# AWH memory usage

**URL:** https://gromacs.bioexcel.eu/t/awh-memory-usage/548
**Category:** User discussions
**Created:** [July 21, 2020, 8:38pm UTC](https://gromacs.bioexcel.eu/t/awh-memory-usage/548 "2020-07-21T20:38:40Z")
**Posts on this page:** 1
**Page:** 1

<div class="post-metadata">

### Author: ![genie](https://avatars.discourse-cdn.com/v4/letter/g/dfb087/32.png) [@genie](https://gromacs.bioexcel.eu/u/genie)
#### Post date: [July 21, 2020, 8:38pm UTC](https://gromacs.bioexcel.eu/t/awh-memory-usage/548/1 "2020-07-21T20:38:40Z")

</div>

GROMACS version: 2020.2  
GROMACS modification: No

Hello,  
I have encountered some unexpected behavior with the AWH functionality. I expected its memory usage to be more or less close to “a few arrays with one element per sampling point”. However, the GROMACS memory requirements actually increase dramatically with increasing force constant, soon hitting the physical memory limit (after which, predictably, it does not do anything useful and can only be terminated by KILL).  
The test setup has 3 dimensions, total sampling space is 7x8x8 nm.

> VIRT RES  
> 9.8g 0.4g no pull/awh active, all other GROMACS functionality  
> 18.1g 8.8g k=250 d=0.099 464k sampling points  
> 26.7g 17.4g k=400 d=0.078 948k sampling points  
> 33.7g 24.6g k=500 d=0.070 1312k sampling points  
> 58.4g 30.7g k=800 d=0.055 2705k sampling points

In addition, the checkpoint (CPT) files are quite large (3-4 **gigabytes** each in these test cases) while the energy (EDR) files where the AWH data are stored are quite reasonable (about 100 megabytes).

Might something be wrong here?