Trying to export a flipbook with TOPs and it gets really slow compared to just skipping to certain frames in my viewport and seeing the result.

If i use out of process as cook type i get 3 workitems cooking simultaneously but its slow. not 3 times slower but more
if i use in process i get one item but its the same speed as in my viewport.

if the in process workitem takes 6 seconds the out of process takes 60.

its difficult to share my scene but basically what is happening is :

1) import dense geo (2mil points) - group prims close to animated camera
1) scatter points on grouped prims each frame. so every frame is different and can be calculated separately, each frame doesn't depend on the previous.
2) Then i distort the points position using chops. point count is 200k.

if I check the info on my OUT_NULL in SOPs it says Memory :16mb(instanced), but when cooking my 64gb ram fills up very quickly.
20gb for in process when cooking just one frame - same speed as viewport
64gb out of process for 3 items - much slower than viewport

I dont understand why its using so much RAM.

i know CHOPs is single threaded, could this be the reason for pdg slow speed? or is this a RAM issue

I got around this by using the "Frames per Batch" so each houdini instance stays open until it finishes a certain amount of frames. to calculate the frames it will be processing i divide the total number of frames by the processors my scheduler is using. So if i set maxprocessors to 6 i have six instances of Houdini calculating 100 frames each for example to complete a 600 sequence. In that way houdini doesn't have to launch and close for every workitem so its very fast like i am exporting from a ROP opengl stills and the only limitation is RAM because all the points i am generating take a lot of space i cannot use all my threads.