Siemens just lately introduced that its Simcenter STAR-CCM+ multi-physics computational fluid dynamics (CFD) software program now helps AMD Intuition™ GPUs for GPU-native computation. This transfer addresses its customers’ wants for computational effectivity, diminished simulation prices and power utilization, and higher {hardware} selection.
Liam McManus, Technical Product Supervisor for Simcenter STAR-CCM+, stated, “Our prospects need to design quicker, consider extra designs additional upstream, and speed up their general design cycle. To try this, they should improve the throughput of their simulations.”
The Simcenter STAR-CCM+ workforce was naturally within the AMD Intuition MI200 collection, together with the MI210, MI250, and MI250X. McManus stated, “We had loads of familiarity, expertise, and success with AMD CPUs. This made us snug exploring what we may obtain with AMD GPUs.”
GPUs speed up backside strains
The computational depth of CFD traditionally burdens conventional CPU-based programs. Whether or not predicting the airflow round a brand new automotive mannequin or optimizing the cooling programs for cutting-edge electronics, there’s at all times a want for quicker design cycles—a problem for industries the place time-to-market and product efficiency are essential. McManus added, “At present, it’s not nearly simulating a part as soon as. A simulation may be run 100 occasions to optimize it and get probably the most environment friendly product doable.”
Siemens discovered that with AMD Intuition GPUs, CFD simulations that when took days might be accomplished in hours and even minutes with out compromising the depth or accuracy of the evaluation. McManus identified, “GPU {hardware} permits us to run extra designs on the similar {hardware} value or begin to take a look at increased constancy simulations throughout the similar timeframe as earlier than.” This newfound velocity permits a extra exploratory strategy to design, permitting engineers to check and refine a number of hypotheses within the time it as soon as took to judge a single idea.
AMD Intuition MI200 GPUs stand aside
The AMD MI200 collection progressive CDNA2 structure gives excessive processing speeds and optimizes power consumption, permitting for environment friendly dealing with of huge datasets and complicated calculations. Superior options corresponding to high-bandwidth reminiscence (HBM), scalable multi-GPU connectivity, and enhanced computational precision collectively improve the GPUs’ efficiency and effectivity throughout various computational duties. The MI250 is additional optimized for the very best efficiency ranges in demanding duties, together with large-scale simulations (HPC), Deep Studying, and complicated scientific calculations. Engineered for scalability and big parallel processing (MPP) skills, the MI250 excels in Excessive Efficiency Computing and synthetic intelligence (AI) workloads as a consequence of its distinctive computational throughput, reminiscence bandwidth, core rely, quick reminiscence, and reminiscence capability.
“Only one AMD Intuition GPU card can present the computational equal of 100 to 200 CPU cores,” stated McManus. In fact, we are able to use a number of GPUs, which means that we are able to supply prospects considerably diminished per-simulation prices.”
Michael Kuron, a Siemens senior software program engineer who led the port, emphasised, “One factor that makes AMD GPUs nice is their excessive reminiscence bandwidth. For CFD, we’re probably not restricted by pure numerical efficiency however by how briskly the GPU can shuffle the info. AMD GPUs supply among the highest reminiscence bandwidth on the market, making them a wonderful platform for CFD functions.” He added, “A number of the world’s quickest supercomputers today use AMD GPUs, so with the ability to run on them actually doesn’t damage.”
AMD ROCm and HIP clean the transition
In fact, {hardware} was solely a part of the consideration. McManus stated, “The AMD ROCm platform has been vital in making certain that our software program may absolutely leverage the computational energy of AMD GPUs. Its open-source nature and complete toolset have considerably eased the event and optimization of our functions.”
Kuron added, “As a result of all the ROCm stack is open-source, I can look below the hood and make things better with out ready for any technical help.” Kuron continued, “Within the ROCm ecosystem, all of the runtime and math libraries, plus all of the stuff constructed on high of these, are open supply. We now have glorious perception when new options and capabilities are available.”
ROCm™ software program’s HIP programing language enabled a clean transition of Simcenter STAR-CCM+’s current codebase. Kuron defined, “Our current CUDA software program interprets nearly one-to-one to HIP, so the porting effort was a lot decrease than rewriting it in one other programming mannequin like SYCL or OpenMP offloading. The precise change between CUDA and HIP was simply a few hundred strains of code. Most likely 95% of the change from CUDA to HIP was achieved utilizing little greater than discover and substitute, and the remaining wasn’t troublesome both.”
