It’s not that simple. Big mining operations like this are more likely to be spending their capital on long(er)-term equipment. Grin is strangely unique in that there’s a balanced C29/C31 cyclehash target ratio (90%/10% on launch, 45%/55% @ t=2 refresh cycles aka 1yr, 0%/100% @ t=4 refresh cycles aka 2yrs). Since a big mining operation is expecting their equipment to last for as long as possible, we can probably assume a 3 yr time horizon. In that situation, they’re more likely to buy Titan RTX or higher (henceforth referred to as Titan RTX+) cards in anticipation for C32+ (which @tromp mentioned needs at least 22GB).
However, Titan RTX+ cards don’t seem to be very price performant on C29, and they need to somehow internally balance or automatically switch between C29 vs. C31 cyclehash algorithms. If too much C31 power comes online, everybody else with a less powerfulGPU will reap higher C29 rewards due to the balance re-adjustment via the ar_scale (aka secondary scale) weighting.
I’m not aware of any other PoW token that launched with such nuanced and equitable considerations. Whether it works or not is another matter, but I’m glad that the plan was transparent ahead of time with clear guidance on which algorithm to mine (if you have the luxury of choice).
The calculations are laid out in @tromp’s two recent GPU posts: