It goes both ways really. It depends on if throughput or latency is more important to your workload. Game engines aren't going to benefit much from more parallel throughput at the cost of latency. They need a lot of serial tasks done in sequence on the CPU side of things. A video encoder on the other hand probably couldn't care less since its workload can be broken up nicely into independent blocks of data and will probably see an improvement overall even if each work unit takes longer than it otherwise would.




Particle's First Rule of Online Technical Discussion:
Rule 1A:
Rule 2:
Rule 2A:
Rule 3:
Reply With Quote
Bookmarks