Maxim Integrated Products, Inc. claims designers of high-performance, high-power artificial intelligence (AI) systems can now achieve highest efficiency (to reduce power cost and heat) and smallest total solution size with the MAX16602 AI cores dual-output voltage regulator and the MAX20790 smart power-stage IC.
Leveraging the current ripple cancelation benefit from Maxim Integrated’s patented coupled inductor, this AI multi-phase chipset provides a 1% efficiency improvement compared to competitive solutions, enabling greater than 95% efficiency at 1.8V output voltage and 200A load conditions.
Furthermore, this increase in efficiency translates to a 16% reduction in wasted power. It also allows 40% less output capacitance compared to competitive solutions, reducing total solution size and capacitor count. The chipset provides a scalable solution for various output current requirements and is customisable to support multiple form factors. In addition, the chipset enables AI computing at the edge as well as cloud computing at the datacentre.
Hyperscale datacentre designers grapple with increasing peak power and related higher thermal levels as they boost computing power to take on the rapidly rising demands of AI applications and deep learning. AI systems implemented with the MAX16602 and MAX20790 multi-phase chipset generate less heat compared to competitive solutions.
Due to Maxim Integrated’s patented coupled inductor technology and monolithic integrated dual-side cooling power stage ICs, power loss is reduced as a result of a 50% lower switching frequency. The monolithic integrated approach practically eliminates the parasitic resistance and inductance between FETs and drivers to achieve the high efficiency.
Space constraints also challenge designers as they seek to boost AI functionality. This chipset enables the smallest total solution size and allows developers to reduce component count and bill-of-materials (BOM) costs. Furthermore, Maxim Integrated’s low-profile coupled inductor technology supports higher saturation current per phase compared to a discrete inductor offered by competitive solutions. This allows designers to overcome space limitations by having lower phase count than competitors while also lowering the total cost of ownership.
Advantages
- High Efficiency/Low Heat and Power Dissipation: Maxim Integrated’s patented coupled inductor technology reduces switching frequency by 50%, allowing for 1% higher efficiency
- Small Total Solution Size: A decrease in solution size is achieved by reducing output capacitance by 40% and providing a solution with lower phase count compared to competitive solutions
- Flexible: Solution is scalable from 2 to 16 phases for different output current requirements (thermal design current is typically 60A to 800A or more); Low profile (<4mm) coupled inductor is customisable to support multiple form factors such as peripheral component interconnect express (PCIe) and OCP accelerator modules (OAM)
“The real estate for AI power solutions is becoming limited while the need to provide high power density in a pre-designed small space is more critical. As a result, designers need a scalable solution to customise their design objectives,” says Steven Chen, director of business management for the cloud and data Centre business unit at Maxim Integrated. “This multi-phase AI power chipset by Maxim Integrated powers AI hardware accelerators such as GPUs, FPGAs, ASICs and xPUs to increase solution efficiency and reduce solution size for different form factors such as PCIe and OAM.”
Availability and pricing
- The MAX16602 is available at Maxim Integrated’s website, as well as authorised distributors; Pricing available upon request.
- The MAX20790 is available at Maxim Integrated’s website, as well as authorized distributors; Pricing available upon request.
- The MAX16602CL8EVKIT# evaluation kit for both products is available.
- Samples for MAX16602 and MAX20790 are available at Maxim Integrated’s website.
- EE-Sim models are now available. For details, click here.
Comment on this article below or via Twitter: @IoTNow_OR @jcIoTnow
