Hey all, fairly new Chia farmer here so bear with me.
I was running K32 plots before and hardly ever ran into any issues.
Now I’ve been trying to do K33 plots but they keep freezing/stalling right around the exact same spot every time - right around 66%. I can sit there watching the disk activity on the Temp drive and it’s going constantly, until it just stops and goes to 0% activity.
Last thing in the log:
Progress update: 0.66
[P3-1] Table 3 took 752.965 sec, wrote 6879846511 right entries
another plot running at the same time made it to 79% and stopped:
Progress update: 0.79
[P3-1] Table 5 took 444.968 sec, wrote 7065467419 right entries
System is i7-7700K OCed to 5GHz, 64GB RAM @ 3600, Temp drive = Samsung 980 Pro M.2 SSD 2TB
Temp drive is not anywhere close to filling up.
Same issue happened even when I swapped to another SSD drive as the temp just to test
For reference: 6 threads, 1x multi, 512 / 128 buckets - using the GUI
Anyone have any ideas of other logs I can check to see what’s going wrong?
I have been able to complete a few K33s before so not sure what changed.
Not running out of memory as far as I can tell. Neither the temp drive nor my RAM is getting near maxed out. Still have about 1TB of space left on the temp drive and RAM from what I’ve observed is only using about half of the available.
Only thing I can thing is maybe because I’m putting the temp drive of 2 plotting instances on the same drive, they may be conflicting or something? But both in separate folders
I’m going to try just running one solo and see.
If that fails I will try with CLI
The legacy Chia plotter uses a buffer of size 3389 MB.
The help for madmax’s executable does not mention the buffer size.
With the above in mind, when I attempted to create a k33 plot (which is roughly twice the size of a k32 plot), via madmax, I increased the buffer to double 3389 and never had a problem.
Perhaps that might help?
EDIT:
I just checked the GUI, and chose madmax.
Scrolling through the options, I did not see an option for changing the buffer.
I wonder, when creating a madmax k33 plot via the GUI, if it sets the buffer accordingly? @Azuka, when a plot starts, it outputs the buffer size. Via the GUI, you would have to check the log for the plotting job.
I use only the command line, and I include “-b 6778”.
When my k33 plotting job begins, it outputs that figure, so I know it is working with that value.
I had a problem before that my plots (k32) would randomly crash in phase 3. It would run fine for a while and then suddenly stop.
In my case it turned out to be a memory issue, maybe want to run memtest and see if there is any issue with the memory
By the way, MM has released an updated version that may be 10-15% faster on your box (7700K). You should use CLI version, as you don’t want to have any extra overhead coming from non-MM code. I am not sure, but don’t think that version will ever make to the official chia release (so you need to do an extra download for it).
Also, when you are plotting for k33, potentially the power consumption growth is non-linear, i.e., one k33 plot will burn more W than 2x k32.
With this in mind, soon you will have an option to use a light compression and run MM on your GPU (5x-10x less W burned per plot), so you may want to start replotting at that time. So, whatever you do right now, keep in that mind.
UPDATE
Sorry, I put there 10-15x faster, where it should be 10-15% faster (it was like that on my i9-10900) as that is the CPU based plotter. The older box got below 10% speedup bump. still worthwhile to download. The GPU plotter will be 5-10x faster, though.
I think you guys may have nailed it with the bad memory thing.
I haven’t run a full pre-boot memtest… but I did just run through a windows based test and it ran into errors within the first couple minutes. So I’m gonna guess one of my (brand new) RAM sticks is bad and try replacing.
Yes I am VERY much looking forward to compression plotting and sounds like my RTX 3060 12GB will work great for chewing out those plots. But I just wanted to get some more plots going in the meantime because who knows how long.
Ya same here. I just don’t normally test my new RAM (though I guess I should! lol)
But thanks for the help all… I wouldn’t have suspected a RAM issue at first… as I’m not seeing other system issues or BSODs or anything, but it seems the RAM works fine… up until it hits a certain block
Nah the memory isn’t overclocked, just the CPU. But I put the CPU back to stock before testing.
Fairly confident it’s the RAM now. It’s consistent now with every plot even K32s, just stalls out in phase 3
Have replacement RAM coming today so will know soon