In my experience (with Madmax at least) changing the bucket size does not affect the total bytes written per plot on temp drives.
1 Like
I think using smaller buckets was a trick to manage to do parallel plotting of 16 or more plots when we were all using the inbuilt plotter to plot with one core at a time. It makes sense to use more ram and fewer buckets if you have enough ram (this is actual ram in use by the system rather than allocated to a ramdrive).
Adding my experiences with 64 vs 128 buckets: Doing 16 4-thread Plots in parallel divided over 4 NVMe’s (4 per NVMe device) and Tmp2 on enterprise SSD’s on my 20 Core (40 Threads) Dual Xeon rig and it made no difference.
I let the system plot for 4 days in a row with each settings (4 days with 64 & Mem: 8000, then 4 days with 128 & Mem 4000).
With 64 buckets, The average total plot times were the same as with 128. So in my case, no need to bother and i’ll stick with 128 as there is no statistically relevant difference in the greater scheme of things. 64 buckets just eats more memory and offers me no benefits.
I don’t worry about wear on the devices as they are write-intensive enterprise grade devices I had lying around.
1 Like