I blame WebODM!

Yes, I mean it!

I bought a gaming rig, I thought it would be enuf for years to come… but NO!

What happened then? Well I found WebODM!

Suddenly I have to up my game, flex my :muscle:!
I maxed CPU and RAM,… but I still feel weak. :disappointed:

So now I feel that I can’t stop, I need more power…

It will not be cheap and it will take time…
As Frodo struggled to get the one ring to Mordor I will struggle to get 16c/32t and 128gb ram!

One CPU to rule them all, One CPU to find them, One CPU to bring them all and in the RAM bind them. :scream:

5 Likes

Beautiful, haha

Yeah, it really makes you appreciate the research-grade hardware that’s out there when you start doing photogrammetry and other intensive workflows. All of a sudden a decent size machine starts looking pretty tiny :stuck_out_tongue:

Look forward to seeing your build when you have it complete!

2 Likes

I hope that GPU acceleration works well soon, I need to get the processing time down.

2 Likes

Yes, for gaming, 16 GB RAM is enough, but not for photogrammetry. Except for small project. Sometimes I use another photogrammetry software, it also heavy. The lightest one is (not free) Reality Capture, then ODM, then Meshroom. Hopefully I can buy more powerfull computer someday :slight_smile:

GPU acceleration fortunatelly works in my notebook with Linux as main OS and using NodeODM.GPU Docker. Sorry, I never test it with Windows as main OS + Docker. It increasing processing time up to 25% in my last test

1 Like

Exactly what I’ve been discovering over the past year for myself. I was so excited when I built this machine with 16GB RAM, but it wasn’t enough when I was doing some GIS work. So it went to 32GB and I (naively) thought that’s more than I’ll ever need!

Now this machine is no different than my 4GB RAM laptop to me… Neither can do the job size/detail I want them to and now I’m looking at 128GB+ like Andreas, which is just nuts to me haha

How may cores do you have? I’m thinking that if I double the cores I double the ram. Got 8 cores now and like to get 16.

1 Like

4C/8T desktop (i7-6700k), 4C/4T laptop (Celeron N3450).

Desktop is likely going i3-12300 if I do a rebuild here shortly.

The Ryzen 3700X is pretty good with 64gig ram, 100 20mp images takes like 4h on ultra settings.

I’m planing to get the Ryzen 5950X with 128gig ram.
If the gpu acceleration works I probably get a GeForce 2060/12gig.

I’ve read an article where they tested different gpu’s with reality capture and 2060 seems to be the best budget offer.

2 Likes

What’s your job that you need so much process power?

Or it’s just for fun?

I wish WebODM network has max power to process on the web what I can’t process at home

1 Like

Have you had jobs not complete on Lightning?

I think Piero has done an excellent job of matching the VM sizes to the service tiers of Lightning. It returns jobs fast and handles up to 3,000 image tasks thus far.

1 Like

I’m considering an upgrade. Any thoughts on this rig?

1 Like

Does Lightning or Open Drone Map installed locally make use of GPU processing? I saw some debate on an old thread and wondered if it was updated yet to take advantage of GPUs

1 Like

Lightning does not. Local can, provided you’re fully up to date and using a supported host OS (Windows, Linux) with a supported NVIDIA card and up to date drivers.

Nice I was planning on a local install if I pull the trigger on the machine I posted above. Lightning has worked well for me.

1 Like

Yes sorry, I did’t want to mean that lightning is not powerfull enough…

Only I’d like to be able to make hihger resolution 3D meshes… More than 300.000 vertices…

That’s the only thing… for orthophotos and point clouds, perfect!

2 Likes

Understood! I didn’t think you were talking badly about Lightning at all, but concerned if you had jobs that didn’t finish or errored out on Lightning, which we obviously don’t want to have happen!

Agreed, it’d be amazing if we could make the VMs a bit bigger on Lightning to accomodate some higher quality settings, but I’m not sure if that is possible in the near term… That is going to be quite costly! We might have to grow a bit more before we can do that :slight_smile:

2 Likes

If you really need cheap resources have a look for netcup and contabo.

1 Like

I too have a gaming rig, currently trying to install ODM as a noob to this specific software.
i7 32gb RAM 8GB GPU (gtx1080)
ODM cant utilize my dedicated graphics card?
I am hoping to leave DroneDeploy so that I can process locally. 300 to 6000 images typically with GCP.

1 Like

Try resizing the images down slightly. CUDA acceleration is incredibly picky with image dimensions for each chipset.

1 Like

Docker only allows you to dedicate RAM to ODM, why can I not dedicate my GPU RAM to ODM?
I still have not been able to get ODM installed. Having linux on the asus rog already has been causing issues and I tried following the instructions for install on the odm page which dont tell you not to use the current version of docker desktop so this has been fun…