Hacker News Reader

My question is why Apple would spend sweat and blood trying to make an “extreme” large chip for a very small handful of customers. To me, the economics don’t add up. The Ultra was already going beyond what they need and that at least was sensibly reusing multiple Max dies. Someone describe the effort:reward ratio to me.

View on HN

> 6. M3 Extreme is going to be insanely power, without any E cores. This chip's integrated GPU will likely be neck and neck with Nvidia's flagship desktop GPUs.

NVIDIA has a market cap of $2.31 trillion.

And to compete against them Apple would need to make a separate GPU, not an integrated GPU.

The claim isn't plausible anyway. A single chip can't expend as much energy as a flagship desktop GPU, and Apple doesn't have some 10x more efficient method of computation.

It could have a use case in the running LLM interfaces so Apple could probably start scoping the possibility of going after NVIDIA on the server side.

I read somewhere (probably here) that apples a device company and so it makes sense for them to focus on device based AI.

With that in mind, having an apple ecosystem where you can prototype and train large models to then pack down to smaller apple devices makes a fair amount of sense.

Reply on news.ycombinator.com