Virtual assistants like Siri and Alexa, facial recognition and self-driving cars heavily rely on deep learning and natural language processing. Deep learning uses more neural network layers to simulate the human brain’s ability to learn and can train computers to accomplish specific tasks by processing large amounts of data and recognizing patterns in the data.

A chip designed specifically for deep learning offers significant advantages over a network of computers – more cores, optimized for AI, huge on-chip memory giving very low latency for lightning-fast processing. This dissipates a lot of heat that must be removed to ensure overall performance and avoid premature failures.

Ensure overall performance and avoid premature failures for deep learning AI hardware with proper thermal management

Watch this webinar to gain insights into AI hardware thermal management for a deep learning machine. Designers can evaluate options such as air cooling using quick simulations. With simulation, designers avoided wasting time on a cooling solution that wasn’t feasible and moved on to characterizing a cooling system that ensured the overall performance of the largest chip ever built.

Cooling optimization for natural language processing AI hardware that dissipates a lot of heat

Systems developed for deep learning dissipate a lot of heat, and designers must prioritize cooling optimization to build successful AI hardware. Simulation is a powerful tool in the designer’s toolkit that can help optimize heat removal. Simcenter Flotherm XT CFD simulations helped optimize the coolant channels in the cold plate manifold.

How the thermal management experts from Electronics Cooling Solutions addresses cooling optimization

See how experts in thermal management at Electronic Cooling Solutions Inc collaborated with Cerebras engineers to develop the cooling solution for the biggest chip ever built. Some of the topics covered include:

• Why liquid cooling was chosen instead of air cooling for this application
• How the cold plate was designed using Simcenter Flotherm XT
• How the overall design ensures each IC operates at close to the same target temperature, which is critical for optimal performance
• The microchannel design that ensures each IC on the wafer gets the correct coolant flow
• The use of a simulation and test strategy to underpin the fidelity of the simulations