You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
I suppose the idea is that either China changes behavior or they fall behind in chips (and consequently some other tech). The former doesn't seem very likely these days but the latter is also better than nothing.
What do you make of the decrease compared to August, @Daniel Nenni? Was the August number exceptionally high, is this seasonality or should we expect that the revenue will trend downward in the coming months?
The improvements (in time on a real GPU) are not very groundbreaking but the fact that we have AI discovering novel algorithms is impressive. That said, the solution seems to be quite specific to the matrix multiplication problem so I don't expect it will transfer to any other algorithm design...
I was thinking about the idea of deterrence by threatening to destroy TSMC for a while. Initially it seemed like an obvious good idea to me but now I'm much less convinced. After learning more about China, it seems to me that semiconductors are not their primary reason for wanting Taiwan, so...
I very much agree with you about "we need" in the sense that it would be very good to do it. But semiconductors are an important source of power and access to their manufacturing (or control of it) is a serious bargaining chip in the geopolitical game so I'm not optimistic about semiconductors...
This whole thread is great!
Thank you @jms_embedded for starting it (and further contributions) and to @Scotten Jones and @mozartct for all the information. I would also buy the book.
> It is far better for China to be a customer than try and take over Taiwan and the sooner they realize this, the better for everyone.
I'm not sure if the semiconductor industry is among the chief reasons for China to want Taiwan. AFAIK, this is a long story that started way before...
I think you are right. When labor gets more expensive, some additional jobs become cheaper to automate (than to have them done by humans).
Further, the pandemic makes factories more dangerous for humans, especially when there's high concentration of humans. If we replace some humans by...
Totally hypothesising here, since I have no idea, but still, perhaps someone will find this useful:
1. Taiwan Silicone Shield has multiple "pillars". One of them is US customers of TSMC, for whom TSMC is irreplaceable, so the US is more likely to defend Taiwan. Another one is Chinese customers...
Most organisms would grow uncontrollably, given an opportunity. In the nature populations are limited by resources or predators. Humans are actually the first species that can limit itself even though it could grow more. Right now the population is shrinking in many developed countries...
I largely agree with your conclusions but my perspective is somewhat different. In my opinion, humanity is not separate from but is part of the ecosystem. Sure, we have unreasonable powers, but we are still part of Earth's biosphere and part of the evolutionary process that runs within it. The...
If the decision to go to war was taken from the perspective of the whole humanity, wars would be quite rare indeed. However, often countries found themselves in lose-lose situations that looked like prisoner's dilemma on steroids, and the incentives were such that escalation was inevitable...
This does sound like a reasonable and simple plan. However, if Intel goes with it, they would be locking themselves into outsourcing cutting edge chips to TSMC for the foreseeable future. That in itself is not necessarily a bad idea, but they have just announced the whole IDM 2.0 thing, and...
I'd be interested in checking it out if you have it published somewhere. Don't have any FPGAs to try it on, but would be interesting to look at the code.
Yeah, that kind of how I was thinking about this: performance depends on the clock frequency, size of caches, size of pipelines, memory latency, out of order smart-assery that the CPU is doing, and god knows what else. To make matters worse different programs will run at different "speeds" and...
So von Neumann preferred random access memory (like in VN architecture) to Turing machines on efficiency grounds -- fair enough. Still, I don't see how this maps to RISC vs CISC debate: both of those work with random access memory. What am I missing?
Interesting move by Intel, I like it! I wonder if they are thinking that with the whole industry drifting towards ARM, RISC-V will be the next stop on the road to CPU unification and commoditization and they are trying to jump directly into the future.
Brutal. I really hope this doesn't repeat. Luckily, similar siege of Taiwan doesn't appear possible because other countries would likely interfere. My impression is that if CCP was to be successful, it would have to be quick. Anyway, this is probably off-topic for this forum.