My problem still actual even in new WM builds, here old post: WM do not allow page data to disk
I have 1 TB HDD, with 200-300 GB free space - and WM never allow me to page data while exports (only while preview - yes)
in 4043 i just started see new text in errors:
Here’s the settings to double check in program settings:
Note that with all of that: When a device is activated, it needs to have all of its data in memory at once. If you have a proclivity to make large checkpoint or layer devices for example, that would require enough ram to bring in 20-30 packets at once to process, which will cause fatal out of memory errors.
You can try to set your windows swapfile size much larger which should that memory allocation to proceed but will cause windows to swap which is terribly slow. A better solution is to re-scope your checkpoints, layers, etc into reasonable logical units to reduce their memory demand.
If you have a proclivity to make large checkpoint or layer devices for example, that would require enough ram to bring in 20-30 packets at once to process, which will cause fatal out of memory errors.
you found my WM weak spot man…
after those words i need to show you my home-made monster: RAM Destroyer 3000 (all purpose of that blueprint is just to be “comfortable” x128 inputs Layer node to combine a lot of shape nodes. I think with Code node functionality i may replace it in future):
created 2 year ago or so
Also thank you for that explanation
It really did help me - now it looks like issue fixed - maybe really all problem was in huge layers nodes… Really thank youuuu 
EHEHEH
I have resorted to using “staggered” Checkpoints to avoid this, though I considered suggesting that they should create “packets” by sibling outputs (of various devices fed into them) automatically. I have wished Checkpoints did internal rerouting, too (I end up using macros for this, so not a crucial request).
Can you explain more what you mean by checkpoints doing internal rerouting?
In a routable checkpoint’s properties, you could add an assignable column for the output order after the name column. Instead of two checkpoints wired together, you have one device that can output in a clean and clear order to following devices. Naturally, the default behavior for input and output is the same order.
Edit: I just thought of adding output grouping in the checkpoint to break up processing packets virtually. That might be a problem given what you said about device activation, unless there is a way to do “virtual” devices inside checkpoint and layer devices (at least for memory allocation and process handling).
