OTOY Forums

I have a weird problem with network rendering:
The slaves don't run at 100% GPU usage when rendering an image. The speed goes up & down between 00 and 99, but most of the time it is at 00. See attached screenshot.
In the meantime the master is running at 100%.

Running the latest 2020.02 stable, with Nvidia driver version 461.4.
The GPUs are not overclocked, they are not running hot. Under Nvidia control panel/3D settings power management is set to "prefer maximum performance".
They are connected to the master, the text is green, the scenes were not too big (5-6GB VRAM is used).
In the Octane settings "minimize network traffic" is turned OFF. (it did the same if it is ON)
I haven't notice this behavior on earlier versions, but I didn't test it again.
Any ideas why this is happening?

Config:
master: Windows 10 Pro, ASUS Prime x299-A, i7-9800x, 96GB RAM, 2x RTX 3090
slaves (2): Windows 10 Pro, ASUS x99-A II, i7-5630k, 96GB RAM, 2x RTX 3090

Thanks!

Rather than being a BUG it should be about time spending on the sending data over network and having minimal time for calculating the image. You may have very intensive data and takes time for sending to clients. Frame calculation may finish on master before clients contribute to calculation.

How do you mean? I'm calculating ONE image, not an image sequence. Those spikes in the GPU usage are happening while rendering one single image.

Hatsize7 wrote:How do you mean? I'm calculating ONE image, not an image sequence. Those spikes in the GPU usage are happening while rendering one single image.

I'm very clear on explanation, it's very often scenario for heavy scenes on first frame.

Please use GPU-Z for monitoring the sensors during rendering.
In particular, check the GPU Load, Temperature, and Power Consumption TDP.

Looking forward for your feedback,

ciao Beppe