![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Your assuming they all don't just say ya consumer hardware. Na.
Judging by what I read even this forums, there seems to be quite a few people OK with streaming......
Yeah. It's quite sad.
There are probably five people who lined up for such a future. Those people are sadists. Probably the same people who went and bought Qualcomm Snapdragon X laptops.
It's the same five people who just so happen to post in this forum.
It would be nice to think gamers on mass will say get bent.
Okay? Not so much, resigned is more like it.
Like for the B580 launch that did nothing, it would have been really hard for them to do so, when a 199mm-128bits gpu from AMD is almost 30% faster than the 272mm-192 bits GPU of intel (or the 204mm-128bits 7600xt and 160mm-128bits from Nvidia of the time matching it), they will have an hard time making actual volume of them at MSRP.
I think Nvidia would rather have a person subscribe to their gaming stream service for their entire gaming lifetime than sell a GPU to them every 2-5 years.
as a general rules, money in front is always better than over time, no risk and no business to run vs having to run a streaming business is always better and here if you sell a $1000 5080 every 5 years, once they pay for electricity instead of you, you do not make more money either way from them by charging them $20 a month (if they game 80 hours a month at $16 cent a kwh, that can be $14 net of electricity, $840 over 5 years coming their way instead of $1000 up front without counting the cost of the setup that need to amortized), the ability to moneytize those gpu in down gaming hours move things a bit, but not sure they mind either way.
There are many more people with GPUs less powerful than a 5080 that might (or maybe likely?) pay $20/month than people that bought a 5080.
yes that what customer deciding the market direciton would look like (target winning audience would be people not wanting to buys 5080s but would not mind renting the near latest best at good price point), a bit like many more people not wanting to buy $17 dvd when oyu can rent a big catalogue for less a month elsewhere, forcing studio to loose their best money printing machine they ever had (or the CD for the music industry)
Do you know how long cloud gaming has been around and everyone has said it's been ready? Before Geforce Now it was called Geforce GRID. Seems like gamers have gotten good at telling cloud gaming to get bent. It was such a failure that Nvidia reinvented it a number of times. Remember when GRID required you to buy games from them? Remember when that didn't work and games you bought from them would also work on Steam? Pepperidge Farm remembers.
They've been starving PC gamers out of hardware since 2016, so why is it different now? Again, the GTX 1060 was the #1 GPU on Steam for a long time. For
My prediction is different. Gamers will game on older hardware until the AI bubble pops and prices go back down to sane levels. Probably even cheaper because there's going to be a rebound effect, as well as new DRAM competitors. I'm not even sure if OpenAI can last by the end of 2026?
Spot on. Shit is going to crash hard. This is just a momentary spike in demand.
arent they not rising like 100 billions of capital in their latest round right now ?
Might also put a dent in future higher end GPU sales. Force people to game on low or medium for a while and they'll start realizing it's not that bad. This is fine, not going to pay up for "ultra" setting anymore. The old days when "low" meant short draw distance and textures looking like shit are long gone. We're well into the era of diminishing returns for more GPU power and higher settings. Chop off a gamer's e-peen with a long period of absurd vid card prices and it might not grow back.
AMD is already ahead of the curve. Last I checked the highest end consumer GPU they make is a 70 class card.
the biggest card will be the biggest card on the market (and what games will limit themselve too), when people bought a tiny 142mm GeForce4 Ti 4600 it still felt like an high-end card for them, same will be if Nvidia leave in 2032 and the biggest AMD/intel/apple will feel like high end to them.
maybe if that a way to avoid steam tax, but in the say geforce now world, why would they care if you play locally or streaming particularly ? and in the netflix type, that would be the first industry to push instead of resist with all they can that model, if people are ready to pay $80 (or more for the big one) on their local device what the gain for the publisher to be made to change that ? They will follow the market of course.
if it ever release (they could cancel this and concentrate on the upcoming Intel-cpu version of this), N1X is way too big of a chip to be sold as a thin streaming device (Geforce now will run perfectly fine on an recent amazon stick or old Intel laptop, no need for a powerful 5070 level gpu laptop), not sure Nvidia particularly wants things here, they will sell 100 millions switch 2 for local gaming gpu without minding (and would do the same with a more recent one the minutes capacity would allow it).
It is all relative. The entire market is affected.
They did make them for consumer devices. I agree they may never actually sell them. Initially I think they believed they would be a good in between. Get some pure Nvidia stuff out there. Jensen probably expected his software guys to find ways to sell it. Some local some streaming. You are probably correct though that now they are thinking they can just go full thin client.
You hit on yet another selling point for the game publishers.
think it’ll actually burst?
with how much work claude opus 4.6-chat gpt codex 5.3 can make unsupervised once you launch an agent into something (and ability to have many working together), like scientific models were doing for science task but for code.... 8 hours of running coherent task is now something that seem can happen, usage seem almost certain to still explode in 2026/2027. Combined with how long it take for the things being made to get online, hard to imagine a pre 2028 bust, would require some really large step function in token/watt of hardware (if we mean a bust in hardware demand to run it), so big that it make up for all the added usage reduced price/time will always add on the demand side of things.
Wasn't Nvidia's 100 Billions delayed because they're not sure if OpenAI can deliver? The whole point of this AI craziness is for AGI and it's not going to happen. Not through machine learning.
You may not even have to settle for low and medium. A lot of this is due to Ray-Tracing, and you could just turn that off. I'd rather have high quality textures then low quality textures with Ray-Tracing. Also, you can leverage upscalers and texture packs to get you visuals that even Ultra quality settings can't reach. You can always texture pack a game and get better visuals without losing performance.
People have always said it has never sucked and assured me it was usable. It's always been about lowering you expectations.
Google did with Stadia and failed. It doesn't get much bigger than Google.
Nvidia isn't the only company that makes GPU's. Let Nvidia sit on a lemon and twist.
Why would they do that? Especially by 2028 when we have no idea if the AI bubble will even last that long.
? Nothing was said about any delay, they seem to be doing 20 billions in the current round (the 100 billions plan was 10 billion by 1 gigawatt that was always over years pace), that Nvidia largest investment ever, not sure:
That not the whole point and not needed at all, an excellent programming tools or cells simulation worth many hundreds of billions does need to never in all domains make an error an human adult would not and able to do everything they do in all domain (if that one mean by AGI). Most task most robots will do will be way more limited than everything an human can do.
and what is called 80 class card can change over time a lot, there is a world where the 5080 was a 510mm-384bits-24GB vram card instead.
I think now most in blind test would not be able to tell which is which( i think it is a very common experience for people that do A/B testing with it, to at some point do an alt-tab to be sure they are not playiong the local version of the game, it was my case but i am really not a latency-high FPS sensitive person).
Isn't that money going right back to Nvidia after OpenAI buys Nvidia's hardware? If so that IS a good investment.
with money being 100% fungible, even if they would say otherwise yes it can be seen like that, but this is straight investment not a structured deal I think, i.e. openAI can use that cash to buy AMD or broadcom or microsoft/amazon stuff, it is the same round than Amazon would be putting 50B and softbank 30B, it is not an hardware credit but actual cash, that new structure could provide less scrutiny than the original plan, being an more regular pure equity investment (even if it could be as much if not more money going from nvidia->openAI than the previous deal in 2026).