3 ms·
> Chips are too imporant to modern war for any country to be unable to make them The chips being used in weapons systems are largely fabbed on legacy nodes (28
by alephnerd 2y ago
> Chips are too imporant to modern war for any country to be unable to make them
The chips being used in weapons systems are largely fabbed on legacy nodes (28nm/48nm) or 14nm at the latest.
You don't need much more than a mid-2010s comparable Intel i5 or Nvidia Maxwell for most defense or weapons systems applications.
- bluGill 2y agoToday but will that continue? We don't know. If ai really works you may need better in the future. Only time will tell
- alephnerd 2y agoWhat is a use case that justifies a Radeon RX 5000 GPU versus an Nvidia Pascal? The same mindset you are describing is a major reason why Deepseek was able to "disrupt" the ML space. At some point, just throwing more and more compute becomes a waste. When building weapons systems, what matters is low unit costs and ease of replaceablity. This is the a major reason why most Defense (as well as Aerospace and Automotive) buyers tend to consolidate on older, cheaper, and more commodified processes There are significant usecases for bleeding edge compute such as in Simulations (heavily used in Nuclear Research and Jet Engine design), but you can fight using legacy compute at scale and still turn the tides. Look at both Russia and Ukraine for example.
- bluGill 2y agoIf it wins the war it isn't a waste. Sometimes we can do better sometimes a lot of processing is needed. I refuse to limit the future which makes your question a stupid question. Maybe nothing will come up maybe it will.
- alephnerd 2y agoStop giving platitudes - answer the question, or don't bother replying to me at all.