Geforcepat
[H]ard|Gawd
- Joined
- Jun 2, 2012
- Messages
- 1,338
Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Let's hope my 3080 doesnt give up the smoke for a few more years...Panic Buy Time.
I’m the idiot with the 10 year Evga warranty on my 3080. If they’re still around I’ll be set.Let's hope my 3080 doesnt give up the smoke for a few more years...
Either way still hot my old old gtx970 should the worse happen, because I'll be damned if I buy a scalper level price hell gpu.
It's more just the AI monster eating everything in its path. A bunch of companies are building datacenters as fast as they can to have the biggest, bestest LLM they can. It's not lie it is easy to scale up production of things like DRAM once you hit the limits of the factory, so prices are going up and the companies can pay because they have massive amounts of investor funds to waste.All of this feels like a concerted effort to push people towards game streaming. You'll own nothing and like it. Looking forward to 600 dollar 6060s.
Lot of the buyer are companies that can pay because of the massive amount of cash flow as well, they are not diluting themselve for this, they just slowing buyback a little bitso prices are going up and the companies can pay because they have massive amounts of investor funds to waste.
Some have been doing it for a long time by now, spending are growing but their revenues explodingAt that point the AI companies have to figure out how to be profitable, which so far they haven't,
It’s the typical startup model these days, hyper scale or whatever then try to figure out how to be profitable.I mean, I'm not saying the bigger firms like Google or MS aren't profitable, they are, I'm saying AI isn't profitable. For all available data it is a money furnace right now. OpenAI went through like $11 billion last quarter. At some point, that party comes to an end and they have to either increase prices, decrease outlays, or both.
And I'm not saying they won't (though I do have some skepticism) just that they will have to cut spending on that path. Unless people are willing to pay a LOOOOOOT more for LLMs than the current prices, they are going to have to cut expenditures. When that happens, stress on the computer supply chain will relax.It’s the typical startup model these days, hyper scale or whatever then try to figure out how to be profitable.
google being profitable but not AI can become a bit messy, pretty much everything that bring revenue that they do is powered by it and there recent explosion in revenues (at least they claim) is due to AI getting better at capturing attention and matching sellers with buyer with it.I mean, I'm not saying the bigger firms like Google or MS aren't profitable, they are, I'm saying AI isn't profitable. For all available data it is a money furnace right now. OpenAI went through like $11 billion last quarter. At some point, that party comes to an end and they have to either increase prices, decrease outlays, or both.
GPT 5 was a lot about cutting cost, we do not know exactly marginal cost per token of course, but the consensus is that they are being sold at a way higher price then marginal compute for inference cost already, the money burning is from the giant training phase/R&D now.And I'm not saying they won't (though I do have some skepticism) just that they will have to cut spending on that path. Unless people are willing to pay a LOOOOOOT more for LLMs than the current prices, they are going to have to cut expenditures
Hey man if i can't be a game dev or software dev any more i can easily work in sales or go back to serving tables in fine dining. I love making games but it's insanely stressful. I use Gemini 3 and GPT 5.1 every day. So far it's a productivity boost, but if it takes over it wont hurt me too too much. The other jobs sound far less stressful at this point. The odd thing is the entire economy will collapse and die if they're successful. The whole thing feels short sighted, but so is our economy.It's more just the AI monster eating everything in its path. A bunch of companies are building datacenters as fast as they can to have the biggest, bestest LLM they can. It's not lie it is easy to scale up production of things like DRAM once you hit the limits of the factory, so prices are going up and the companies can pay because they have massive amounts of investor funds to waste.
In the long run who knows what happens? Maybe LLMs will advance to the point where they just replace almost all workers (this is what the companies are betting on and about the only way this investment ever makes sense) and then, well, GPU prices probably won't matter because you won't have a job to afford one anyhow. Or maybe this will all come crashing down when the economy enters a recession and investors get spooked and pull their money. At that point the AI companies have to figure out how to be profitable, which so far they haven't, and as part of that they'll have to start buying drastically less hardware which should reduce the prices.
Chip makers are the winners in all of this. When there is a gold rush you don't pan for gold to not miss out - you sell shovels and pans.It's more just the AI monster eating everything in its path. A bunch of companies are building datacenters as fast as they can to have the biggest, bestest LLM they can. It's not lie it is easy to scale up production of things like DRAM once you hit the limits of the factory, so prices are going up and the companies can pay because they have massive amounts of investor funds to waste.
