I am having some issues when using NVIDIA Nsight. When launching the GPU Trace Profiler I get either the error “GetInstanceProcAddr unable to locate vkCreateDebugReportCallbackEXT function.” or the error “vkCreateDevice: Failed to create device chain” depending on small changes in the logical device creation process I tried, followed by immediate program termination.
I have tried with Nsight versions 2022.1.1.22067 and 2023.2.0.23146. I have Windows 10 and driver version 522.25.
Since my project is on a github repo, I was able to use versions from 5-6 months ago, when I remember it was working fine, but I am experiencing the same issue.
Do you know what could cause this issue?
In my computer I have installed in the past other tools like NSight Compute and NSight Systems. Could it be related with some wrong system setup?
Thank you for using Nsight Graphics and I am sorry you ran into this issue with GPU Trace Profiler. I will contact the engineering team on your behalf and get back to you on your question.
I also encountered the same problem, please tell me how to solve this problem？Thanks.
It seems that vkCreateDebugReportCallbackEXT is part of VK_EXT_debug_report, which has been superseded by VK_EXT_debug_utils, so we would recommend moving to that extension if possible.
Can you let us know which GPU you were using and a pointer to your github repo if we want to reproduce?
Thanks a lot for your help.
I updated my framework to use VK_EXT_debug_utils (vkCreateDebugUtilsMessengerEXT / vkDestroyDebugUtilsMessengerEXT / VkDebugUtilsMessengerEXT) but unfortunately it is still failing. I have the console output information in the image provided below where three extra validation layers seem to be added by the GPU Trace Profiler (“VK_LAYER_NV_GPU_Trace_release_public_2023_2_0”, “VK_LAYER_bandicam_helper” and “VK_LAYER_NV_optimus”):
Could the issue be due to incompatibilities with other layers / extensions I have? Currently I use:
Instance extensions / layers: VK_KHR_surface, VK_KHR_win32_surface, VK_EXT_debug_report, VK_EXT_debug_utils, VK_KHR_get_physical_device_properties2, VK_LAYER_KHRONOS_validation
Device extensions: VK_KHR_swapchain, VK_KHR_maintenance3, VK_KHR_get_memory_requirements2, VK_KHR_ray_tracing_pipeline, VK_KHR_acceleration_structure, VK_KHR_ray_query, VK_KHR_buffer_device_address, VK_KHR_deferred_host_operations, VK_EXT_descriptor_indexing, VK_KHR_pipeline_library, VK_KHR_dedicated_allocation, VK_EXT_scalar_block_layout, VK_EXT_debug_marker
Unfortunately, currently my framework’s code is not publicly available.
I did some tests on my framework, removing all extensions, and making an older instance version (1.2.162) with the hope I would go past the call vkCreateDevice but still fails when using GPU Trace Profiler.
Out of curiosity, I tried with one demo from the SaschaWillems repo GitHub - SaschaWillems/Vulkan: Examples and demos for the new Vulkan API “deferred” in Release and it crashes as well, with the error “Could not create Vulkan device” which happens when vkCreateDevice fails in VulkanDevice::createLogicalDevice. I tried other demos from this repo and they crash as well in both Debug and Release.
I tested on NVidia Nsight 2023.2.0.0, with driver version 522.25. I have a RTX 2060 and Windows 10 Pro version 22H2.
Could this be a system issue? I have installed several products and versions in the last 3 years of your products in my computer so maybe something could have got misconfigured.
I have an update. I tested a matrix of different NVidia driver versions (472.12, 522.25 and 536.99) and NSight Graphics versions (2021.2, 2022.1, 2023.3). Both Frame Debugger and GPU Trace work fine for 522.25 and 2022.1, in the other cases I tested it does not work or it crashes.
Thanks for testing with the latest version of Nsight Graphics and the recent driver version and sorry you are running into these issues. Can you provide a repro of the crash you saw with Nsight Graphics 2023.3 and Driver version 536.99? You can send an email to email@example.com so we can discuss more there.