Without logs, there is nothing we can do to help you.
Please provide:
%ProgramData%\Looking Glass (host)\looking-glass-host.txt
and the client command line output also
GitHub issues are for bug reports only, not performance/configuration issues. You have something wrong with your system configuration, on a properly configured system at 1200p LG never exceeds more than 10% GPU usage, and even less for the next release.
LG is certainly designed for gaming.
my apologies.
here are the host logs:
looking-glass-host.txt (542.3 KB) looking-glass-host-service.txt (15.2 KB)
and here is the client output: client loutput - Pastebin.com
Why are you forcing the legacy OpenGL renderer?
no reason in particular, at one point i thought it might make a difference and EGL might be the problem, this turned out not to be the case and i just forgot to remove that from my LG macro.
here is the output of â./looking-glass-client -f /dev/kvmfr0 -c 10.13.74.200 -p 5901 -F input:rawMouse=yes -g autoâ
different client renderers dont seem to have any impact on host application GPU usage whatsoever.
[I] 10606950427 main.c:1064 | main | Looking Glass (B4)
[I] 10606950468 main.c:1065 | main | Locking Method: Atomic
[I] 10606952244 ivshmem.c:127 | ivshmemOpenDev | KVMFR Device : /dev/kvmfr0
[I] 10607064154 egl.c:274 | egl_initialize | Double buffering is off
[I] 10607064204 main.c:671 | tryRenderer | Using Renderer: EGL
[I] 10607066385 x11.c:303 | x11Init | X11 XInput 2.0 in use
[I] 10607069381 x11.c:925 | x11GetEGLDisplay | Using eglGetPlatformDisplayEXT
[I] 10607180567 egl.c:634 | egl_render_startup | Multisampling enabled, max samples: 4
[I] 10607181644 egl.c:685 | egl_render_startup | Single buffer mode
[I] 10607186572 egl.c:701 | egl_render_startup | EGL : 1.5
[I] 10607186599 egl.c:702 | egl_render_startup | Vendor : AMD
[I] 10607186602 egl.c:703 | egl_render_startup | Renderer : AMD Radeon RX 5700 (NAVI10, DRM 3.42.0, 5.14.3-xanmod1-mathew2214-xankrnl2, LLVM 12.0.1)
[I] 10607186608 egl.c:704 | egl_render_startup | Version : OpenGL ES 3.2 Mesa 21.2.1
[I] 10607186611 egl.c:705 | egl_render_startup | EGL APIs : OpenGL OpenGL_ES
[I] 10607186612 egl.c:706 | egl_render_startup | Extensions: EGL_ANDROID_blob_cache EGL_ANDROID_native_fence_sync EGL_CHROMIUM_sync_control EGL_EXT_buf
fer_age EGL_EXT_create_context_robustness EGL_EXT_image_dma_buf_import EGL_EXT_image_dma_buf_import_modifiers EGL_EXT_swap_buffers_with_damage EGL_KHR_cl_event2 EGL_KHR_config_att
ribs EGL_KHR_create_context EGL_KHR_create_context_no_error EGL_KHR_fence_sync EGL_KHR_get_all_proc_addresses EGL_KHR_gl_colorspace EGL_KHR_gl_renderbuffer_image EGL_KHR_gl_textur
e_2D_image EGL_KHR_gl_texture_3D_image EGL_KHR_gl_texture_cubemap_image EGL_KHR_image EGL_KHR_image_base EGL_KHR_image_pixmap EGL_KHR_no_config_context EGL_KHR_reusable_sync EGL_K
HR_surfaceless_context EGL_KHR_swap_buffers_with_damage EGL_EXT_pixel_format_float EGL_KHR_wait_sync EGL_MESA_configless_context EGL_MESA_drm_image EGL_MESA_image_dma_buf_export E
GL_MESA_query_driver EGL_NOK_texture_from_pixmap EGL_WL_bind_wayland_display
[I] 10607403757 main.c:954 | lg_run | Host ready, reported version: B4
[I] 10607403795 main.c:955 | lg_run | Starting session
[I] 10607403926 main.c:371 | main_frameThread | Using DMA buffer support
[I] 10607585602 main.c:498 | main_frameThread | Format: FRAME_TYPE_BGRA 1920x1080 stride:1920 pitch:7680 rotation:0What is your CPU?
The physical host machine is a Ryzen Threadripper 1950x. running Debian 10 on debianâs Linux 5.10 with Vendor-Reset module added.
this setup worked flawlessly on B2, and i have changed nothing about the setup since i last used B2.
Are you using VM to VM here?
Note there was a regression with the kvmfr module, be sure you have upgraded it also. If still no success please update to bleeding edge and re-test.
yes i am using vm to vm. is that a problem? it just worked in B2, and seems to work fine in B4.
the VM for the Client is running Debian 11 buster with Xamodâs Linux 5.14.
here is the xml for the client VM: https://gist.github.com/mathew2214/aaa4d982ddf0e3332714ca3388871c08
i am using kvmfr module version 0.0.7. i will compile the bleeding-edge module from git, reboot and report my findings.
B4 changed how things work here, thatâs all.
I will compile the bleeding-edge module from git, reboot and report my findings.
Not just the module, upgrade the host and the client too please. Also there is no need to reboot, just rmmod/modprobe the new version.
updating to B4-565-gdf4a964496 make it noticeably better, but LG host application still uses around 80% of the host GPU according to task manager.
and the host still doesnt get gaming performance anywhere near what it gets without LG. it is better, but still far from playable.
AMD GPUs are known to not take into account the capture load in their performance power budgeting. We have had people reporting that they had to set the GPUs performance profile to maximum performance to rectify this issue.
Please also be aware that youâre on a NUMA CPU that has a large impact on GPU memory transfer performance to/from system ram if you have not configured things correctly, this could also be impacting your performance.
i cannot seem to find âperformance profileâ anywhere in Windowsâ settings or Radeon software.
my current NUMA config hasnt caused any issues before. that being said, i know for a fact that the 2 GPUs are on different NUMA nodes.
Sorry, I am not sure where these settings might be, I just know they have been reported in Discord a few times.
You might have better luck getting help there too as many people camp out in the LG discord with hardware configurations like yours and are more familiar with the issues that you may be facing.
Note that the experience youâre having is not the norm for LG, there is certainly something wrong with your setup to cause such a slowdown.
updated to the LG build 2021-10-18 16:17:54.
updated Radeon Software to 21.10.2
no change in behavior.
still waiting for a fix.