I see, if it wasn't closed source though, I wouldn't be surprised if someone from the crypto community could manage a faster more gpu intensive way though. Let's be reasonable, if our gpu's aren't maxed it's unlikely it's fully optimized.
A couple of things are the cause of this...
1. Folding at home is not an algorithm. See a previous explanation I gave in this post: (
https://bt.irlbtc.com/view/603757.msg6708528#msg6708528). Therefore, it is highly dependent on the project and the core it is using what the GPU is doing. So no, people who program for crypto currencies will know fuckall about maximizing gpus for F@H. This is not technically feasible because
1a. the code must be hardware agnostic (read: NOT optimized) in order to run on the most combination of platforms, gpu generations, hw vendor configurations, driver versions.
1b. Any optimizations would only apply to specific projects running on cores, etc.
2. Some offtopic commentary about "fully optimizing algorithms". Of course, in a project such as F@H, it would be good to maximize GPUs. But for proof of "work" cryptocurrencies, it's wasteful and INEFFICIENCY should be valued. Consider scrypt, which runs the hottest on GPUs and keeps it at almost peak wattage. How does this give end users ANY benefit? Every gpu gets a share of its hash as a % of the total hash of everyone else using GPUs (not going to bother talking about asics). Therefore, the less efficient the algorithm, the better, as everyone gets the same % while using less electricity.
