[citation][nom]warmon6[/nom]It truthfully a mix relationship on the arguments between cpu's and gpu's. While i cant vouch for WCG, for F@H while they use both, cpu's (to me) appear better overall. The main issue with folding on a gpu is that your limited with what you can run on it. Memory size on a gpu is smaller than what on the system (F@H has there client setup to mainly run there WU's on the faster gpu memory), you have to run a client per gpu instead of one client for all gpu's, and the way some proteins needed to be folded.... is just not possible to run a simulation on gpu. Forcing a CPU to take the job.CPU's, there slower than the gpu but unlike the gpu clients, the cpu clients can take all the cores that are thrown at them. You'll be amazed there are folding nuts out there that will run multi-cpu servers just for folding. Common big servers for folding are AMD quad-g34 socket, 12 core cpus servers. Having this many cores at your disposal can be just as fast as a mid to high range gpu (on small multi-core wu's) while requiring far lower power. (although most guys running an enhanced version of the multi-core F@H client and thats call bigadv. Which are huge wu's (100+ MegaByte) that needs powerful computers and needs these WU's done very quickly.)As pointed out above, cpu's use less power than gpu's. Which probably is why this fellow able to run multiple motherboard off 1 PSU.[/citation]
Nice explanation 😀 Thanks!!
Facebook
Twitter
Instagram