NVIDIA’s Post

Lattice Boltzmann method (LBM) offers a completely novel take on #computationalfluiddynamics. By using methods derived from statistical mechanics (Boltzmann transport equations), we can calculate the velocities of individually discretized particle parcels and then recover the overall fluid motion and macroscopic properties. This is completely different than finite volume methods that use brute-force differential equation solving on macroscopic fluid equations. Why do you care? Because LBM is inherently parallelizable and therefore **made** for GPU architecture. We can send each cell to its own GPU thread.... and modern NVIDIA GPUs have thousands of cores. (If you don't know know who NVIDIA is, then maybe you should just retire because you haven't been paying attention to technology or the stock market). Those particle parcels come in individual topological flavors according to the number of speeds connecting each neighbor lattice (Q) and the dimension of the discretization (D). That's what D3Q19, D3Q27 or D2Q9 is referencing. Because the grid is regular and Cartesian, we easily add a subgrid turbulence model (Smagorinsky LES, usually) by summing the eddy viscosity to the molecular viscosity. There's rumors that the higher dimensional creatures are simulating us on a D5Q119 hyperlattice, but this post is too small to contain proof of the conjecture. Do you need CFD consulting, finite element analysis, ASME pressure vessel design, heat exchanger analysis, welding engineering or a professional engineer to just stand around and point at things? Contact: ben.turner@latticept.com #computationalfluiddynamics #cfd #smagorinsky #hyperlattice #asme #nvidia

Ansys Enables #3DVisualization of Next-Gen 3D-IC Designs with NVIDIA Omniverse https://ow.ly/OioC50SlW2Y #HPC

If you dont think that the Blackwell GB200NVL72 will begin to change everything, you need to look at the performance increase over the Hopper: 1. NVIDIA Blackwell vs. Hopper Efficiency: NVIDIA's GB200 Grace Blackwell Superchip, unveiled at GTC, provides generative AI on trillion-parameter LLMs with up to 25x less cost and energy consumption compared to the previous Hopper architecture, driving significant efficiency improvements. 2. Enhanced Double-Precision Performance: Blackwell GPUs deliver 30% more FP64 and FP32 FMA (fused multiply-add) performance than Hopper, crucial for scientific computing and physics-based simulations that rely on double-precision formats. 3. Faster Analog Circuit Design: The Cadence SpectreX simulator runs analog circuit design simulations 13x quicker on the GB200 Grace Blackwell Superchip compared to traditional CPU-based systems, enhancing design efficiency and speed. 4. Superior Computational Fluid Dynamics: Cadence Fidelity software running on GB200 systems can execute CFD simulations up to 22x faster than on CPU-powered systems, thanks to the advanced capabilities of the Blackwell architecture. 5. Significant Speedup in AI Inference: The GB200 architecture's second-generation transformer engine offers a 30x speedup in inference workloads for large language models like the 1.8-trillion-parameter GPT-MoE, compared to the H100 generation, unlocking new possibilities for high-performance computing applications. What do you think what will happen with the Rubin? Infrastructure Masons Compass Datacenters Hewlett Packard Enterprise NVIDIA IBM Dell Technologies #HPC #NVIDIA #BLACKWELL

Discover how the new #CUDA math library NVIDIA cuDSS 0.2.0 (Preview) was integrated into Honeywell Unisim Design to achieve up to a 78x performance increase over traditional sparse linear equation solvers. https://nvda.ws/48Nazga

