I am testing various setups for render machines. The current one is a dual Xeon system running Windows 10. I have 7 GPUs connected (GTX 10 series), which show up correctly in Windows device manager. In Blender (2.82a) system preferences I only see 6 GPUs under CUDA. Is there a limit within Blender on the number of GPUs?
There’s no hardcoded limit in Cycles. You could try running Blender with --debug-cycles
to see if there is more info.
Or use a utility like nvidia-smi or CUDA-Z to check if all those devices are actually available for CUDA.
Thanks, that narrows down the possibilities. I just tried on Linux (Ubuntu 19.10), and I am seeing all 7 of the cards. Once I finish the Linux tests, I’ll head back to Windows and see if I can figure out what’s going on.