Agisoft 2.0: Reason: VK Error : VkResult is "ERROR_TOO_MANY_OBJECTS" at line 12

[JPG_Z]

Hi,

So i'm getting the error following error doing texture blending in tiled model:

Cannot use GPU. Reason: VK Error : VkResult is "ERROR_TOO_MANY_OBJECTS" at line 12
Performing blending on CPU...

Currently using a 7900xtx with the latest drivers.


Any ideas why?


Best regards,
JM

[Alexey Pasumansky]

Hello JM,

Can you please share the log from the beginning of Build Tiled Model task (or at least its texturing phase) and also the starting lines of Metashape from the Console pane?

[JPG_Z]

Hello JM,

Can you please share the log from the beginning of Build Tiled Model task (or at least its texturing phase) and also the starting lines of Metashape from the Console pane?

Hi ALex,

Here it is the log

2023-01-10 18:31:26 Agisoft Metashape Professional Version: 2.0.0 build 15597 (64 bit)
2023-01-10 18:31:26 Platform: Windows
2023-01-10 18:31:26 CPU: AMD Ryzen 9 5900X 12-Core Processor
2023-01-10 18:31:26 CPU family: 25 model: 33 signature: A20F10h
2023-01-10 18:31:26 RAM: 63.9 GB
2023-01-10 18:31:26 OpenGL Vendor: ATI Technologies Inc.
2023-01-10 18:31:26 OpenGL Renderer: AMD Radeon RX 7900 XTX
2023-01-10 18:31:26 OpenGL Version: 4.6.0 Compatibility Profile Context 22.12.2.221216
2023-01-10 18:31:26 Maximum Texture Size: 16384
2023-01-10 18:31:26 Quad Buffered Stereo: not enabled
2023-01-10 18:31:26 ARB_vertex_buffer_object: supported
2023-01-10 18:31:26 ARB_texture_non_power_of_two: supported
2023-01-10 18:31:26 Using simple console. Rich console can be enabled in Preferences dialog.

2023-01-10 18:32:00 BuildTiledModel: source data = Depth maps, pixel size = 0.00375611, tile size = 256, face count = High, ghosting filter = 1, quality = Low, depth filtering = Moderate, PM version
2023-01-10 18:32:00 Generating depth maps...
2023-01-10 18:32:00 Preparing 1666 cameras info...
2023-01-10 18:32:00 cameras data loaded in 0.44 s
2023-01-10 18:32:01 cameras graph built in 0.726 s
2023-01-10 18:32:01 filtering neighbors with too low common points, threshold=50...
2023-01-10 18:32:01 avg neighbors before -> after filtering: 166.956 -> 21.4004 (87% filtered out)
2023-01-10 18:32:01 limiting neighbors to 16 best...
2023-01-10 18:32:02 avg neighbors before -> after filtering: 21.4004 -> 8.46098 (39% filtered out)
2023-01-10 18:32:02 neighbors number min/1%/10%/median/90%/99%/max: 1, 3, 7, median=16, 16, 16, 16
2023-01-10 18:32:02 cameras info prepared in 2.754 s
2023-01-10 18:32:03 saved cameras info in 0.023
2023-01-10 18:32:03 Partitioning 1666 cameras...
2023-01-10 18:32:03 number of mini clusters: 34
2023-01-10 18:32:03 34 groups: avg_ref=49 avg_neighb=43.5 total_io=189%
2023-01-10 18:32:03 max_ref=50 max_neighb=108 max_total=158
2023-01-10 18:32:03 cameras partitioned in 0.023 s
2023-01-10 18:32:03 saved depth map partition in 0.002 sec
2023-01-10 18:32:03 loaded cameras info in 0.055
2023-01-10 18:32:03 loaded depth map partition in 0 sec
2023-01-10 18:32:03 already partitioned (50<=50 ref cameras, 23<=200 neighb cameras)
2023-01-10 18:32:03 group 1/1: preparing 73 cameras images...
2023-01-10 18:32:03 tie points loaded in 0.007 s
2023-01-10 18:32:04 AMD Radeon RX 7900 XTX (gfx1100): no SPIR support
2023-01-10 18:32:04 Found 1 GPUs in 0.034 sec (CUDA: 0.024 sec, OpenCL: 0.01 sec)
2023-01-10 18:32:04 Using device: AMD Radeon RX 7900 XTX (gfx1100), 48 compute units, free memory: 24557/24560 MB, OpenCL 2.0
2023-01-10 18:32:04   driver version: 3504.0 (PAL,LC), platform version: OpenCL 2.1 AMD-APP (3504.0)
2023-01-10 18:32:04   max work group size 256
2023-01-10 18:32:04   max work item sizes [1024, 1024, 1024]
2023-01-10 18:32:04   max mem alloc size 20876 MB
2023-01-10 18:32:04   wavefront width 32
2023-01-10 18:32:04 Using device: CPU AMD Ryzen 9 5900X 12-Core Processor (using 19/24)
2023-01-10 18:32:04 Loading kernels for AMD Radeon RX 7900 XTX (gfx1100)...
2023-01-10 18:32:04 Kernel loaded in 0.002 seconds
2023-01-10 18:32:13 group 1/1: cameras images prepared in 10.566 s
2023-01-10 18:32:13 group 1/1: 73 x frame
2023-01-10 18:32:13 group 1/1: 73 x uint8
2023-01-10 18:32:13 group 1/1: expected peak VRAM usage: 123 MB (64 MB max alloc, 1368x1368 mipmap texture, 16 max neighbors)
2023-01-10 18:32:13 AMD Radeon RX 7900 XTX (gfx1100): no SPIR support
2023-01-10 18:32:13 Found 1 GPUs in 0 sec (CUDA: 0 sec, OpenCL: 0 sec)
2023-01-10 18:32:13 Using device: AMD Radeon RX 7900 XTX (gfx1100), 48 compute units, free memory: 24557/24560 MB, OpenCL 2.0
2023-01-10 18:32:13   driver version: 3504.0 (PAL,LC), platform version: OpenCL 2.1 AMD-APP (3504.0)
2023-01-10 18:32:13   max work group size 256
2023-01-10 18:32:13   max work item sizes [1024, 1024, 1024]
2023-01-10 18:32:13   max mem alloc size 20876 MB
2023-01-10 18:32:13   wavefront width 32

