RADIOSS Update and Intel Haswell Performance

I’m happy to share the latest developments for RADIOSS, including some dramatic performance improvements in recent testing with Intel!

First, HyperWorks 13.0 is about to be released officially; for solvers like RADIOSS, this will mean even better performance that you’ve grown used to expecting with this fast, scalable solver. We are continuing to push the scalability of the code under hybrid parallelism based on MPI and OpenMP, which makes RADIOSS one of the fastest solvers out there — and one of the best-suited to take advantage of the high-performing Intel® Xeon® line. In particular, we’ve finely tuned the OpenMP parallelization in order to scale further on latest high-end multi-cores from Intel.

Intel E5-2600 v3, codename Haswell, is a great platform to test and tune RADIOSS speed, scalability and performance. This new processor has just been announced by Intel this week… and I cannot resist posting our first performance numbers observed with RADIOSS:

While for other CAE applications an improvement of 1.25X is generally noticed, with RADIOSS we achieved a superior speedup of more than 1.4X!

This acceleration is achieved when comparing RADIOSS v12.0 running on one node of Intel E5-2696 v2 at 2.5 Ghz with 2×12 cores to RADIOSS v13.0 running on one node equipped with this new E5-2697 v3 at 2.6 Ghz with 2×14 cores.

See below for the performance results from this testing:


This exceptional performance gain emphasizes the continuous effort Altair dedicates to optimizing our code for the latest high-performance computing (HPC) technologies — and the fruitful long term collaboration we have with Intel. Highly scalable hardware together with highly scalable software is the key to achieving such speed and performance improvements!

Users can achieve this tremendous scalability and level of performance when RADIOSS is deployed on very large scale clusters powered by efficient processors. Again, what makes this possible is RADIOSS’s massively parallel architecture — using Intel MPI coupled with OpenMP to achieve a Hybrid Massively Parallel Program (HMPP), for the best scalability in the marketplace.

We’re proud that RADIOSS is the most scalable solver in its class, and one of the top solvers in auto, aero, consumer goods, electronics, sports and other industries. Read more about RADIOSS performance evolution on past Xeons here.

Eric Lequiniou
Eric Lequiniou

About Eric Lequiniou

An expert in software optimization and parallelization on clusters, multi-core architectures, accelerators and coprocessors, Eric developed the Hybrid MPP parallel version of RADIOSS finite element software. After receiving an MSc degree in computer science, Eric started his career in 1994 at CNRS, Laboratoire Informatique du Parallélisme. He joined Mecalog in 1994 and worked for this company until its acquisition by Altair in 2006. Eric holds an Executive MBA from HEC French business school obtained in 2007.