The key feature on intel platforms is the hardware de-noise acceleration (NVIDIA OptiX also works well.) Note, AMD OpenCL works quite well for some renderings, but blender flamenco likes consistent cluster hardware.
For 8k HDR10 media or 3+ screens the rtx 5090 32G model is going to be the minimum card people should buy. Just because you see 4 DP ports, doesn't mean the card can push bit-rates needed to fill an HDR10 display >60Hz.
The Mac Studio Pro unified >512GB ram/vram is a better LLM lab solution (Apple recently NERF'd it to 256GB.) Who cares if a task completes a bit slower, it doesn't matter given the lower error rates... and not costing $14k like an rtx 6000. =3
Great tutorial on getting blender to behave on mid-grade PC and laptops etc. :
Blender supports CUDA, HIP, OneAPI, and Metal. So Intel GPUs are performing poorly using their native API.
The key feature on intel platforms is the hardware de-noise acceleration (NVIDIA OptiX also works well.) Note, AMD OpenCL works quite well for some renderings, but blender flamenco likes consistent cluster hardware.
For 8k HDR10 media or 3+ screens the rtx 5090 32G model is going to be the minimum card people should buy. Just because you see 4 DP ports, doesn't mean the card can push bit-rates needed to fill an HDR10 display >60Hz.
The Mac Studio Pro unified >512GB ram/vram is a better LLM lab solution (Apple recently NERF'd it to 256GB.) Who cares if a task completes a bit slower, it doesn't matter given the lower error rates... and not costing $14k like an rtx 6000. =3
Great tutorial on getting blender to behave on mid-grade PC and laptops etc. :
https://www.youtube.com/watch?v=a0GW8Na5CIE