Learning Forum

[lq2556937429]

hello
We were amazed by the GPU's improvement in simulation speed when simulating FDTD. We would like to purchase a computer specifically designed for GPU accelerated Lumerical FDTD simulation. We have the following questions:
1. Does FDTD support parallel connection of two graphics cards to obtain larger video memory? For example, if we have a simulation file that requires 20GB of memory, can we use two 16GB RTX 4060Ti graphics cards to equivalent to a 32GB GPU and obtain simulation acceleration support for Lumerical FDTD
2. Is a better CPU still necessary? If it's not necessary, can we purchase a computer consisting of a good GPU and a poor CPU
Please note that the above issue is based on Lumerical FDTD 2024R1
If you have any other suggestions or tips, please let me know!
Looking forward to your answer

[Kirill]

Dear Subscriber,

• I strongly recommend upgrading to the latest version of Lumerical, as many improvements have been implemented since 24R1.
• Yes, having more GPU memory can be beneficial. Please keep in mind that using multiple GPUs may require additional licenses. For more information, refer to the article Getting started with running FDTD on GPU.
• Also note that certain parts of the simulation - such as meshing and data post-processing - still run on the CPU. Please ensure that this does not become a bottleneck and affect your overall simulation time.

Best regards,
Kirill