+44 (0)24 7671 8970
More publications     •     Advertise with us     •     Contact us
 
Loading...
News Article

Expanding NAND storage for AI applications

News

Phison Electronics has expanded its IMAGIN+ service to include AI computational models and AI services solutions.

Phison’s aiDAPTIV+ utilizes Phison's innovative integration of SSDs into the AI computing framework and expands NAND storage solutions in the AI application market.

“ADLINK and Phison have had many years of cooperation and synergy in the industrial storage application field, and recently, we have joined forces to expand into the AI application market”

Phison has been deeply engaged in NAND controller innovations for more than 23 years. In recent years, the company has developed and integrated AI and machine learning technologies into NAND controllers and algorithms in order to enhance the computational performance and reliability of the company’s NAND storage solutions.

Additionally, Phison has established a data science team, developed and deployed relevant AI technologies, and created a world-class validation laboratory. These efforts have accelerated the adoption of Phison's storage solutions in various applications, including PCs, mobile phones, automobiles, gaming platforms, enterprise data center & edge applications, supply chain logistics, and more.

With the recent rise of large-scale generative AI models such as ChatGPT, the possibilities for artificial intelligence to assist both businesses and individuals in the future are significant. The rapid growth of AI models has also led to a significant increase in the hardware infrastructure cost associated with providing AI services.

The primary reason for increased costs is that current AI models run primarily on GPUs and DRAM. However, the growth rate of future AI models will far exceed the capacity that GPUs and DRAM can provide. According to a recent research report by Microsoft, the growth rate of AI models will soon be 200 times that of DRAM in GPU cards. In other words, the current hardware architecture for AI computing may already be insufficient to meet the future demands of AI applications.

To improve overall operational performance and efficiencies for AI hardware architecture, aiDAPTIV+ integrates SSDs into the AI computing framework. Based on the application timeline, aiDAPTIV+ structurally divides large-scale AI models and collaboratively runs the model parameters with SSD offload support. This approach maximizes the executable AI models within the limited GPU and DRAM resources, effectively reducing the hardware infrastructure cost required to provide AI services.

The initial application scenario for this technology is the world's first mass-produced Artificial Intelligence-Automated Optical Inspection (aiDAPTIV+ AOI) system integrated with SSD cooperative computing. Phison’s aiDAPTIV+ AOI services accelerate the transition of Surface Mount Technology (SMT) factories into Industry 4.0 and improve detection accuracy while eliminating delays and costs associated with instability caused by human inspection.

Phison’s aiDAPTIV+ AOI service is applicable to various SMT product lines, including SSDs, graphics cards, DRAM, motherboards, and more. Approximately 10 SMT partners and customers have already successfully adopted the Phison aiDAPTIV+ AOI service, improving the accuracy of their existing AOI systems and reducing the escape rate caused by human errors to below 10 defective parts per million (DPPM), significantly increasing throughput rates to more than 99%.

Initial users are enthusiastic about the results they are seeing upon deployment.

“ADLINK and Phison have had many years of cooperation and synergy in the industrial storage application field, and recently, we have joined forces to expand into the AI application market,” said Ethan Chen, General Manager of Edge Computing Platforms BU at ADLINK Technology. “Phison has integrated AI mechanisms into the design of NAND controllers, enhancing product performance and reliability. Coupled with ADLINK's next-generation NVIDIA Jetson Orin platform, ADLINK and Phison will be able to provide customers in the industrial and AI fields with more complete and reliable solutions. Currently, ADLINK's edge AI inference platforms, DLAP-211-Orin Nano and DLAP-211-Orin NX, have become the first certified edge computing devices for Phison's self-developed aiDAPTIV+ AOI automatic optical inspection system, and we will continue to expand our collaboration in the future.”

“Phison’s aiDAPTIV+ architecture integrates SSD technology in an innovative approach. This new structure can significantly reduce the cost of hardware deployment in specific AI applications. The aiDAPTIV+ AOI platform launched by Phison is a concrete result of this new architecture,” said Benny Lan, deputy general manager of Giga Computing. “Giga Computing is pleased to be the first partner with product certification for the aiDAPTIV+ AOI platform, which has been successfully deployed in many customers’ SMT production lines using GIGABYTE’s customized workstations and AMD GPUs, which were paired with aiDAPTIV+ computing architecture. This system improved the efficiency in detection of defects and greatly reduced possible human errors. Giga Computing and Phison will continue to expand cooperation in the field of AI applications.”

