alt Hacker NewsEdit: it seems like this is likely one chip and not 10. I assumed 8B 16bit quant with 4K or more context. This made me think that they must have chained multiple chips together since N6 850mm2 chip would only yield 3GB of SRAM max. Instead, they seem to have etched llama 8B q3 with 1k context instead which would indeed fit the chip size.
This requires 10 chips for an 8 billion q3 param model. 2.4kW.
10 reticle sized chips on TSMC N6. Basically 10x Nvidia H100 GPUs.
Model is etched onto the silicon chip. So can’t change anything about the model after the chip has been designed and manufactured.
Interesting design for niche applications.
What is a task that is extremely high value, only require a small model intelligence, require tremendous speed, is ok to run on a cloud due to power requirements, AND will be used for years without change since the model is etched into silicon?
Replies
Where are those numbers from? It's not immediately clear to me that you can distribute one model across chips with this design.
> Model is etched onto the silicon chip. So can’t change anything about the model after the chip has been designed and manufactured.
Subtle detail here: the fastest turnaround that one could reasonably expect on that process is about six months. This might eventually be useful, but at the moment it seems like the model churn is huge and people insist you use this week's model for best results.
A lot of NLP tasks could benefit from this
> What is a task that is extremely high value, only require a small model intelligence, require tremendous speed, is ok to run on a cloud due to power requirements, AND will be used for years without change since the model is etched into silicon?
Video game NPCs?
I'm thinking the best end result would come from custom-built models. An 8 billion parameter generalized model will run really quickly while not being particularly good at anything. But the same parameter count dedicated to parsing emails, RAG summarization, or some other specialized task could be more than good enough while also running at crazy speeds.
Alternatively, you could run far more RAG and thinking to integrate recent knowledge, I would imagine models designed for this putting less emphasis on world knowledge and more on agentic search.