I'm in the process of building a plotting machine for Chia coin mining. It requires a ton of disk IO to the temporary drive. I have two U2 drives on PCI-E adapter cards that I plan to software RAID0/stripe in order to have one big and fast temporary drive. The OS will be the latest version of Win10.
The MB in question is an AsrockRAck EP2C622D16-2T (Socket 3647). Link in case you're interested: ASRock Rack > EP2C622D16-2T
As you can see in the attached block diagram, the PCI-E lanes are split (unequally, but still split) between CPUs. Would it be better to have both drives/halves of the stripe on the same CPU to minimize latency, or better to split them b/t CPUs in order to distribute the load b/t CPUs?
Keep in mind the CPUs will be at 90-100% usage for the computationally-intensive part of the mining process. Thanks in advance.
The MB in question is an AsrockRAck EP2C622D16-2T (Socket 3647). Link in case you're interested: ASRock Rack > EP2C622D16-2T
As you can see in the attached block diagram, the PCI-E lanes are split (unequally, but still split) between CPUs. Would it be better to have both drives/halves of the stripe on the same CPU to minimize latency, or better to split them b/t CPUs in order to distribute the load b/t CPUs?
Keep in mind the CPUs will be at 90-100% usage for the computationally-intensive part of the mining process. Thanks in advance.
Attachments
-
188.3 KB Views: 8