In our game we have about 1k drawcalls in nanite compute basepass and 2/3 of their instance are 0, and the nanite basepass took 5-8ms at some place on GTX1060. I want to know that if they took much time and does it worth to make a lot of change to asset to reduce the drawcall?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Detectnet training appears slower | 0 | 402 | November 3, 2016 | |
| Unexpected dramatic performance drop when doubling drawcalls | 1 | 1042 | June 23, 2016 | |
| memcopy time taken | 1 | 1338 | June 13, 2008 | |
| Extremely slow glDrawElementsInstanced compared to glDrawArraysInstanced | 0 | 1102 | May 28, 2019 | |
| why did my first cudaMalloc() cost so much time? | 1 | 4934 | November 12, 2009 | |
| Nvml overhead on latency of applications running on the GPU | 0 | 447 | August 17, 2020 | |
| Why my wall time is much worst using TITAN X than GTX 1050? | 0 | 385 | April 9, 2019 | |
| Strange timing of command buffers for d3d11 dispatches | 1 | 951 | November 8, 2013 | |
| Memory copy latency and kernel launch overhead | 3 | 7445 | May 17, 2011 | |
| Is there performance problem in CUDA and Windows? | 2 | 517 | March 22, 2017 |