You don't have to choose between accuracy, speed and visualization with M-Star Simulations. All three are standard features of our GPU-native CFD solver that runs exclusively on DESKTOP NVIDIA cards. Harlley Parno shows off again by validating the pressure drop of a shell and tube heat exchanger at multiple flow rates. In this simulation, we use the experimental data of Chen et al (2019). (Source: https://lnkd.in/eWaeNuv7) Because M-Star is inherently transient, we actually script the flow rate as a step impulse that changes as the simulation runs. We can then watch the pressure drop respond exactly as you would expect (overshoots and all) and compare it side-by-side to the lab data as a reference point of accuracy. The 'noise' is the actual turbulent field captured by our subgrid model. To make this simulation even more mind-blowing, we get within THREE PERCENT of all experimental values with just SIX VOXELS between the smallest gap. That's a very coarse grid on a very hard problem! How is this even possible? Well, because our solver guy (Brian DeVincentis) crafted his own novel filtered wall function to stabilize pressure oscillations streaming from the wall. Not only is it extraordinarily numerically stable, but also works with our fourth-order accurate cumulant turbulence model. Only 25% of the total turbulent energy is explicitly resolved by the filter in these simulations. John Thomas built an incredible piece of software that's designed to specifically run on the fastest hardware available: NVIDIA GPU's. Contact me to learn more or how Latticept can transform your industrial simulations. #cfd #computationalfluiddynamics #gpu #hpc #nvidia #heattransfer

COUPLED DEM + CFD ON NVIDIA GPU's WITH M-Star Simulations! Air on the left. DEM on the right. This is actually a fluid pump, pellet conveyor and a packed bed all in one! I simulated this in a few hours this morning on a desktop RTX4090 using Dotmatics incredible GPU-native solver. Harlley Parno's love of the cyclone separators inspired me to make a machine that does absolutely nothing. The vortex that forms at the top is beautiful and totally unexpected, but shows the power of being able to visualize time-accurate physics in corporate engineer (not academic) timescales. This is an inherently parallel lattice Boltzmann algorithm with LES turbulence interacting with a million particles. These range in sizes from 1 to 10mm. At every single time step the particles create turbulence AND the particles move from the turbulence. The screw conveyor moves the pellets to the top of the bin AND moves the air with it, creating an area of low pressure that sucks the air through the bed. Have a hard problem in the chemicals, oil/gas, pharmaceutical, food or basically any other industry? Do you live inside of a fluid like I do and interact with it every day? Then you probably have a use for M-Star or my consulting services. John Thomas created it from the ground up to run on the most powerful hardware that exists: the NVIDIA GPU. Learn more at: www.mstarcfd.com DM me here for consulting services or ASME pressure vessels or email: ben.turner@latticept.com #cfd #computationalfluiddynamics #hpc #gpu #nvidia #cfddem

Did you hear? Ansys Speos GPU has been recognized among the best by the 2023 Laser Focus World Innovators Awards. An esteemed and experienced panel of judges from the optics and photonics community recognized Speos GPU as a Silver honoree! https://ansys.me/3P9Px3K Ansys Speos creates accurate virtual prototypes for design and validation of illumination, imaging, and display systems, from automotive lighting to augmented reality devices. Speos is the premiere #optics simulation software for #Photometry/#Radiometry simulation in every domain where light is involved. It allows design of #illumination systems as well as virtual validation through any type of measurement, including radiance for proper human perception or camera sensor. Speos #GPU takes the capabilities from traditional Speos #simulation to the GPU hardware for an increase in speed that not only accelerates result delivery but revamps the overall design and validation experience. All achieved while maintaining the accuracy of the simulation. With overall increased simulation speed of x60, complex simulations that took hours or days now take minutes, shortening the time to market while allowing more innovation time. Included in the Ansys 2023 R2 Release are even more Speos features that enable designers to maximize their innovation thanks to Speos GPU. Here are some highlights. Interactive Live Preview Speos 2023 R2 introduces a brand-new explorative workflow using Live Preview. Any modification in the definition of Speos objects is seen almost instantaneously thanks to the update button powered by Speos GPU. Speos GPU - ray file support Simulations defined with ray file sources now run with Speos GPU direct simulation, providing the same result accuracy with GPU speed. Any illumination project can now benefit from GPU speed thanks to the support of ray file. 3D Irradiance sensor 3D irradiance sensor permits designers to analyze light contributions for each face of a geometry by providing a 3D illumination map. Each face becomes an irradiance sensor. Simulations defined with 3D Irradiance sensors can now run in Speos GPU direct simulation, providing the same result accuracy with GPU speed. Output faces of irradiance sensor When irradiance sensor is used in inverse simulation, rays are emitted in all directions, which may not be efficient in some cases, such as camera lens objective. Output faces permit to set the faces rays starting from the sensor are going to aim during the inverse propagation to improve performances. This functionality of the irradiance sensor is now compatible with Speos GPU inverse simulation, providing the same result accuracy with GPU speed. Join our upcoming webinar to learn more about the features in the Ansys Speos 2023 R2 Release! Ansys 2023 R2: Ansys Speos What’s New September 13, 2023 / 8:00am & 12:00pm PDT Register Today: https://ansys.me/3EoHJ9n