2023-01-10 20:35:16 Avg camera fetch time: 0.431827 sec
2023-01-10 20:35:16 Total stalled time: 47.913 sec
2023-01-10 20:35:16 Avg camera processing time: 0.0130877 sec
2023-01-10 20:35:18 Cannot use GPU. Reason: VK Error : VkResult is "ERROR_TOO_MANY_OBJECTS" at line 12
2023-01-10 20:35:18 Performing blending on CPU...
2023-01-10 20:35:18 rendering 5149 pages
2023-01-10 20:35:22 calculating mesh connectivity... done in 0.029 sec
2023-01-10 20:35:22 detecting outliers...

is it enough?

Its always the same problem in lowest, low, medium or high.

Thank you.


Best regards,
JM

[JPG_Z]

Gonna use this thread to another problem:

[GPU 1] Using device: AMD Radeon RX 7900 XTX (gfx1100), 48 compute units, free memory: 24557/24560 MB, OpenCL 2.0

The 48 Compute Units are 'wrong' (?), It says 'Compute Units 96' https://www.techpowerup.com/gpu-specs/radeon-rx-7900-xtx.c3941


Could it be a bug?

edit. its because they are 2 chiplets with 48 in each one. But agisoft only saw 1 chip (48 cu) rather then the 2 chips (96 cu).

Any idea how to fix this, if possible?

[Alexey Pasumansky]

Hello JM,

Thank you for additional information regarding the texturing issue, we will try to reproduce it on our side on a similar configuration and analyze the problem.

As for your question about GPU initialization line output, for OpenCL devices Metashape gets the "compute units" number from the driver reply upon CL_​DEVICE_​MAX_​COMPUTE_​UNITS call. It seems that for modern ADM graphic cards with RDNA architecture the command returns number of WGP (so-called "workgroup processors") which are actually dual compute units.
Additional information on this matter can be found here:
https://www.amd.com/system/files/documents/rdna-whitepaper.pdf
https://en.wikipedia.org/wiki/RDNA_(microarchitecture)#Differences_between_GCN_and_RDNA

However, Metashape is not relying on this output, the task of parallel computations on all available compute units is performed by GPU driver. You can check GPU load using dedicated utilities, such as GPU-Z, for example.

[JPG_Z]

Hello JM,

Thank you for additional information regarding the texturing issue, we will try to reproduce it on our side on a similar configuration and analyze the problem.

As for your question about GPU initialization line output, for OpenCL devices Metashape gets the "compute units" number from the driver reply upon CL_​DEVICE_​MAX_​COMPUTE_​UNITS call. It seems that for modern ADM graphic cards with RDNA architecture the command returns number of WGP (so-called "workgroup processors") which are actually dual compute units.
Additional information on this matter can be found here:
https://www.amd.com/system/files/documents/rdna-whitepaper.pdf
https://en.wikipedia.org/wiki/RDNA_(microarchitecture)#Differences_between_GCN_and_RDNA

However, Metashape is not relying on this output, the task of parallel computations on all available compute units is performed by GPU driver. You can check GPU load using dedicated utilities, such as GPU-Z, for example.

Alex thank you for the explication.

Regarding the 'Texturing' problem, if it reduce the pixel size from the original 0.0037561 to, for example pixel size = 0.006, i dont get constantly the error (just a 1 or 2 times).

Hope it helps!


best regards,
JM