RC is crashing my PC, but not my laptop. Both are new and high spec (see below). PC crashes completely and restarts on medium alignment settings, 16bit HDR images (tiff). Drivers all up to date.
Hi bajp2019,
what kind of crash do you have? Is there some error message?
What are your alignment settings and how many images are you trying to align?
33 images, tiff ~266mb each, 16bit converted to 8bit on import to RC.
PC Settings:
max features per mpx: 50,000
max features per image: 10,000
image overlap: low
image downscale factor: 1
max feature reprojection error: 2.00000
Laptop settings (no crash)
max features per mpx: 10,000
max features per image: 40,000
image overlap: low
image downscale factor: 2
max feature reprojection error: 2.00000
Hi bajp2019,
this second crash was also without any error? Just crash?
Is this also happening for other dataset? Is it possible to compute Preview model?
Probably there could be a problem with number of cores. For example you have 24 cores * 2 hyperthreading * 4 gb of ram per thread = 192GB RAM for a good performance in high compute calculation. Otherwise it is swapped on C disc.
Could you test disabling some of your cores or disabling hyperthreading in your BIOS? And turn the reconstruction test again?
One more thing I would try is to manually set windows swap to 150% of your current RAM so you would have 32GB+ 150% on windows swap (keep in mind to use an SSD disk for that when you will all the time have at least 50 GB of storage left)
I’m having a similar issue. Video drivers are current (RTX4070, 64gb of RAM) but it seems to overuse the gpu and run it hot until the BSOD. I can typically get through the alignment phase but reconstructing in normal detail results in the failure.
Running 1.2.1 on an old machine with less spec hardware and no issues whatsoever.
Hi, I’ve tried all the suggestions and the problem persists. I’ve also added 2 nvme sticks for dedicated project and cache storage. I’ve updated my bios as well, and adjusted various RAM speed and core settings. Factory reset also did not help. I can render a preview mesh, but otherwise the system totally crashes after alignment.
I’ve run prime95 and my system is stable under these stress tests.
Do you have any recommendations for RAM speed or core settings, or other bios settings?
Hi, it is still really strange.
Which version of RealityCapture are you using?
So, is the alignment successful? The meshing crash is there after a while or is it a direct crash? What is written in the console view before the crash? What kind of storage are you using? Is it external or internal storage for the data? For meshing are you still using the default settings?
Can you try to disable on GPU and try to compute the model using the only one?
One possible solution of one customer: I had the temp folder path on an SSD which has a second installation of windows inside and there was a write permission issue. When I changed the temp path to another disk it started working perfectly!
HW usage during reconstruction (can you check in which step is it crashing for you?):
Depth map calculation: (Storage, RAM, GPU - CUDA, Vram only)
Decompose for Large scene reconstruction, Depth map Grouping: (CPU, RAM, Storage only)
Model creation / Meshing: (CPU, RAM, Storage only)
Meshing: (CPU, RAM, Storage only)
Also, some customers have similar problem and it was connected with a voltage of the RAM. This could also happen when overclocking as the voltages could be higher than they should be.
Storage: Internal NVME (1 for OS, 1 for Project, 1 for Cache)
For meshing are you still using the default settings? Yes
The crash is on depth map calculation: console shows 0% progress for 1-2 seconds and then total PC crash. So it may be a RAM issue? I have tried reducing the speeds but not the voltages…
So I should rule out these one by one: Storage, RAM, GPU - CUDA, Vram only?
I don’t think it is storage, I just added two new nvme sticks and it didn’t help.
RAM: I have 4 x 16GB sticks, should I remove two sticks?
Two GPUs do sip a lot of power, and depth map calculation does use all GPUs unless you disable some, so it sounds plausible that disabling one makes it work when the power supply is at its limit.
