To test RADIOSS, I used a public domain finite element model called Neon refined with 1 million of elements. It represents a frontal car crash on a wall (you can see a video demonstration of RADIOSS here). I ran RADIOSS v12.0 leveraging the 60 cores of this 4-way Intel® Xeon® processor E7-4890 v2 @2.80GHz.
The computing power is really incredible, with a performance improvement of more than 2.35 times compared to the previous generation based on 4-way Intel® Xeon® processor E7-4870 @2.40GHz (codenamed Westmere EX)!
And even better, when comparing my results to a latest 2-way Intel® Xeon® processor E5-2696 v2 @2,5GHz (codenamed Ivy Bridge EP), it shows a tremendous performance increase of more than 2.75! You can see these results in the figure below:
One thing that makes this performance possible is the fact that RADIOSS is one of the most scalable solvers in the market. RADIOSS v12.0 is the so-called HMPP (Hybrid Massively Parallel Program) version based on Intel MPI and OpenMP, a highly parallel version that offers a unique, proven method for rich scalability over thousands of cores for finite element analysis (FEA). With RADIOSS v12.0 on Intel® Xeon® processor E7-4890 v2, engineers can achieve higher efficiency on large HPC clusters, with easy tuning of MPI & OpenMP and perfect repeatability in parallel.
For our clients, running RADIOSS on the latest generation of Intel® Xeon® processor(s) opens new ways of designing their products — using more accurate simulations, running more parametric studies for optimization of the structures and confidence on results… and at the end, producing better and safer products for end-customers.