Support for SMP/MPP/GPU Solver

JMAG has a track record of continued pursuit for acceleration through the development and fine-tuning of calculation algorithms. JMAG utilizes the speed of single core acceleration. It also effectively uses hardware to deal with large-scale models to work in parallels.

JMAG released a GPU solver and High parallel solver with enhanced highly parallel processing (HPC) solution.

SMP Solver

Speedup in processing is widely expected in small to medium size models with the shared memory multiprocessing (SMP) parallel solver. JMAG is engaged in development that makes the best use of multi-core machines, such as reducing memory usage and assigning core for multi-thread processing.

Measurement model (partial)
Application Analysis type Element count
PM motor TR 0.35M
Generator TR 0.42M
Stepping motor TR 0.75M
Transformer FQ 1.0M

Case Studies:

Other, related materials

MPP Solver

This is a parallel solver that uses multiple nodes (machines). JMAG is developing technology to reduce traffic between the nodes. This technology enables the calculation of a large number of parallels in a cluster system. Speedup in processing massively parallels such as 1024 parallels is promising in models with elements on a scale of one million to tens of millions of elements.

Measurement model (partial)
Application Analysis type Element count
PM motor TR 5.1M
Axial gap motor TR 2.1M
Retarder TR 2.2M
Transformer FQ 2.0M
Wireless power supply FQ 2.0M
Induction motor TR 2.1M
Induction heating FQ 5.1M

Case Studies:

Other, related materials

GPU Solver

JMAG is focusing on GPGPU (General-purpose computing on graphics processing units) used for numeric calculations with the exception of image processing and provides solvers that support GPGPU.
Speedup is conducted efficiently by using GPU in calculation parts of the linear solver, as well as simultaneously using CPU parallels for other processing.
In particular, a noticeable increase in processing speed can be seen in a model where calculation time for a single step (there are a large number of elements or calculation for the linear solver takes time) takes a long time.

Measurement model (partial)
Application Analysis type Element count
Induction motor TR 1.4M
PM motor TR 1.1M
PM motor TR 1.4M
Speaker TR 1.2M
Magnetic head TR 1.1M
Retarder TR 1.3M
Induction heating FQ 1.0M
Bus bar FQ 1.5M

Hardware Specifications
GPU NVIDIA GeForce RTX 2080 Ti
CPU Intel Core i9-10800KF
Clock frequency (GHz) 3.7
Number of cores / processor 10
Number of processors / node 2
Memory (GB) 64

[ System requirements ]