I don’t think that is weird. Internally on our implementation the two methods should be generating similar HW instructions and I don’t expect more than 2-3% variance across runs.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Trouble with VK_NVX_device_generated_commands | 12 | 2716 | July 13, 2017 | |
| New: Vulkan Device Generated Commands | 0 | 546 | August 25, 2020 | |
| Strange timing of command buffers for d3d11 dispatches | 1 | 970 | November 8, 2013 | |
| CUDA vs DX execution times DX GPGPU code --> CUDA = slower | 15 | 13385 | January 30, 2008 | |
| Why is DirectCompute 2x faster than CUDA for my kernel? | 23 | 6733 | November 11, 2010 | |
| improved texfetch to exploit all of texture hardware | 21 | 9006 | May 4, 2007 | |
| Attention Lucky GTX 480/GTX 470 Owners! Please run some benchmarks for us. :) | 88 | 22763 | May 5, 2010 | |
| problem running demos | 9 | 8264 | January 1, 2009 | |
| Device generated commands extension support | 0 | 497 | September 3, 2020 | |
| Is GPU worth it? GPU currently too slow. | 16 | 6127 | December 8, 2008 |