In general, trying to split up frame rendering between multiple GPUs runs into issues with latency because either you need to share data between the GPUs (so some reads/writes go to the remote GPU, which takes longer and then there’s cache coherency issues on top of that) or you need to mirror it (so all data transactions between GPU and CPU need to be doubled, bus bandwidth might need to be shared), and there could still be cross-gpu dependencies that require eating those longer latencies.

100 fps means each frame needs to be rendered in 10ms, so even round trips of 1ms can be costly and risk missing the frame deadline (which results in a stutter). The whole idea of multi-gpu is to get even higher frame rates, so the higher they get pushed, the bigger an impact that latency has.

Though maybe raytracing workloads would work well with it. Something that depends more on massive parallelization than anything else and where each ray is independent. But load balancing still isn’t trivial as it will depend on the scene or even the camera and lighting position and the software complexity might not be justified by how small the market would be.

It’s not just money but also expertise required to build, run, and maintain the fabs, which are each different specialized skills (or more accurately an umbrella of different specialized skills). While these people certainly exist, I don’t think there’s a large pool of them waiting to get hired. And with memory production being so concentrated in Taiwan and South Korea, it will be difficult to poach workers because (as I understand) both cultures are big on loyalty to the nation.

That’s not to say I disagree that anyone else should try to get into the memory fab game (or chip fab in general) as I think the market needs more players, but it’s easier said than done, even if you have a ton of money to throw at the problem. Look at Intel, they struggled on the fab side and already had fabs going as a leading chip producer. Or AMD, who sold their fabs because they were struggling to keep them profitable.

Though I also think that governments would be smart to subsidize (or outright socialize) chip fabs, as depending on others to make these is giving up on a lot of economic power (which can affect military power, so spreading chip production out more could increase world stability).

“You just want an open relationship so you have an excuse to cheat!”

• person who doesn’t think about things or understand them

It will always be an arms race. If someone develops a new style that doesn’t look like AI, then it’s only a matter of time before an AI model can generate that style.

And the future I see for this arms race is eventually it will be better than humans at producing art that doesn’t look like AI.

I’m at the point where I think AI generated shit should be judged on its own merit rather than for whether or not it’s AI because eventually we won’t be able to tell from other factors and any attempts will end up hurting humans more than the AIs (because they have no feelings to be hurt and will just continue while some portion of the humans will get discouraged and stop bothering to try).

Plus, at least with programming, saying code sucks because it’s AI generated is weak. If a code submission is shit, there will be technical reasons that it is shit that have nothing to do with who (or what) wrote it.