upvote
Not defending Nvidia here, but the RTX 50 series was released in 2025, was it not?

Regardless, they aren't planning on sticking to the two year release cadence any more which does suck.

reply
RTX 5090 - release price $2000, average price today $6500

RTX 5080 - release price $1000, average price today $2300

Nvidia also stopped selling its own version of these cards directly to consumers and is happy with letting partners and resellers/scalpers control the price.

So it is pretty obvious that the company doesn't particularly care about this market segment.

reply
> So it is pretty obvious that the company doesn't particularly care about this market segment.

Which I think is foolish, every AI shop is making their own alternatives, it's not going to be a winning bet for them.

reply
Ironically, if consumer GPU progress stalls, then slowly PC becomes the best console. Everyone has the same top-end GPU that can run any game.
reply
Or game streaming actually becomes a thing. Because the best new games only play in the cloud on computers we could never afford.

Almost like going back to the Arcade era except we all pay monthly subscriptions instead of 1 quarter at a time.

reply
Game streaming isn't going to happen. It's been tried several times now and failed every time.
reply
I guess gamers are tech savvy enough that they care about ownership on principle? Or why is it that game streaming hasn't taken off?
reply
Almost every gamer can immediately tell the difference in fidelity and latency. And only a very small handful of gamers have the internet quality for game streaming at low enough latency and high enough fidelity to be “sort of on par”.

The quality and internet connection stuff will improve over time. The latency problem essentially never will.

reply
I can see a world where the edge is built out a la Netflix. https://openconnect.netflix.com/en/
reply
GPU-style platforms (lots of compute connected to a big shared pipe that connects to a large pool of fast memory) are inherently more efficient for centralized inference.

Local inference gets truly interesting when the proper next-gen, natively AI and not just a repurposed GPU, stuff starts to land. Initially things like compute-in-memory and fancy flash, and later possibly the truly exotic stuff like various RMM (read-mostly-ram, for example stacked FeFET on logic) tech.

IMO there is no chance for ubiquitous local AI to happen if that means people using GPUs. But the industry is chasing all possible solutions for more efficient inference platforms, and if something like pachinko machine architecture with FeFETs ends up winning, suddenly buying a local ai inference solution in a box is no longer as insane.

reply
If consumers were willing to pay more for GPUs like datacenters do, NVIDIA would have sold plenty of consumer GPUs. Or would you press a button to halve your revenue for some reason?

(I say this as someone who would have liked a 5090...)

reply
> Hoping AMD steps up to the plate

Nvidia has nothing close to the Strix Point, Strix Halo, Gorgon Halo, etc.

My GPD pocket 4 handheld can run Qwen 3.8 27b only because of AMD making embedded mobile inference cheap.

reply
> Nvidia has nothing close to the Strix Point, Strix Halo, Gorgon Halo, etc.

Is there a reason why DGX Spark doesn't qualify? (genuinely curious)

reply
> Hoping AMD steps up to the plate

AMD's aggressive marketing as an AI-first company leaves me pessimistic about that.

reply
I'm running qwen3.8-flash-next at 4-bit quantization on two AMD Radeon AI PRO R9700 cards at 75 tok/s. And I can actually run two models in parallel at around 50 tok/s, so you can get by with only one card. I switched from NVidia to AMD for a fully open-source stack for my local inference.

AMD Radeon AI PRO R9700 is $1600 right now (I bought them last year for $1300).

They are also quite power-efficient. My setup typically uses around ~300W during inference.

reply
No gut feeling necessary. They announced weeks ago that they are going to be $1200-$1800 (Xbox specifically, PS6 somewhere in that range too).
reply
Not to mention the required monthly subscription... Income inequality like 1929, might as well bring sharecropping back.
reply
Yes, ten the home AI market would be similar to the serious PC gaming market used to be when they were the main market for the latest GPUs.
reply
> we all could be running better models locally by now

Apple silicon says "hold my beer", which I know because i've been doing this on my 3 year old quite beefy mac studio, for example. it's good enough for many tasks.

reply
Can't wait for M7! I hope the rumours are true.
reply