In the long run who knows what happens? Maybe LLMs will advance to the point where they just replace almost all workers (this is what the companies are betting on and about the only way this investment ever makes sense) and then, well, GPU prices probably won't matter because you won't have a job to afford one anyhow. Or maybe this will all come crashing down when the economy enters a recession and investors get spooked and pull their money. At that point the AI companies have to figure out how to be profitable, which so far they haven't, and as part of that they'll have to start buying drastically less hardware which should reduce the prices.
When I started this thread 4 years back, I anticipated something like the 9060xt 16gb to be affordable, right around this time (before a potential refresh with Super cards)
It is. Has finally dropped below MSRP. for now.
But immediate future looks bleak.
9060xt will last a long timeI'm seriously considering a 9070 or xt though I should probably just get a 9060xt 16gb.
Not saying that not the timeline, but if is mid-end of 2027 for the 10070 to came out, AMD will have went a full 5 years without clearly and cleanly beat the AD103, a 379mm-TSMC 5-256 bits of 2022.... that many called a 4070 in disguise.9070 / 9070xt will get replaced in 2 years time by the 10070
25% according to TPU bench if it make things feel a little bit better, they have the 6700xt at 80% of an 9600xt16GB (https://www.techpowerup.com/gpu-specs/radeon-rx-6700-xt.c3695)I just dislike buying a 128bit card and it is supposedly only 20% faster than a 6700xt.
What's the price difference?I'm seriously considering a 9070 or xt though I should probably just get a 9060xt 16gb.
In the next several years, there's probably gonna be one GPU from nVidia, one from Intel, one from AMD and you get to decide which subscription you want depending on how much power you want to unlockIm glad I got my XFX 7900 XTX, should be good for a while now. Doubt I will run into any problem with games for the next several years. Plus by then Ill be due for a whole system upgrade anyway.
This is exactly why I snagged my 9070xt back in early February; just wanted to get in while I could still stomach the cost. Knew the price increase was coming, but thought it'd take effect a bit sooner then what's reported above.Igor says, if you are planning on buying the 9070xt, then buy it immediately
Will Radeon RX 9070 XT Soon Cost 200 Euros More? Insights into Material and Manufacturing Costs, Cost Calculations, and Manufacturer Margins
22. May 2026 05:30
Igor Wallossek
The RX 9070 XT is currently less a normal product in price comparisons than an indicator of a delayed cost wave. The visible price apparently does not yet reflect the current FOB pressure everywhere. That is precisely why the card still appears comparatively attractive in May 2026, even though it is objectively not cheap. The real message is therefore not that graphics cards have become more expensive. The real message is that part of the price increase may not yet have fully reached the shelf.
https://www.igorslab.de/en/mid-range-gpus-2026-material-manufacturing-costs-margins-trends/
isn't H2 2026 cap-ex a new record high and by a lot ? Not sure anything has slowed yet.So far, there's no indication of stopping the eventual building of the AI datacenters, even though funding has slowed up.
Isn't nVidia including gaming sales as part of this "revenue" stream? That tells me they're doing everythign they can to artificially inflate the numbersisn't H2 2026 cap-ex a new record high and by a lot ? Not sure anything has slowed yet.
Data center is one group, everything else falls into a group bucket.Isn't nVidia including gaming sales as part of this "revenue" stream? That tells me they're doing everythign they can to artificially inflate the numbers
my guess is the party ends in 2028So far, there's no indication of stopping the eventual building of the AI datacenters, even though funding has slowed up. Even if AI isn't profitable, they're just going to rent out the hardware as cloud/compute. The best case scenario is somehow they just resell the hardware to consumers, but that'll be at least 2-3 years down the road. By then, newly released graphics cards costing $5,000 will be the norm, so you'll be paying $2,500 for a used four-year-old RTX 5070 Ti.
An AI consultant tells Axios that one of their clients recently spent half a billion dollars in a single month after failing to put usage limits on Claude licenses for employees.
https://x.com/edzitron/status/2060006709557690841?s=20
I think you are the first one to suspect Nvidia of using secret gaming revenues to boost their datacenter number.Isn't nVidia including gaming sales as part of this "revenue" stream? That tells me they're doing everythign they can to artificially inflate the numbers
Amount of compute per robots and how many of thoses there will be make this an very hard guess game, 2 intelligent robots per humans seem reasonable, the compute to train that fleet and to operate it being about 2x5080 per robots has well, thats 30 billions high end GPU worth of compute in the next 15 years just there.my guess is the party ends in 2028