Dr. Wei Lin, Head of Phison's AI R&D team and Assistant Professor at the College of Artificial Intelligence, National Yang Ming Chiao Tung University, said Phison's core expertise and competitive advantage is in the development of NAND controllers. Expanding the connection between NAND storage and AI applications has been Phison's focus in recent years. Phison's AI R&D team has successfully integrated SSDs into a newly developed AI computing framework, which not only effectively reduces hardware infrastructure costs for AI servers but also enables the application of this aiDAPTIV+ computing architecture in various AI application scenarios.

The globally pioneering SSD-Boosted aiDAPTIV+ AOI optical inspection system announced today demonstrates Phison's successful integration of SSDs and the creation of a new AI computing framework. Moving forward, Phison will continue to expand the development of aiDAPTIV+ technology and collaborate with partners from various fields to further foster the application of NAND storage in the AI domain. Through the Phison aiDAPTIV+ AOI, Phison aims to assist SMT partners and customers in improving production efficiency and reducing costs.

Phison will invite global ecosystem partners and customers to visit their cutting-edge NAND controllers at the Flash Memory Summit in the United States, which will take place from August 8th to August 10th in Pacific Time. The showcased solutions will include the world's first PCIe 4.0 DRAM-Less client SSD controller PS5027-E27T that supports speeds of up to 3600MT/s for 3D NAND, the world's first PCIe 5.0 DRAM-Less client SSD controller PS5031-E31T, as well as the live demo of Phison's AI service solution, aiDAPTIVE+ AOI. Phison will engage in discussions with global partners and customers to further expand the market for NAND storage applications.

Reality AI Explorer Tier offers free AI/ML development access
AEM introduces new generation of Automated Burn-In Systems
NPUs are emerging as the main rival to Nvidia’s AI dominance, says DAI Magister
Camtek receives a $20M order from a Tier-1 OSAT
Free samples of every STMicroelectronics NPI IC available from Anglia
NY CREATES and SEMI sign MoU
Major government investment to 'propel' Canada
QuickLogic announces $5.26 million contract award
ASNA and Athinia collaborate
Global sales forecast to reach record $109 billion in 2024
AMD to acquire Silo AI
Nanotronics unveils 'groundbreaking' Gen V AI Model
Aitomatic unveils SemiKong
Biden-Harris Administration reveals first CHIPS for America R&D facilities and selection processes
Collaboration to produce cutting-edge AI accelerator chips
Oxford Ionics breaks global quantum performance records
Adeia wins ECTC Award for paper on “Fine Pitch Die-to-Wafer Hybrid Bonding”
AEM introduces new generation of Automated Burn-In Systems
Sydney council forges vital semiconductor agreement
EV Group's EVG880 LayerRelease wins 2024 Best of West
Biden-Harris Administration to invest up to $1.6 billion
Cyient sets up subsidiary for semiconductor business
EMD Electronics network leaders honoured
Accenture acquires Cientra
Introducing Park FX200
Deep learning chipset market to surge to $72.8 billion by 2033
PI Innovation Award 2024: Nanobodies for Diagnostic and Therapeutic Applications
Graphcore joins Softbank Group
Kaman Measuring highlights high precision displacement sensors
Si2 reveals recipients of Annual Power of Partnerships Award
Advantest achieves ASPICE Level 2 Certification for V93000 SmarTest 8 Software
Semiconductors to become a trillion-dollar industry by 2030
×
Search the news archive

To close this popup you can press escape or click the close icon.
Logo
×
  • 1st January 1970
  • 1st January 1970
  • 1st January 1970
  • 1st January 1970
  • 1st January 1970
  • 1st January 1970
  • 1st January 1970
  • 1st January 1970
  • View all news 22645 more articles
Logo
×
Register - Step 1

You may choose to subscribe to the Silicon Semiconductor Magazine, the Silicon Semiconductor Newsletter, or both. You may also request additional information if required, before submitting your application.


Please subscribe me to:

 

You chose the industry type of "Other"

Please enter the industry that you work in:
Please enter the industry that you work in: