Some experiments with Vulkan API: And it should be the opposite, isn't it? Go on: 80 Mb difference! Vulkan build is significally larger, works slower and unstable. After 1 minute game crashes.
which version of Unity did you try and on what device? The initial implementation we released in Unity 5.6 surely isn't perfect. We made various improvements (stability, performance, features) in more recent versions. Also some of the earlier Vulkan drivers have issues (stability and performance). Regarding size, Vulkan SPIR-V shaders tend to be quite a bit larger than GLSL shaders. Also in some earlier versions we accidentally always included GLSL shaders. I'll have to check if the fix for that was back-ported.
Hi! Thank you for reply. I using most recent version 2017.2. Device - Xiaomi Redmi Note 4 (Snapdragon 625). Please note, that device's Vulkan API implementation is fine. I tried UE4 SunTemple demo (GL ES3.0 and Vulkan) and Vulkan build works better. I can share video if it is necessary.
Whoops - just tried a Vulkan build with Unity 2017.3, and the frame rate drops from 60+ to 20... switching back to DirectX 11 for now.
Still on that project? I recently discovered vulkan so I don't know that much about it, I'm using unity 2017.4, please tell me something : did you test on this version as well? If yes how did it do? Secondly, what will you recommend me to use as of today (considering that I be upgrading to unity 2018) Thank you
Sorry, but I haven't looked at it again since my 2017.3 test. I was just experimenting with it at that time. I honestly don't know too much about it either.
Let's keep up with the necro: On Unity 2018.2.17, still the same poor performance on Android Vulkan. Switching from GL ES 3.1 to Vulkan cuts the fps to less than half! (from 60+ to 22!)
My friends OnePlus 5 has better performance on Vulkan vs OpenGL3, and my Pixel XL has worse performance on Vulkan vs OpenGL3... Soo, Vulkan isn't even consistently worse...which is a bummer because we can't just omit Vulkan from our Android builds...but it'll also be impossible to test every possible device to see whether it runs better with Vulkan or OpenGL3. Not to mention that even if we did have a master list of all Android devices and which performs better with what, it's not like we have scripting API access to decide which graphics API will be used, right? (ie I can't write if(Pixel XL) use(OpenGL3) else use(Vulkan)).
True, I forgot about that! But it is still not practical. I can't test every single device in existence.