re: (Score:2)
... and bring World Piece at the same time?
Do you mean World Peas, since they want everybody to go vegetarian?
Re: (Score:2)
You thought IBM wanted... everybody to go vegetarian? lololol
You're a little confused about the culture of those guys from NY in bright white shirts, ties, and shiny shoes.
Re: (Score:2)
Meanwhile, IBM is the #1 AI Cloud vendor, with a $60B market. But they don't call the software Watson anymore.
Maybe the name change confused you?
Shocking! (Score:2)
Company That Sells Processors Says More Processors Are Needed!
Who would have thought!
Re: (Score:2)
Doesn't mean they're wrong. Doesn't mean they're right, either - if the American government got rid of all the cable and telephone companies, switched the entire nation to fibre to the home and shifted 10% of their fossil fuel subsidies into upgrading the entire telecoms infrastructure to something from this millennium, you'd achieve the required network latency.
Foveated rendering (Score:3)
Someone clue these fools in on foveated rendering. Future VR headsets will use foveated rendering, which means only about half a megapixel needs to be rendered photorealistically as long as the frame rate is over 120fps. If you do not believe me, try to read this paragraph without moving your eyes.
Re: (Score:2)
Didn’t Samsung say they were working on lenses with this design like 3-5 years ago
From what i remeber i was sold on that.
Re: (Score:2)
Oculus already provides foveated rendering options in their SDK, and many games already use it (I noticed it while playing a game and happened to look at the extreme periphery you can see the much lower detail and blockiness of the lower resolution. So its already here, right now, the future versions are just to provide eye tracked foveated rendering so it applies to everything except where you are actually looking instead of a general periphery area
Second Life, Active Worlds, etc. (Score:3)
We were doing the "metaverse" thing 15-20 years ago...
Re: (Score:2)
Facebook has about 3 billion active users. They won. They are one of the most successful things ever. This is just Mark Zuckerbergs way of saying
I SPENT 2 BILLION DOLLARS ON OCCULUS! ! ! WTF was I thinking?
Take Occulus and shoehorn it into EVERYTHING.
Re: (Score:2)
Most companies are one trick ponies that lucked out, rather than actually understanding why they were successful at all. Which is why they buy other companies of success. But in many cases they end up ruining the companies they buy because they don't understand why those were successful as well.
Re: Second Life, Active Worlds, etc. (Score:2)
Re: (Score:2)
Fusion is only delayed because nobody spends any meaningful money on it. It's not pocket-money science. It costs serious dollars to build a reactor of suitable size, superconducting electromagnets of both sufficient strength speed to be able to dynamically control plasma, sufficient sensors of sufficiently high resolution (that also won't fail under fusion conditions) to be able to see what every last fraction of the plasma is doing, and a computer of sufficient speed to be able to process that kind of volu
Pushing the boundary of physics (Score:5, Interesting)
1000x efficiency in computational efficiency is almost like reaching the limit of physics. Everytime, there is a bit flip, the electron energy level transition happens which results into loss of energy of about 1 eV. So we lose about 1eV of energy. So the best we can get is 6 exa bit flips per Joule. But this is too idealistic condition. Semiconductor doping is not uniformly spread. So you need minimum about 5-10 electrons per transistor. 1 multiplication need n^2 bit flips where n is the word size. So 32 bit multiplication needs 1000 bit flips. With all these, the best we can get 100-200 terra flops per Joule. We are already getting over 100 giga flops per J. So 1000x efficiency is almost impossible in classical semiconductor based computers unless someone finds a way of capturing energy released when elecron jump to lower energy level.
Re: (Score:2)
Re: (Score:2)
Nah he is only going to use them as a power supple /smirk
Re: (Score:2)
Well, that's only if the electronics have to be 1000x as efficient. That's not actually a requirement here.
Let's say that instead of using general-purpose microprocessors on standard silicon, you use processors that are optimally designed for this specific workload and aren't designed to be used for anything else, using 7N pure silicon-28 (which you can do cheaply with small linear accelerators) or equally purified GaAs. You can also potentially hard-wire some of your low-level special-purpose libraries as
Re: (Score:1)
It is also ridiculous to say it's a requirement for Metaverse.
Metaverse can be implemented today - infact a version of it was implemented 18 years ago with Second Life, which was itself inspired by SnowCrash.
The requirements for reasonable performance are mostly on the client side - the server side only has to provide models and texturing info, and a relatively little amount of live stream info for position and poses of avatars. Even a raspberry pi has enough processing capability for basic 3d models - and
Re: (Score:2)
1000x efficiency in computational efficiency is almost like reaching the limit of physics.
This is Intel we're talking about. Samsung just announced new chips that could have phones run for a week on a single battery charge. Intel just announced ones that can solve the European gas crisis by doubling as room space heaters.
Faster CPUs don't mean much (Score:2)
Easy... (Score:2)
Just stop programming by using libraries that use libraries that use...
Seriously, the amount of bloat in modern software is crazy, all because processors got so fast that we didn't have to care. Using unnecessary libraries also opens a wider attack surface (cough, log4j, cough).
Re: (Score:2)
That wouldn't be so bad if optimisers were moved from compilers to static linkers. If you delayed optimisation to that point, and the optimisers were good enough, you could flatten the code everywhere it would give a speed gain.
Also, there's nothing that stops some low-level libraries being moved into silicon. How much does BLAS actually change, these days? It shouldn't be difficult to port BLAS to a hardware language (is SystemC good enough?) and have it as a co-processor or on a card. There may be other s
Billions of humans? (Score:2)
It's gonna be another Second Life clone, except with strict censorship, so all the cringe, none of the appeal. I'l be surprised if they break 5 million users.
Re: (Score:2)
Only governments can censor, same way only governments can impose taxes even though private companies can charge subscriptions. They may serve similar roles, but the private and public sectors are fundamentally different places.
Re: (Score:2)
Only governments can censor,
What kind of alternate reality do you exist in? Are you a Roblox character maybe?
Metaverse? (Score:2)
What IS the metaverse? (Score:2)
A few weeks ago it was just the new aspirational name for the company formerly known as Facebook's product portfolio. But all of a sudden un related companies are now talking about it as if its ubiquitous and everyone already knows what it is. This article makes me think Intel thinks metaverse is the freaking matrix or something.
Software engineering (Score:2)
Leading chip-maker Intel has stressed that building Metaverse -- at scale and accessible by billions of humans in real time -- will require a 1,000-times increase in computational efficiency from what we have today. Insider reports:
So, what they are saying is that we have the hardware for it but not the software engineering skill. Think about it, our computers are millions of times more powerful than they were 30 years ago, but we only have a hundredfold increase in actual usability. There is a thousandfold increase possible in compute power if we learned how to code more efficiently. Strike that. We do know how to code more efficiently, it just takes more effort than the CEOs are willing to pay for.
Laughing all the way to the bank (Score:2)
Am I the only one baffled by this vision? (Score:2)
So We Aren't Matrix Ready Yet (Score:2)
Not enough computational power yet to put everyone in The Matrix.
This is assuming of course that we are not all already in The Matrix.
Trust Moore law (Score:2)
There are better uses of those CPU cycles (Score:2)
Instead of even trying to make this vision of an alternative immersive world championed by somebody severely lacking social skills and seems to have misread some sci-fi (hint, many really are dystopian, not utopian), let's try something else, like more supercomputers for science and research?
We don't need to be connected more, we need proper connections to the communities we are in and to learn more tolerance and acceptance of differences, not to escape constantly into an echo chamber.
It's way too easy to f