Escalating product complexity begets more complex simulations to prove out designs. This trend in turn generates numerically large computing problems. The continuous cycle around product development is upping the ante for high-performance computing (HPC) resources, and a longstanding partnership between ANSYS and Intel is trying to address this.
Intel recently released its Xeon E5-2600 processor platform, which is designed to power servers, workstations, and the next generation of storage and communications systems. And it worked with ANSYS to achieve the scalability and performance targets for using the ANSYS suite of simulation offerings on the new platform.
Officials at Intel said in a press release that the Xeon E5-2600 family can support up to eight cores per processor and up to 768GB of system memory, and the processors can improve applications and systems performance up to 80 percent over the previous-generation Xeon 5600 series. The new family also supports Intel's Advanced Vector Extension technology, which "increases the performance on compute-intensive applications such as financial analysis, media content creation and high performance computing up to 2 times."
Wim Slagter, lead product manager at ANSYS, told us:
The goal of our partnership is to offer ANSYS applications that deliver outstanding performance on Intel processing platforms such that joint customers can tackle ever more complex and high-fidelity simulations while still achieving the turnaround time required for product development decision making. For ANSYS software to effectively leverage today's computer hardware, efficient execution on multiple cores is essential. As core counts continue to increase, and with the advent of many core processors and availability of graphics processing units, ANSYS continuously has to make software architecture changes in order to maintain productive use of HPC.
That software architecture evolution includes refined support for parallel processing efficiency, and that's where the Intel relationship comes into play, Slagter said. The partnership's support of the E5-2600 and its ability to handle the additional cores (eight vs. six previously) and larger memory capacity (64GB vs. 32GB) will help ANSYS CAE customers significantly reduce simulation throughput times for the largest models.
What exactly does this mean when put in the context of real engineering work? Slagter said the Intel relationship is important because the resulting work lets customers conduct higher-fidelity simulations and study entire systems and the interaction among components. In layman's terms, Slagter said that means output with more geometric detail and the ability to simulate larger, more complex systems and physics. Enhanced performance throughput is another offshoot of the partnership. That means customers can consider more design ideas and tradeoffs.
Intel and ANSYS conducted a joint demonstration project to highlight the fruits of their partnership. In the "50:50:50" project, automotive engineers conducted CFD simulations to study the aerodynamics of 50 shape variants of a baseline vehicle model. The engineers leveraged morphing, advanced CFD solver numerics, HPC environments, and process automators in this project. The high-fidelity CFD simulations deployed a computational mesh of 50 million cells to simulate each design point. Such simulations would typically take 50 hours after initial setup. Using the Xeon E5-2600 processors, Slagter said, those simulations took 34 hours -- an improvement of 48 percent.
ANSYS and Intel are continuing other collaborative efforts in scaleout processing, including the MIC architecture, which combines numerous Intel processing cores on a single chip.