Earlier quoted context omitted.
Last time I did the math, by the time I spent 4090 money on compute, I could have been done with a PHD. Now, if all you need is a 3060? Yeah you could probably make that work, but you wouldn't be doing LLM inference even quantized
Either you only use the GPU sporadically or your math is very different from Tim Dettmers` [1]: "The break-even point for a desktop vs a cloud instance at 15% utilization (you use the cloud instance 15% of time during the day), would be about 300 days ($2,311 vs $2,270):" And that was written before the current GPU crunch and assumes availability in the cloud, which currently is not a given at all. [1] https://timdet…
TSMC warns AI chip crunch will last another 18 months
101–107 of 107 posts
Re: TSMC warns AI chip crunch will last another 18 months
#102I'd bet $50 in 18 months "the crunch" will still be in full swing (and Raspberry Pi will still be "bringing huge volumes to the market any week now"). There is no way in hell all these expensive fabs will pay for themselves if chips are too cheap. So no one (especially not TSMC) will risk oversupply by building too many fabs.
The Raspberry Pi wouldn't have issues if not for the fact that they are explicitly an outlet for Broadcom's excess chips, and they have never paid full price for their CPUs. If they doubled the price of the boards, they would be able to get the parts.
Re: TSMC warns AI chip crunch will last another 18 months
#103Earlier quoted context omitted.
Why don't they? The current market price is 3 or 4 times the MSRP and has been for years--the Raspberry Pi has lost its place as the affordable hobbyist board. I suppose it would be admitting defeat, but if they'd actually pay for parts and make new Pis again, I'm sure they could at least bring the market prices down.
> the Raspberry Pi has lost its place as the affordable hobbyist board I keep hearing this - but I've seen no evidence of any successor board. Which board would you say is now the go-to board for hobbyists?
The price/performance (+size) is about the same as for the Pi, the power draw is not as bad as with older Intel/AMD machines, and your I/O options and x86 software compatibility is vastly better.
The only drawback is that you're buying used instead of new, and that your starting price for a small home lab (albeit one that starts at 2-3x the power of a single Pi) is larger, which harms students and other makers on a shoestring budget.
Re: TSMC warns AI chip crunch will last another 18 months
#104Earlier quoted context omitted.
Yes but it’s hard to disentangle the effect of the crypto crisis ending last year, the pandemic boom in consumer hardware sales coming to a close, and the pricing of Ada from one another. So many GPUs were sold to miners and made their way to the secondary market that gamers have lots and lots of options that aren’t new sales.
Maybe. But the traditional gaming market doesn't have a lot of growth and I have to believe the high-end may be a dedicated niche but it's a niche.
Re: TSMC warns AI chip crunch will last another 18 months
#105Earlier quoted context omitted.
RPi's are freely available nowadays. https://rpilocator.com/
More available than before, yes. But still heavily constrained. In the US, there's no CM4 anywhere, and spotty availability of RPi4. CM4s are still being sold for hundreds of dollars on eBay.
Re: TSMC warns AI chip crunch will last another 18 months
#106Earlier quoted context omitted.
From my undertanding, Fat Man requires a very precise design to focus the shock wave properly, or the bomb will never detonate properly. This design is quite non-trivial even today. It took several of the brightest minds of all time to get it just right back then (like von Neumann). Even with modern tools, you still need a team of quite capable engineers/scientists to pull something like that off, even if you do have…
Totally agree, the science / design part of it is still not trivial. Also I know next to nothing about it. The machining bit so, uaing proper machinery, is easy. And the necessary engineers and technicians readily available.
What makes it easier, though, is that the general principle is relatively well known, so one could always brute force the solution using computer simulations, even if unable to do the math by hand.
Re: TSMC warns AI chip crunch will last another 18 months
#107Time to open source the chip making industry... ... ... This whole fabrication thing is a biiig chokepoint.
This is exactly the kind of thing that will never be open source. Any entity with the extremely high level of resources to fabricate chips at this node have zero incentive to open source anything, and no number of amateurs with 3D printers can change that.