raul.salas Posted January 15, 2019 Share Posted January 15, 2019 (edited) Hi, I am working on app that needs 6x1 walls so I'm using AppWall on 2.5 version. I found an issue on the GPU performance in the profiler. It takes 14ms in the "Render GPU" field at the moment when i set the first camera even if there are no objects to paint. This is a big problem in performance when you have to load objects in the scene. I have tried the AppWall samples on version 2.5 and 2.7 and can be checked through the profiler how it affects. In the case of 2.7, in the field "Waiting GPU". I need to understand if the problem is mine or the engine to be able to put a solution in my application. Thanks. Edited January 15, 2019 by raul.salas Link to comment
silent Posted January 15, 2019 Share Posted January 15, 2019 Hi Raul, Which GPU are you currently using? You also can check with render_info command in console how many VRAM is used to render a single screen and then multiply it by six to calculate how many data you already transferring per frame. Probably that data transfer is the current bottleneck. Alternatively you can check VRAM usage with any 3rd party tool like GPU-Z. Thanks! How to submit a good bug report --- FTP server for test scenes and user uploads: ftp://files.unigine.com user: upload password: 6xYkd6vLYWjpW6SN Link to comment
raul.salas Posted January 15, 2019 Author Share Posted January 15, 2019 Hi Silent, Thanks for your response. I have a NVIDIA Geforce GTX 970. I tried to use the command in the AppWall sample and shows 330.7Mb on total. That's a screenshot of the profiler: Gpu-Z gives the VRAM info adding all processes, but if i open and close the app i can see an average of 600Mb on the difference. Is just the sample, i don't know how performance can decrease with nothing to paint. Link to comment
silent Posted January 15, 2019 Share Posted January 15, 2019 Hi Raul, Each additional screen in AppWall can result in significant performance drop. "Normal" resolution for GTX970 is rather FullHD than 6xFullHD screens (it also can be related to the PCI-E bandwith). If you can upgrade your GPU it would be the best option. However, 6 fullHD outputs sounds too heavy for a single GPU (at least with all rendering features enabled) for deferred rendering. Thanks! How to submit a good bug report --- FTP server for test scenes and user uploads: ftp://files.unigine.com user: upload password: 6xYkd6vLYWjpW6SN Link to comment
Recommended Posts