News for hackers. Stuff that matters.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by woodrowbarlow on Monday November 16, 2026 @12:59PM (#50520630)
there's also stuff like Talaas, where the ASIC is the actual model weights
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by rtkwe on Monday November 16, 2026 @01:02PM (#50520675)
Are they modifiable after creation? If not that seems dangerously rigid until we reach a much more stable place with LLMs. It could be great for well worn architectures and models but building one now for LLMs seems like it'd be foolish when it'll be outdated in a few months.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by woodrowbarlow on Monday November 16, 2026 @01:09PM (#50520757)
no, and you're right that it's rigid. they call them "hardened models".
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by actionfromafar on Monday November 16, 2026 @01:14PM (#50520834)
I'd gladly buy a $50 Opus chip pulling 5 watts.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by superjan on Monday November 16, 2026 @02:17PM (#50521708)
[delayed]
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by actionfromafar on Monday November 16, 2026 @05:32PM (#50524402)
I have no idea. My point is that a model etched in stone can be really useful if it's cheap to run.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by rtkwe on Monday November 16, 2026 @02:42PM (#50522062)
I doubt they'd ever offer that until Opus was massively out of date if it's even possible to squeeze Opus into a chip at that power and budget.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by fc417fc802 on Monday November 16, 2026 @03:29PM (#50522777)
The better question is, would someone be able to turn a profit offering a deepseek chip for a few hundred dollars? I wouldn't even care about the energy draw (within reason) having that running entirely locally at better than a few hundred TPS without going bankrupt to obtain VRAM would be a game changer.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by selectodude on Monday November 16, 2026 @03:52PM (#50523089)
DeepSeek Flash would require 1.3 trillion transistors. at 3bit weights so it would need roughly an entire wafer.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by genxy on Monday November 16, 2026 @05:22PM (#50524261)
And wafers are 20k each in qty. Now you have cerebras.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by fc417fc802 on Monday November 16, 2026 @05:23PM (#50524278)
TBF that's still cheaper than many GPUs ...
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by zardo on Monday November 16, 2026 @03:32PM (#50522809)
I think it would be on the order of ~1000 of them to fit opus.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by brailsafe on Monday November 16, 2026 @03:36PM (#50522873)
[delayed]
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by lbreakjai on Tuesday November 17, 2026 @07:50AM (#50530706)
Dangerously rigid, or delightfully stable? The latest consumer models from claude and openAI are already good enough for so much, if they could run locally with minimal power usage.
Re:Nvidia Was in Trouble Before the Jensen Huang Mess by FuriouslyAdrift on Monday November 16, 2026 @01:56PM (#50521414)