Posted by todsacerdoti 7/4/2025
This isn't true. People were buying brackets with 10 series cards.
that's like they purposely not selling because they allocated 80% of their production to enterprise only
I just hope that new fabs operate early as possible because these price is insane
It became obvious when old e-waste Xeons were turned into viable, usable machines, years ago.
Something is obviously wrong with this entire industry, and I cannot wait for it to pop. THIS will be the excitement everyone is looking for.
> THIS will be the excitement everyone is looking for.
Or TSMC could become geopolitically jeopardized somehow, drastically increasing the secondhand value of modern GPUs even beyond what they're priced at now. It's all a system of scarcity, things could go either way.
If no good use is found for high-end GPUs, secondhand models will be like AOL CDs.
By your logic people should be snatching up the 900 and 1000-series cards by the truckload if the demand was so huge. But a GTX 980 is like $60 these days, and honestly not very competitive in many departments. Neither it nor the 1000-series have driver support nowadays, so most users will reach for a more recent card.
Also, it's not "a logic", it's not a cosumer recomendation. It was a fluke in the industry that to me, represents a symptom.
High-end GPUs are already useless for gaming (a low-end GPU is enough), their traditional source of demand. They're floating on artificial demand for a while now.
There are two markets that currently could use them: LLMs and Augmented Reality. Both of these are currently useless, and getting more useless by the day.
CPUs are just piggybacking on all of this.
So, lots of things hanging on unrealized promises. It will pop when there is no next use for super high-end GPUs.
War is a potential user of such devices, and I predict it could be the next thing after LLMs and AR. But then if war breaks out in such a scale to drive silicon prices up, lots of things are going to pop, and food and fuel will boom to such a magnitude that will make silicon look silly.
I think it will pop before it comes to the point of war driving it, and it will happen within our lifetimes (so, not a Nostradamus-style prediction that will only be realized long-after I'm dead).
This is the exact model in which WWII operated. Car and plane supply chains were practically nationalized to support the military industry.
If drones, surveillance, satellites become the main war tech, they'll all use silicon, and things will be fully nationalized.
We already have all sorts of hints of this. Doesn't need a genius to predict that it could be what happens to these industries.
The balance with food and fuel is more delicate though. A war with drones, satellites and surveillance is not like WWII, there's a commercial aspect to it. If you put it on paper, food and fuel project more power and thus, can move more money. Any public crisis can make people forget about GPUs and jeopardize the process of nationalization that is currently being implemented, which still depends on relatively peaceful international trade.
Dude, you're describing the 80s. We're in 2025.
GPUs will be used for automated surveillance, espionage, brainwashing and market manipulation. At least that's what the current batch of technologies implies.
The only thing stopping this from becoming a full dystopia is that delicate balance with food and fuel I mentioned earlier.
It has become pretty obvious that entire wealthy nations can starve if they make the wrong move. Turns out GPUs cannot produce calories, and there's a limit to how much of a market you can manipulate to produce calories for you.
There are satellites and ISR platforms taking images and data and data centres are processing that information into actionable targets.
Bombs that fly between continents or are launched from submarines for any "big scale" war.
From a market perspective, LLMs sell GPUs. Doesn't even matter if they work or not.
From the geopolitical tensions perspective, they're the perfect excuse to create infrastructure for a global analogue of the Great Firewall (something that the Chinese are pioneers of, and catching up to the plan).
From the software engineering perspective, LLMs are a nuissance, a distraction. They harm everyone.
Really? What about textures? Any ML that the new wave of games might use? For instance, while current LLMs powering NPC interactions would be pretty horrible, what about in 2 years time? You could have arbitrary dialogue trees AND dynamically voiced NPCs or PCs. This is categorically impossible without more VRAM.
> the perfect excuse to create infrastructure for a global analogue of the Great Firewall
Yes, let's have more censorship and kill the dream of the Internet even deader than it already is.
> From the software engineering perspective, LLMs are a nuissance, a distraction. They harm everyone.
You should be aware that reasonable minds can differ in this issue. I won't defend companies forcing the use of LLMs (it would be like forcing use of vim or any other tech you dislike), but I disagree about being a nuisance, distraction, or a universal harm. It's all down to choices and fit for use case.
Do not mistake adjacent topics for the main thing I'm discussing. It only proves my point that right now, all silicon talk is bullshit.
This is not the case if you want things like ray tracing or 4K.
Not for me. I prefer Intel offerings. Open and Linux friendly.
I even hope they would release the next gen Risc-V boards with Intel Graphics.
NVIDIA Keynote from the upcoming RISC-V Summit China: "Enabling RISC-V application processors in NVIDIA compute platforms"
And you can build mythologies around falsehoods to further reinforce it: "I have a legal obligation to maximize shareholder value." No buddy, you have some very specific restrictions on your ability to sell the company to your cousin (ha!) for a handful of glass beads. You have a legal obligation to bin your wafers the way it says on your own box, but that doesn't seem to bother you.
These days I get a machine like the excellent ASUS Proart P16 (grab one of those before they're all gone if you can) with a little 4060 or 4070 in it that can boot up Pytorch and make sure the model will run forwards and backwards at a contrived size, and then go rent a GB200 or whatever from Latitude or someone (seriously check out Latitude, they're great), or maybe one of those wildly competitive L40 series fly machines (fly whips the llama's ass like nothing since Winamp, check them out too). The GMTek EVO-X1 is a pretty capable little ROCm inference machine for under 1000, its big brother is nipping at the heels of a DGX Spark under 2k. There is good stuff out there but its all from non-incumbent angles.
I don't game anymore but if I did I would be paying a lot of attention to ARC, I've heard great things.
Fuck the cloud and their ancient Xeon SKUs for more than Latitude charges for 5Ghz EPYC. Fuck NVIDIA gaming retail rat race, its an electrical as well as moral hazard in 2025.
It's a shame we all have to be tricky to get what used to be a halfway fair deal 5-10 years ago (and 20 years ago they passed a HUGE part of the scaling bonanza down to the consumer), but its possible to compute well in 2025.
Dude, no one talks about this and it drives me up the wall. The only way to guarantee modern CPUs from any cloud provider is to explicitly provision really new instance types. If you use any higher-level abstracted services (Fargate, Cloud Run, Lambda, whatever) you get salvation army second-hand CPUs from 15 years ago, you're billed by the second so the slower, older CPUs screw you over there, and you pay a 30%+ premium over the lower-level instances because its a "managed service". Its insane and extremely sad that so many customers put up with it.
It's also way faster to deploy and easier to operate now. And mad global, I've needed to do it all over the world (a lot of places the shit works flawlessly and you can get Ryzen SKUs for nothing).
Protip: burn a partition of Ubuntu 24.04 LTS which is the default on everything and use that as "premium IPMI", even if you run Ubuntu. you can always boot into a known perfect thing with all the tools to tweak whatever. If I have to even restart on I just image it, faster than launching a VM on EC2.
But I do spend a lot of effort finding good deals on modern ass compute. This is the shit I use to get a lot of performance on a budget.
Will people pay you to post on HN? How do I sign up?
The spoiled gamer mentality is getting old for those of us that actually work daily in GPGPU across industries, develop with RTX kit, do AI research, etc.
Yes they’ve had some marketing and technical flubs as any giant publically traded company will have, but their balance of research-driven development alongside corporate profit necessities is unmatched.
Also, nobody ever said they hate their researchers.