Hello! I recently built a new machine and have not been able to complete any model constructions. About 1/3 in, I will get a depth-map for filtering is corrupted. This same dataset builds no problem on my other machine, and I’ve noticed i can sometimes get Normal detail done, but never high
So far, I’ve tried:
A deep uninstall of RC, including registry entries,
Run memtest (passed)
Tried having my datasets on/off the machine’s local drives.
7950x
128 GB of memory
Win 11
Latest Nvidia drivers (first thing i tried)
Hi Fredugolon,
this is not a very usual error.
If you get a message “A depth-map is corrupted or missing” the cause could be that one of your images is corrupted. Please try to check your dataset. You can as well try to create a model one more time. This should fix the issue. Also, try to check topology, then clean.
You can also try cleaning your cache (application properties), changing a Cache location folder and starting again.
This helped to one customer before: “I changed the cache files to system that alone didnt fixed then I updated my drives installed RC into C: and underclocked my GPU. That fixed the calculation freezes.”
Thanks. I just tried setting cache to system, moved over to Nvidia Studio drivers, still no luck. Other GPU-bound apps like Redshift have worked fine on this machine so far, so I’m stumped.
Not sure how Check Topology will help when i can’t get past alignment to begin with
The datasets are also not the issue, I’ve tried multiple sets that all have worked on my main workstation
Can you try also some of the previous drivers if it will be the same?
Also, in that customer’s advice is mentioned underclocking of GPU, have you tried that?
I’ve tried 3 driver versions now, 2 versions of RC (1.3.1xx and 1.3.0xx), tried updating system drivers, and have underclocked the GPU as much as Afterburner allows me, to no avail.
I understand this is a very rare error, is there a diagnostic route i can take other than throwing everything I can at this machine in terms of updates? Someone also told me to try removing half my ram, which is not a great option imo XD
I also tried a build running Depthmap Algorithm V1, instead of V2, which still gave the same error, though after a much longer time
Additionally, I tried disabling installing iGPU drivers, as well as entirely disabling iGPU, same result
At this point I’m thinking of trying a fresh windows install on a completely new SSD to rule everything out that i can
Hi,
how many images are you using in this dataset? Are you processing them in one step?
Are you using SLI or NVidia bridge? If so, disable the bridge and run the cards in parallel.
Also, one customer with the same issue was able to solve this by: The problem eventually was a corrupt stick of memory that was put into the wrong slot during the switching of the memory. This is what caused the weird error’s I received in Reality Capture. After re-inserting the memory in the right slots (on the MB) everything worked as expected
One GPU, ~350 images (though many datasets of varying size fail), and i do alignment followed by buiding the mesh.
Thanks for the heads up on memory. I’ll run Memtest over night, then even if that clears, I’ll try removing ram to see if i can hunt down this gremlin.
@OndrejTrhan Thanks for your help, you can add my case to the “check your memory” list.
My motherboard bios poorly configured my RAM when XMP/expo was set up, timings were all off. Reset my bios and now the build completes, just slower than I’d like which means I will work on finding a memory configuration that is both fast and stable. So much fun being my own IT person
