My OS SATA SSD 240 Gb is screaming at 100 % all the time even thou I’m plotting in a 2 Tb NVMe and my final directories are one of the three HDDs I have. Why is chia using 100 % of my OS SSD?
Did you oversubscribe on memory and is it pagefile thrashing?
3 Likes
Same problem read this post
if is this an other problem please send us your NVMe info ex. id ,model
No no.
My Plotting m.2 ssd seems really normal and the speeds are fine with althought it rougly uses 30-40 % when plotting 5 in parallel. My problem is that for some unknow reason the SATA SSD that i only have my OS in (and some random software) is always at 100 % when plotting.
Motherboard: Asus Z97-Pro
CPU: i5-4690k (turboed to 3,89 Ghz)
RAM: Kinsgton HyperX Fury 8 Gb 1866Mhz CL11 DDR3(downclocked to 1600Mhz while I’m waiting for new RAM to arrive and overclock my CPU)
Plotting SSD: PNY 2TB XLR8 CS3030 M.2 NVMe Gen3
OS SSD: Kingston 240 GB SATA (I don’t have the specs right know)
5 parallel plots on 8Gb of memory? That would definitely be oversubscribing the amount of memory available, hence using the pagefile on the operating system drive to compensate.
Each plot by default is 3400Mb of ram x 5 plots is wanting 17Gb of ram minimum.
4 Likes
Also, your CPU is 4C/4T. That is a 1 to 2 parallel plot maximum right there. Anything beyond and you will be starving processes for threads.
Well yea anything above 3 is really slow but 3 seems to be the sweet spot as the time to plot one is the same as three (within reasonable deviation)
What Ive seen, if single plotting I can make 2,5-3,5 plots a day. Three parallel completes in circa ~9 h. Are my settings fucked up or something?
Is it a bad thing if I’m starving my processors? I have my computer plotting when Im sleeping/working and just three plots will always complete before I can add the next one. Or is there a way that I can queu parallel plots?
Im getting Crucial Ballistix 4x4 GB 1600 Mhz CL8 and hope to overclock my cpu to something like 4.4 Ghz.
How should I make my queu if I can check my computer before going to bed, waking up and coming from work? (~8 h intervals)
Yes. If your processes are fighting over threads, your OS will try to service them all and it will end up running each a very little bit, with a very large overhead of constantly switching between them.
Phase1 is b far the most CPU intensive. Given your system, you should strive to never have more than 1 process in phase 1 at the same time
Use a plot manager like Swar to keep it feeding plot jobs(and any restrictions for jobs in phase 1 etc).
Where can I find this manager?
It still baffles me that I can either queu plots to have one plot complete in ~8 h or what I tried today 2 in paralel one in queu at it was also ~8 h.