Thanks to you, we’ve steadily added more computing power to our CPU farm over the past 4 months. We’re happy to report the farm has at least 70% more CPU processing power now than it did 4 months ago. The additional processors were purchased in direct response to the overall farm load we saw on an averaged basis. We continue to monitor farm loads, in order to make the best hardware purchasing decisions in the future.
If you’ve used Pixel Plow recently, you may have noticed the CPU farm rarely gets into an over-subscribed state now. When it does, it doesn’t stay there for long, and compute throughput seems fairly unaffected. Now that Harpoon is keeping those nodes busy whenever there’s work to do, the farm can very quickly recover from over-subscription, as well as maintain great performance when heavily loaded. Consistent and predictable compute throughput remains our goal.
So keep sending in those CPU render jobs!