3 reasons to convert your HPC infrastructure to Tesla K80s today
- Details
- Category: Our Blog
- Created on 03 February 2015
We run extensive testing of K80s using GROMACS MD App. We used 4 different GROMACS models to run the comparison. And we do want to share our findings with you.
GROMACS models we used for the tests:
1.96pope + 32popg in water
2.ADH Cubic (ftp://ftp.gromacs.org/pub/tmp/adh_cubic.tar.bz2)
3.PROTEIN (ALPHA-AMYLASE)
4.B. subtilis model membrane
G10 HPC infrastructure we used for the test:
1)Next gen. G10 HPC based on NVIDIA Tesla K80
2)Current gen. G10 HPC based on NVIDIA Tesla K20
3 reasons why you should consider upgrading to G10 HPC based on Tesla K80 starting from today:
1. New G10 is 2 times faster
You’ll need 2-times smaller K80-based system to get the same performance as with K20-based system.
2. With G10 you save ~36% on CAPEX
Even though new Tesla K80 GPUs are more expensive than Tesla K20s, the cumulative effect of building G10 HPC system on K80s result in ~36% fund savings comparing to the system of the same performance built on Tesla K20s. In other words the cost of research just became more than 1/3 cheaper!
3. With G10 you save ~43% on electricity and cooling
We compared the power consumption of both systems under load and the resulting data showed that Tesla K80 based G10 consumes almost half of the electricity than the Tesla K20 based system!
Summary:
The next gen. G10 is not only twice as fast as the current gen. systems, but also is much more affordable. The cost of running new G10 is almost half of that of the current gen. New G10 consumes twice as less rack space and is a state of engineering art and technological progress.
HPC