Did you hear? Ansys Speos GPU has been recognized among the best by the 2023 Laser Focus World Innovators Awards. An esteemed and experienced panel of judges from the optics and photonics community recognized Speos GPU as a Silver honoree! https://ansys.me/44HoQsI Ansys Speos creates accurate virtual prototypes for design and validation of illumination, imaging, and display systems, from automotive lighting to augmented reality devices. Speos is the premiere #optics simulation software for #Photometry/#Radiometry simulation in every domain where light is involved. It allows design of #illumination systems as well as virtual validation through any type of measurement, including radiance for proper human perception or camera sensor. Speos #GPU takes the capabilities from traditional Speos #simulation to the GPU hardware for an increase in speed that not only accelerates result delivery but revamps the overall design and validation experience. All achieved while maintaining the accuracy of the simulation. With overall increased simulation speed of x60, complex simulations that took hours or days now take minutes, shortening the time to market while allowing more innovation time. Included in the Ansys 2023 R2 Release are even more Speos features that enable designers to maximize their innovation thanks to Speos GPU. Here are some highlights. Interactive Live Preview Speos 2023 R2 introduces a brand-new explorative workflow using Live Preview. Any modification in the definition of Speos objects is seen almost instantaneously thanks to the update button powered by Speos GPU. Speos GPU - ray file support Simulations defined with ray file sources now run with Speos GPU direct simulation, providing the same result accuracy with GPU speed. Any illumination project can now benefit from GPU speed thanks to the support of ray file. 3D Irradiance sensor 3D irradiance sensor permits designers to analyze light contributions for each face of a geometry by providing a 3D illumination map. Each face becomes an irradiance sensor. Simulations defined with 3D Irradiance sensors can now run in Speos GPU direct simulation, providing the same result accuracy with GPU speed. Output faces of irradiance sensor When irradiance sensor is used in inverse simulation, rays are emitted in all directions, which may not be efficient in some cases, such as camera lens objective. Output faces permit to set the faces rays starting from the sensor are going to aim during the inverse propagation to improve performances. This functionality of the irradiance sensor is now compatible with Speos GPU inverse simulation, providing the same result accuracy with GPU speed. Join our upcoming webinar to learn more about the features in the Ansys Speos 2023 R2 Release! Ansys 2023 R2: Ansys Speos What’s New September 13, 2023 / 8:00am & 12:00pm PDT Register Today: https://ansys.me/44HoQJe

Ansys, a leading engineering simulation software company, has recently announced its collaboration with Nvidia Omniverse to revolutionize the visualization capabilities of its physics solver for 3D IC designers. This partnership aims to provide real-time and accurate visualizations of electromagnetic and thermal package effects for 3D IC designs in cutting-edge technologies such as 5G/6G, Internet of Things , artificial intelligence /machine learning , cloud computing, and autonomous vehicles. The shift towards 3D vertically stacked chips has significantly reduced the form factor of electronic devices. However, this design approach poses challenges related to electromagnetic interference and thermal stress effects. Understanding and tracing the origins of these complex issues have become more critical than ever. To address these challenges and enable more advanced applications, the integration of 3D multiphysics visualization has become a necessity for effective design and diagnostics. #electricalengineering #electronics #embedded #embeddedsystems #electrical #computerchips Follow us on LinkedIn to get daily news: HardwareBee - Electronic News and Vendor Directory

Ansys has taken a significant step in advancing #semiconductor system #design by integrating NVIDIA Omniverse APIs. This #collaboration empowers 3D-IC #designers with real-time visualization of results from Ansys solvers, including Ansys HFSS, Ansys Icepak, and Ansys RedHawk-SC. #electronicsnews #technologynews