Another question, if you're plotting with the GPU plotter with 'direct write' so it optimizes and plots at the same time, how much slower does it go then normal plotting? It also looks like it freezes when you start plotting, but it's still writing to the disk, no status text is updated. Does stagger size matter with 'direct write'? If it's properly defraged couldn't you effectively have a stagger size of like 1024 and it wouldn't matter?
Bump on this... It takes so long to plot, a few questions help a lot.
This thread should answer your questions.
Specific details in the middle of that thread
https://burstforum.com/index.php?threads/gpu-plot-generator.45/page-12#post-3252Hopefully these conversation can move to the forum and off of this thread for better organization.
Good-Luck
I already checked through that thread and read the relevant posts. It doesn't answer the above questions though as I was asking for other peoples experience with direct write and stagger size with direct write. That thread is also not active, so it's sorta pointless to post there.
Also, I finished optimizing one of my plots using the optimizer. However the newly 'optimized' plot is now smaller then the old plot by 5GB and of course the sizes don't match when I start up blagos miner. Is this normal? Did something go wrong along the way?
Hmmm... after adjusting the name of the file it only finishes 3s faster then the old plots (also it appears to be corrupt now).
-
https://burstforum.com/index.php?threads/gpu-plot-generator.45/page-11#post-3014Why Direct is Slower than Buffer -
https://burstforum.com/index.php?threads/gpu-plot-generator.45/page-11#post-3034Plot size will remain the same after optimization, if it is not then something went wrong along the way. Did you run out of disc space?
Information about Stagger setting when Direct Plotting -
https://burstforum.com/index.php?threads/gpu-plot-generator.45/page-13#post-3348During a direct right method, the stagger setting is used to setup a ram buffer for pre-ordering of the nonces.
In your example, a stagger size of 1024 during a direct right method will just use 1024KB of RAM for the nonce buffer re-ordering.
In this thread and the thread on the forums a few people have complained about the reliability of the direct right method, also the performance hit for direct right compared with just plotting and optimizing does not appear to be worth it.
Maybe someone else can add some information to this topic.
Just because no one has posted in the BurstForum thread about the GPU plotter doesn't mean it's dead. This seems like the perfect type of question to ask in a Burst forum within the GPU plotter post. I'm guessing the regulars have had their questions answered you would be a perfect candidate to awaken the thread and maybe Cryo will answer your questions directly
