Melyon Posted April 18, 2016 Share Posted April 18, 2016 Hi everybody, We have developed an online 3D city called Melyon (www.Melyon.ir ).But we have a little problem here: When roaming in the city, no matter in which scene we are, from the 512 MB memory dedicated by GPU to the browser, the max amount engaged in processing is just about 5 MB or less and the rest of memory needed is provided directly from RAM wich is as you know slower than GPU memory. Why is that and how we can fix it? We want most of the resources needed for processing comes from that 512 MB GPU memory dedicated for Browser and as much as possible not to engage Ram directly. Thanks for your answers. Quote Link to comment Share on other sites More sharing options...
JCPalmer Posted April 18, 2016 Share Posted April 18, 2016 All memory is slow. Memory access is the new I/O. It is 100 times slower to have to wait for memory access as opposed being in a processor cache. Memory access sucks, period. That said, memory is not fungible across processors, unless it is "pinned" meaning it is used by both to transfer data. If you are disappointed with your performance, start profiling to reduce CPU in the render loop. A good start is reducing the number of draw calls by merging meshes of similar material. This both reduces GPU draw calls, and CPU overhead of of managing the drawing process. Having too many objects of the same material is usually biggest bang for the buck. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.