I wonder what kind of PPD would be achieved from running the SMP client on 3 cores and the GPU client on the remaining core. The only thing about the Windows SMP is that it is extremely prone to getting the 30xx WU, which finish super close to the deadlines when running on only 2 cores (from my experiences at least).
Some of the others (such as Xilikon, who you've been talking to) would know more about this than I would. I just personally don't like running so close to deadlines so I don't know if this would be a good solution for it.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}