Kuron stated, “Reaching one-to-one parity was a major milestone that ensures our software program delivers exact and dependable outcomes persistently, whether or not working on AMD or some other {hardware}.”
Collaborating to serve the client
Collaboration was pivotal to the venture’s success. “AMD was very aware of our suggestions, working carefully with us to refine the mixing,” famous Kuron. “The chance to speak immediately with the AMD workforce members who implement these options and perceive the technical particulars has been extremely useful.”
McManus stated, “It’s nice to collaborate with AMD. They’re creating the GPU options and we are able to work carefully with them to make sure our software program runs on it. Siemens and AMD have the identical goal: to get the client to the reply as quick as doable.”
Waiting for MI300 and past
Seeking to the brand new AMD Intuition MI300 collection, Kuron stated, “We’re wanting ahead to the rise in reminiscence bandwidth on the MI300 platform. The tighter coupling between CPU and GPU of the MI300A platform may assist get rid of bottlenecks and velocity up simulations that require some components to run on the CPU.”
McManus provides, “The rise in reminiscence capability, as much as 192 gigabytes for the MI300X, will cut back constraints on simulation complexity and permit bigger drawback sizes to be addressed extra successfully.
We’re additionally exploring hybrid computational methods for some CPU-bound simulation challenges and we’re significantly intrigued by the chances provided by the unified reminiscence of the MI300A.”
Collectively, Siemens and AMD are addressing the evolving wants for faster, more cost effective design processes. Integrating Simcenter STAR-CCM+ with AMD Intuition GPUs broadens the vary of instruments out there for computational fluid dynamics challenges, providing excessive simulation velocity and price effectivity and providing engineers a wider array of {hardware} choices. The AMD MI300 collection guarantees to develop these capabilities additional, catering to an more and more various and complicated array of simulations, and really dynamic markets.
Disclaimers
The data introduced on this doc is for informational functions solely and should comprise technical inaccuracies, omissions, and typographical errors. The data contained herein is topic to vary and could also be rendered inaccurate for a lot of causes, together with however not restricted to product and roadmap modifications, part and motherboard model modifications, new mannequin and/or product releases, product variations between differing producers, software program modifications, BIOS flashes, firmware upgrades, or the like. Any pc system has dangers of safety vulnerabilities that can’t be utterly prevented or mitigated. AMD assumes no obligation to replace or in any other case right or revise this info. Nevertheless, AMD reserves the fitting to revise this info and to make modifications infrequently to the content material hereof with out obligation of AMD to inform any particular person of such revisions or modifications. THIS INFORMATION IS PROVIDED ‘AS IS.” AMD MAKES NO REPRESENTATIONS OR WARRANTIES WITH RESPECT TO THE CONTENTS HEREOF AND ASSUMES NO RESPONSIBILITY FOR ANY INACCURACIES, ERRORS, OR OMISSIONS THAT MAY APPEAR IN THIS INFORMATION. AMD SPECIFICALLY DISCLAIMS ANY IMPLIED WARRANTIES OF NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR ANY PARTICULAR PURPOSE. IN NO EVENT WILL AMD BE LIABLE TO ANY PERSON FOR ANY RELIANCE, DIRECT, INDIRECT, SPECIAL, OR OTHER CONSEQUENTIAL DAMAGES ARISING FROM THE USE OF ANY INFORMATION CONTAINED HEREIN, EVEN IF AMD IS EXPRESSLY ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. AMD, the AMD Arrow emblem, [insert all other AMD trademarks used in the material here per AMD Trademarks] and combos thereof are logos of Superior Micro Gadgets, Inc. Different product names used on this publication are for identification functions solely and could also be logos of their respective firms. [Insert any third party trademark attribution here per AMD’s Third Party Trademark List.]
Third-party content material is licensed to you immediately by the third celebration that owns the content material and isn’t licensed to you by AMD. ALL LINKED THIRD-PARTY CONTENT IS PROVIDED “AS IS” WITHOUT A WARRANTY OF ANY KIND. USE OF SUCH THIRD-PARTY CONTENT IS DONE AT YOUR SOLE DISCRETION AND UNDER NO CIRCUMSTANCES WILL AMD BE LIABLE TO YOU FOR ANY THIRD-PARTY CONTENT. YOU ASSUME ALL RISK AND ARE SOLELY RESPONSIBLE FOR ANY DAMAGES THAT MAY ARISE FROM YOUR USE OF THIRD-PARTY CONTENT.
© 2024 Superior Micro Gadgets, Inc. All rights reserved.