1

To be brief. I'm rendering relatively small scenes (mostly crystal structures) in blender, and I recently got an RTX 2060. Interestingly, when I use the default settings (with auto-tile size) Blender 2.79 indicates that my GPU has 12 threads. This results in noticeably worse performance as shown below:
12 Threads = 2:59.02
1 Thread = 0:56.30
Does anyone know why this is the case? Is it a case of hardware misidentification? The difference is pretty significant.

Thanks

BesselFunct
  • 121
  • 1
  • https://blender.stackexchange.com/questions/63151/why-does-cycle-rendering-only-use-one-core-from-an-amd-gpu – Duarte Farrajota Ramos Apr 10 '19 at 16:29
  • 1
    This is not what I'm asking. I realize that CPUs render with multiple cores, while GPUs tend to render with a single "super"-core. What's happening is that Blender is rendering with 12 threads with my GPU by default, which I didn't think was possible. Is it subdividing the single large GPU core into multiple smaller units, each with its own dedicated VRAM? – BesselFunct Apr 10 '19 at 17:09
  • @BesselFunct Pretty interesting. I've tried it on RTX 2080 and there are no differences but as well Blender has in Auto-Detect values from CPU by default. – cgslav Apr 10 '19 at 21:50

0 Answers0