Hi everyone,

I’m currently mapping out an upgrade to a high-TDP (80W) Intel Panther Lake Mini PC, which I expect to have running by early February.

I’m very interested in how the Xe3 architecture (Arc B390) performs on the NumberFields@home GPU app. With the restored hardware-native FP64 support and the move to the 18A node, I'm expecting a significant increase in efficiency and PPD over previous Intel iGPU generations.

I’ll be running 64GB of DDR5-9600 to ensure the GPU tasks have a massive VRAM buffer and high memory bandwidth. I’m happy to share my initial clinfo logs and benchmarks once the hardware arrives in Denmark. Is there anything specific the science team needs to see to ensure these new chips are properly utilized and whitelisted?

There's nothing we need to see and your iGPU should receive tasks as long as you have beta apps allowed. I am curious to see how well it does.

Hi Eric,

​Thanks for the clarification!

I've gone ahead and enabled "test applications" in my project preferences to make sure the Panther Lake iGPU is ready to pull tasks as soon as it's online.

​I'll be sure to post the initial clinfo logs and some early performance data once the system is stable in February.

It will be interesting to see how the Xe3 architecture handles these specific number theory workloads compared to previous generations.

​Looking forward to sharing the results!