# What hardware do you recommend for COMSOL Multiphysics?

 Solution Number: 866 Title: What hardware do you recommend for COMSOL Multiphysics? Platform: All Platforms Applies to: All Products Versions: All versions Categories: Installation Keywords: hardware recommendation license memory

## Problem Description

I am going to buy a new dedicated computer for running COMSOL Multiphysics. What hardware do you recommend? What about CPU speed and amount of RAM?

## Solution

Selecting appropriate hardware for optimal COMSOL Multiphysics performance is dependent primarily upon the size of the models that you wish to solve. Different models have different optimal hardware, so it is advised to choose your hardware based upon the largest models that you anticipate solving.

### Model Size and RAM Memory Requirements

The single most important factor in choosing hardware is to determine the physical memory (RAM) needed to solve your models. If the amount of memory needed to solve the your models exceeds the available physical RAM then you will notice a marked decrease in solution speed. Predicting RAM requirements is done by solving smaller models that contain the same physics that you want to solve in your largest models. Monitor the memory used and the degrees of freedom from these smaller models. Fit a curve of the form A x (dof)^N, where A and N are fitting coefficients and dof is the number of degrees of freedom, to the data and use this to predict the memory requirements for your larger models.

Be aware that memory usage versus degrees of freedom can be very different between combinations of physics, so you may want to repeat this procedure for every type of model you wish to solve.

### Hardware Recommendations

Once you know your memory requirements, you can select optimal hardware

#### Models using 2GB or less

CPU speed is the most important in this range. Purchase a single-processor system with up to four cores and high clock speed. Use at least dual-channel memory bus architecture, and make sure that two RAM chips (DIMM's) are used, one per each memory channel.

#### Models using between 2-16GB RAM

Memory bandwidth is the most important in this range. Purchase a single-processor (single-socket) system with at least four cores, a quad-channel memory bus architecture and the fastest possible memory bandwidth. Bandwidth is often reported as the QPI, measured in GigaTransfers/second (GT/s).

Make sure that the memory channels are evenly populated, there should be one DIMM per memory channel for a total of four. Most hardware vendors will do so by default. Make sure not to use less DIMM's than memory channels. For example, if you anticipate needing 16GB RAM, purchase a computer with 4 x 4GB DIMM's.

Speed with respect to number of cores in the CPU will increase linearly up to four cores, beyond which there will be a smaller increase (and in rare cases, even a decrease) in speed as more cores are used.

#### Models using between 16-256GB RAM

In this range there is a benefit from adding an additional CPU, which will provide additional memory channels. Purchase a dual-processor (two-socket) system with quad-channel memory bus architecture. Make certain that the memory channels are evenly populated, and that no more than two of the four open slots per memory channel are used, since the memory speed will be reduced if more than two slots per memory channel are populated.

For example, if you anticipate needing 128GB RAM, purchase a dual-processor computer with either 8 x 16GB DIMM's and make sure that there is one DIMM per each of the eight memory channels, or 16 x 8GB DIMM's, with two DIMM's per memory channel. There is no advantage in using 4 x 32GB DIMM's since this will leave four memory channels unused.

There is no benefit at this time of going to four-socket or eight-socket systems since these do have slower memory bus speed than two-socket systems.

#### Models using greater than 256GB RAM

In this range solving on a distributed memory computer (a cluster) becomes a good option. Please contact COMSOL Technical Support for personalized guidance.

### Parametric Sweeps

If you plan to solve for many geometric variations, different sets of materials, different parameters or excitation frequencies within each unique model you will be using the Parametric Sweep functionality. For example, a sweep over 10 variations of a part dimension along with a sweep over 10 different materials and 10 different model parameters would require solving a similar model 1000 times, and the solution time will be just about exactly 1000 times greater.

If you expect to be performing many parametric sweeps it is highly advised to use a Floating Network License which will allow you to distribute your parametric sweeps on a cluster computer using as many nodes as you have available. You can run on your own cluster or use remote resources such as the Amazon EC2. Each node of the cluster need only meet the requirements described for running a unique model. For further guidance on cluster hardware, see Knowledge Base 1116.

Always consider if you can avoid large sweeps by using the Optimization Module.

### General Recommendations

The following recommendations are general and apply to all cases described above.

#### OS

A 64-bit OS is recommended and needed for models needing more than ~1.5GB RAM. The choice between Windows, Linux, or Mac OS X will not significantly affect performance.

#### Graphics

We recommend modern AMD FirePro or NVIDIA based dedicated graphics cards. A list of tested graphics cards can be found on the system requirements page. The larger the memory in the graphics card, the more complex models can be visualized. Note that models will large memory requirements to solve need not necessarily require large video cards. When using the FNL it is also advantageous to use Client-Server mode.

#### GPU's

General-purpose computing on graphics processing units (GPU, also referred to as GPGP and less often GP²) is currently not supported by COMSOL.

#### Hard Drives

Hard drive speed does not affect COMSOL performance as long as the problem fits into main memory. If main RAM memory is insufficient, the hard drive swap will affect performance severely, regardless of which hard disk you have.