garageCube

So my current setup is with an RTX2070 super with 3 datapaths + 1 control montor for a total of 13 outputs.

But I came across a pair of Quadro RTX4000 8GBs and decided to experiment. I removed the RTX2070, installed the dual quadros and installed quadro drivers. I didn't tinker with the nvidia software (is that required?). Is nDisplay required?

I saw the outputs immediately in madmapper - outputted the project successfully which is numerous video loops encoded in HAP.

Within minutes the primary GPU climbed to 80-95'c, while the 2nd gpu was at 45-65'C. I swapped the GPU's around on the motherboard, and the same thing happened again. The primary gpu remained extremely hot, while the 2nd gpu was relatively cool. Eventually, the computer would become unstable after about 1 hour of use.

I'm ready to throw away this experiment and just stick with what works. But I'm curious if there are Quadro settings that I should have enabled to get a better result?

On Windows, MadMapper will use only the GPU assigned to the Main Display in Windows settings. The only way to dispatch process accross multiple GPUs is to make them one, with NVLink or the AMD equivalent, but then you loose the outputs of the second GPU. Best is to focus on a good GPU and use splitters on the output

On Windows, MadMapper will use only the GPU assigned to the Main Display in Windows settings.Retro Bowl The only way to dispatch process accross multiple GPUs is to make them one, with NVLink or the AMD equivalent, but then you loose the outputs of the second GPU. Best is to focus on a good GPU and use splitters on the output

Why does MadMapper only use the GPU assigned to the Main Display on Windows, and is there any workaround to utilize multiple GPUs without losing display outputs?

There's a work in progress to avoid being linked to the "main display". But not for 2025, and not sure we'll dispatch processing over multiple GPUs but it would become possible (after getting rid of OpenGL)