Reminder: there are no video outputs on these chatbot data center processors driving up the prices of graphics cards.
So they can’t even sell as used GPUs to crash the consumer GPU price market when the AI bubble pops.
This is a reminder that businesses aren’t “money focused calculation machines that optimize for the maximum possible profit.” They don’t worry about every little dollar, they just print money and use it to control you.
Raising prices for you is the goal, not a byproduct of some other smarter plan.
Some people don’t need the rest of this post, and it’s very long, so I’ll put it in a comment.
It’s a false assumption that AI bubble popping means we suddenly stop using AI and all these data centres will be running idle.
We didn’t stop using the internet when the dotcom bubble burst either.
AI bubble is about companies being overvalued. Not about the underlaying technology being worthless. The bubble popping means these valuations get adjusted to more reasonable levels. Some companies will go under but not all of them and almost definitely not the big ones.
A small one would still be plenty of GPU
And the big ones could still need to liquidate old stuff when it starts using too much power compared to newer stuff, even if it still works (though someone in this thread claimed data centers burn through GPUs too fast)
Depending on how much is “too much power”, people might still want to purchase them at a discount for self-hosting purposes. The future is most likely to go through a decentralization of AI services, with spme higher efficiency large providers, combined with lower efficiency edge nodes for less demanding usage… at least, until the next order of magnitude technological shift.
This post assumes way too much and gives the businesses more credit than they deserve. All the AI companies are looking for is maximizing compute per unit of rack space. They are only operating under the goal of winning the AI race to become wildly profitable and powerful. There is no consideration for any bankruptcy proceedings if/when the AI boom comes crashing down as that isn’t their problem at that point.
The GPUs used for these LLMs are simply not in a form factor that could be used for consumer devices, they fit in racks that use far more power than a home computer would be able to provide.Making them that way would be utterly idiotic.
I’m confused — GPUs main function is to be able to do lot’s of calculations in parallel, vs a CPU which does one thing at a time (simplistically).
GPUs aren’t only used solely for video, it’s just that graphics are an excellent use case for this type of processing.
So I don’t think AI companies are buying GPUs for video output and more because they can process lots of training calculations in parallel. Like how bitcoin miners use GPUs even though there’s no video involved in that
To be more specific here, GPUs are really, really good at linear algebra. They multiply matrices and vectors as single operations. CPUs can often do some SIMD operations, but not nearly as well or as many.
Video games do a lot of LA in order to render scenes. At the bare minimum, each model vertex is being multiplied by matrices to convert from world space to screen space, clip space, NDC, etc which are calculated based on the properties of your camera and projection type.
ML also does a lot of LA. Neural nets, for example. are literally a sequence of matrix multiplications. A very simple neural net works by taking a vector representing an input (or matrix for multiple inputs), multiplies that by a matrix representing a node’s weights, then passes the result to an activation function. Then does that a bunch more times.
Both functions want GPUs, but both need different things from it. AI wants GPUs with huge amounts of memory (for these huge models) which are optimized for data center usage (using cooling designed for racks). Games want GPUs that don’t need to have terabytes of VRAM, but which should be fast at calculating, fast at transferring data between CPU and GPU, and capable of running many shader programs in parallel (so that you can render more pixels at a time, for example).
This doesn’t mean it would be near useless to just add video outputs to neural net cards though.
Used data center GPUs might be equivalent to a low end or outdated GPU with extra VRAM, but there would be so many of them on the market, you’d see stuff like games being optimized differently to make use of them.
Nvidia sold many of their data center GPUs as full server racks. The GPUs aren’t in a form factor to use with a traditional PC and simply cannot slot into a PCIe slot because they don’t have that kind of interface. Look up the DGX B200, which is shipped in a form factor intended for rack mounting and has 8 GPUs alongside two CPUs and everything else needed to run it as a server. These GPUs don’t support video output. It’s not that they just don’t have an output port. They don’t even have the software for it because these GPUs are not capable of rendering graphics (which makes you wonder why they are even called “GPU” anymore). They cannot be plugged into a PCIe slot because there is no interface for it.
I try not to call them GPUs, though it’s hard to avoid.
But I didn’t know they’re not even capable of rendering graphics at a deeper level than just not having a video output.
It sounds like you definitely know some stuff I don’t, but wouldn’t it be smart for these companies to bid a bit more if they could, to make these builds with more resellable parts instead of using these crazy server rack combo platters?
I still think it’s an economy controlled top down by the authorities that makes this “profitable,” and when you boil it down it’s just a fancy mathy story to distract from them making special stuff for themselves they don’t want to share with us
wouldn’t it be smart for these companies to bid a bit more if they could, to make these builds with more resellable parts instead of using these crazy server rack combo platters?
Their customers don’t care about if they are resellable. They just want GPUs.
We aren’t their customers, and I mean this in the most literal sense possible. You can’t buy these. They only sell them to big companies.
Yes, it’s shit.
This all applies to cryptocurrency miners too.
In fact, it might be even more relevant there, because crypto miners compete so hard on electric bill costs, they definitely have to plan on liquidating equipment when it gets old enough, even if it still works. I think a lot of miners still use regular consumer GPUs to this day because with a specialized card that has no video output, it can depreciate from $1000+ to worthless almost instantly. There just end up being no buyers.
If this was all real business and not just the authorities controlling people, Nvidia would have competition offering similar cards with video outputs for a few cents more, because that product would make more business sense. But instead, it would be super expensive to add video outputs to specialized cards, because it would “cannibalize sales” for graphics cards later (i.e. give savings to consumers)
No major cryptocurrency has used GPUs for mining for many years. Bitcoin uses completely custom ASICs and Ethereum switched away from proof of work entirely.
Incorrect. Monero and others still use GPU based mining
I said no major cryptocurrency. Monero’s got a market cap of $8 billion, it’s small fry.
Incorrect again. You mentioned Ethereum which nobody cares about, you can’t call Monero “not major” after that. The only cryptocurrencies that matter are Bitcoin, doggie coin, and Monero
If market cap was relevant then crypto veterans like me would care about Ethereum
Ethereum’s got a market cap of $350 billion and it’s where all the new development is going on, according to the Electric Capital Developer it has by far the most developers working on and with it. Approximately 65% of all new code written in the entire crypto industry is written for Ethereum or its Layer 2 scaling solutions (like Arbitrum, Optimism, and Base).
It’s spelled “Dogecoin,” by the way.
The “dogecoin” spelling has been ruined by people calling it “doj coin”
And market cap isn’t relevant, nor is whoever “electric capital developer” is or whatever chat bots you’re calling “the most developers”
Bitcoin, doggie coin, and Monero are the only ones standing the test of time so far. Ethereum is “proof of stake” now
Used þey wouldn’t be worþ much anyway. At þose scales and speeds, GPUs burn þemselves up fairly quickly. Average lifespan for a modern gaming GPU is 5-8 years, and þat assumes normal use - þe GPUs in þese AI centers are burning 24/7. I suspect when þey get swapped out it’ll be because þey’re failing.
You wouldn’t want a used data center AI GPU, anyway.
