Understanding the Role of CPUs, GPUs, and DPUs in Data Processing
The central processing unit (CPU) and graphics processing unit (GPU) each manage distinct types of data. The CPU is a fundamental component found in virtually all devices, including computers, smartphones, tablets, smartwatches, and televisions. Its primary function is to process general information, ensuring that software operates smoothly. In contrast, the GPU complements the CPU by handling more complex tasks such as supercomputing, artificial intelligence, machine learning, and large-scale data analysis, which require high numerical precision.
The COVID-19 pandemic triggered a global chip shortage, prompting chip manufacturers to explore alternatives to conventional chips to secure a stable supply of semiconductors. Concurrently, companies are seeking efficient methods to reduce costs and enhance productivity at the chip level. Enter the data processing unit (DPU), a recent innovation designed to work alongside CPUs and GPUs. The DPU offloads network communication tasks from the CPU or GPU, optimizing workloads in cloud and data centers while lowering costs by alleviating the burden on more expensive CPUs and GPUs.
One notable DPU developer, MangoBoost, recently raised $55 million in a Series A funding round for its DPU hardware and software solutions. These innovations assist enterprises and data centers in managing substantial data volumes to optimize performance. While the startup has not disclosed its latest valuation, insiders estimate it to be around $300 million, bringing its total funds raised to $65 million.
Based in Seattle and Seoul, MangoBoost asserts that its DPU solution enables data centers to decrease power consumption while enhancing performance cost-effectively and securely. The company claims its DPU can deliver up to three times the performance of existing technologies and reduce CPU usage by as much as 95% when used in conjunction with Samsung’s Petabyte SSD storage system.
The investment will primarily support the young startup in advancing the development and operation of its product line, which includes DPU hardware IP, DPU software, FPGA (Field Programmable Gate Array)-based DPUs, ASIC (Application-Specific Integrated Circuits)-based DPUs, AI-driven DPU solutions, custom DPU designs, and DPU-optimized systems. Additionally, MangoBoost plans to double its workforce, expanding from 58 employees—of which 71% are focused on research and development—by the end of next year.
Jangwoo Kim, CEO of MangoBoost, emphasizes that the company's efficiencies stem from over nine years of research and development conducted at the Seoul National University laboratory focused on DPU technology for data centers.
As competition in the DPU market intensifies, tech giants and semiconductor firms like Intel, Nvidia, AMD, Amazon, and Microsoft are investing in DPU vendors to enhance their data center optimization technologies. Various competitors also offer DPU-like technologies, such as Microsoft’s FPGA Smart NIC and Amazon Nitro.
Kim points out that Intel, Nvidia, and AMD typically provide their DPUs as PCIe cards equipped with proprietary DPU chips. He highlights that MangoBoost differentiates itself through an extensive array of customizable DPU features tailored to meet specific client needs, including AI server DPUs, big data server DPUs, and cloud server DPUs.
“Our mission is to deliver comprehensive DPU hardware and software solutions that cater to diverse customer requirements,” states Kim.
The startup recently participated in discussions with potential partners and presented its collaborative efforts with Samsung during the OCP Summit 2023, showcasing how MangoBoost’s customizable DPU technology can enhance Samsung’s Petabyte storage solutions.
South Korean venture capital firms IMM Investment and Shinhan Venture Investment co-led the Series A funding round, with participation from Korea Development Bank, KB Investment, Hong Kong-based IM Capital, and Premier Partners.
Ranking
