Thank you for the correction Anton! You are right - it certainly does. I assumed it would keep the separate process going, but it looks like it kills the render job once you exit Houdini.
@@cgforge okay then I didn’t do anything wrong, good to hear 😄 do you know if you can start TOP nets in the same way? Often I have to cache many sims at once do that would be great
How do you do lookdev on a simulation if you set the cache memory to "0", instead of "5000"? You do want to hit play and see the simulation to make changes right? Confused?
When set to 0, I believe it tells Houdini to only display the current frame and to not display previous frames. So, if you simulated something up to frame 45, for example, it would only take frame 45 and store that to memory.
Great content! It worked for me to render in Mantra, but I haven't been able to make it work in Karma. Do you know how to do it in the Karma-Solaris pipeline?
Hey @grn7, one thing to consider is that Karma only takes in USD information. So, if there isn't a network setup in LOPs to convert your scene into USD, then it might be related to that. I would set up your LOP network first, and try using the cmd line to send of a render at the end of that network rather than going into ROPs.
The substeps shouldn't matter. If you have checkpoints, the simulation will just start from there and substep as normal. Is that what you're thinking about?
@@cgforge sort of but I am facing something strange. For example I am rendering with 10 substeps, and I stopped on frame "100". The last checkpoint I am getting on my checkpoints folder is "1000". In checkpoints I am not getting frame number. I am getting frame number acording to substeps. For 2nd frame I am getting 20. So then how can I resume my caching from 2nd frame if the last saved frame in my checkpoints folder is 20?
Hi. First of all, thank you for all this good information. I tried to cache a flip simulation with The save to disk in background option, but my RAM goes to 98%. Si there something that Indon't understand about that? It supose that The CPU goes with all and RAM get free.... PD. Sorry for my English....😅
Hi Alberto, Even if you do all the right things that I talk about in this video, it is still possible to reach 98% of your RAM. Fluid simulations can take a lot of ram, so you might need to turn down the quality settings on your simulation or purchase more ram for your computer.
Hey Invader Zim, go to your houdini install directory... which is probably your C drive --> Program Files --> Side Effects Software --> Houdini version --> bin and then find the hcmd.exe to bring up the command line tool
I've never tried it with Linux, but you might be able to look some of that up in the documentation. Or the sidefx forums might have some Linux users that can help you out
The "caching" process itself isn't something that you would want to use a GPU for, so from my understanding, no, that's not really possible. The caching part is only saving a file to your hard drive, and it doesn't usually take that much time to do so. Instead, it's better to use the GPU for the simulation part. Is there a specific situation that you have in mind?
@@cgforge Yes, absolutely correct about writing to the computer. Do you know of any way to use the GPU for the sim since it will only remember a few frames at a time?
It depends on the solver that you're trying to use. Not all solvers are designed with the GPU in mind, but if you take a look at the documentation, it will often have a section that talks about it. As an example, take a look at the pyro solver documentation here: www.sidefx.com/docs/houdini/nodes/dop/pyrosolver.html press ctrl + f to do a keyword search and type in "gpu" It will bring you to the advanced section, and there you can learn all about OpenCL and how to use it.
First time doing a big sim and the Checkpoint advice really saved me. Thank you
Glad I could help! Cheers
Thank You very much, Tyler!
Awesome, thank you. When you said: you can also just save to disk in background and close Houdini, it does not continue caching after closing...
Thank you for the correction Anton! You are right - it certainly does. I assumed it would keep the separate process going, but it looks like it kills the render job once you exit Houdini.
@@cgforge okay then I didn’t do anything wrong, good to hear 😄 do you know if you can start TOP nets in the same way? Often I have to cache many sims at once do that would be great
Yes, but instead of using "render" as the command, you need to say, "topcook" and then browse for the top node that you're trying to cook.
looks like it does keep caching after closing now. Maybe they fixed it
Very informative, thanks!
new cap congratulations
😊😊
by the way good tutorial as always 👍👍
super tutorial. help me lot thank you bro
Glad I could help Alagar! Cheers
How do you do lookdev on a simulation if you set the cache memory to "0", instead of "5000"? You do want to hit play and see the simulation to make changes right? Confused?
When set to 0, I believe it tells Houdini to only display the current frame and to not display previous frames. So, if you simulated something up to frame 45, for example, it would only take frame 45 and store that to memory.
Great content! It worked for me to render in Mantra, but I haven't been able to make it work in Karma. Do you know how to do it in the Karma-Solaris pipeline?
Hey @grn7, one thing to consider is that Karma only takes in USD information. So, if there isn't a network setup in LOPs to convert your scene into USD, then it might be related to that. I would set up your LOP network first, and try using the cmd line to send of a render at the end of that network rather than going into ROPs.
But can you tell how can i resume my caching if i have substeps
The substeps shouldn't matter. If you have checkpoints, the simulation will just start from there and substep as normal. Is that what you're thinking about?
@@cgforge sort of but I am facing something strange. For example I am rendering with 10 substeps, and I stopped on frame "100". The last checkpoint I am getting on my checkpoints folder is "1000". In checkpoints I am not getting frame number. I am getting frame number acording to substeps. For 2nd frame I am getting 20. So then how can I resume my caching from 2nd frame if the last saved frame in my checkpoints folder is 20?
Thanks boss
Hi.
First of all, thank you for all this good information.
I tried to cache a flip simulation with The save to disk in background option, but my RAM goes to 98%.
Si there something that Indon't understand about that? It supose that The CPU goes with all and RAM get free....
PD. Sorry for my English....😅
Hi Alberto,
Even if you do all the right things that I talk about in this video, it is still possible to reach 98% of your RAM. Fluid simulations can take a lot of ram, so you might need to turn down the quality settings on your simulation or purchase more ram for your computer.
Where do I find the Command Line Tools? I don't know if I have it installed
Hey Invader Zim, go to your houdini install directory... which is probably your C drive --> Program Files --> Side Effects Software --> Houdini version --> bin and then find the hcmd.exe to bring up the command line tool
@@cgforge thanks! That worked
Any advice for linux?
I've never tried it with Linux, but you might be able to look some of that up in the documentation. Or the sidefx forums might have some Linux users that can help you out
Can you recommend some versions of Linux?
Is it possible to cache sims with GPU?
The "caching" process itself isn't something that you would want to use a GPU for, so from my understanding, no, that's not really possible. The caching part is only saving a file to your hard drive, and it doesn't usually take that much time to do so. Instead, it's better to use the GPU for the simulation part. Is there a specific situation that you have in mind?
@@cgforge Yes, absolutely correct about writing to the computer. Do you know of any way to use the GPU for the sim since it will only remember a few frames at a time?
It depends on the solver that you're trying to use. Not all solvers are designed with the GPU in mind, but if you take a look at the documentation, it will often have a section that talks about it.
As an example, take a look at the pyro solver documentation here: www.sidefx.com/docs/houdini/nodes/dop/pyrosolver.html
press ctrl + f to do a keyword search
and type in "gpu" It will bring you to the advanced section, and there you can learn all about OpenCL and how to use it.