“Creative, magical, beautiful, inspirational; these are words that we use to describe the core of who we are as humans, ” said Praveen Vaidyanathan, Vice President and General Manager of Compute Products Group, at Micron Technology. “There are words we just used to describe technology. These are words I will use today to connect humans and technology.”
Vaidyanathan took the stage as the first presenter of the forum sessions on the second afternoon of COMPUTEX, which focused on memory and storage for the age of Artificial Intelligence. According to Vaidyanathan, current obstacles in AI tech related to memory and storage are bandwidth, capacity and power.
For Micron, Vaidyanathan’s “AI Unleashed: Powering the Future with Micron Memory and Storage” forum talked about the company’sHBM3E (High Bandwidth Memory 3E) solution which was engineered to meet the growing demand for data storage by generative AI models, boasting industry-leading performance and energy efficiency to empower AI.
However the latest product that Vaidyanathan introduced was the Micron GDDR7, a graphic card that delivers an experience with over 1.5 TB/s of system bandwidth.
Taking the stage second was Eddie Ramirez, the VP of Marketing Infrastructure Line of Business at Arm who assists partners develop using solutions from his company into the cloud, networking and edge markets.
Ramirez’s “Building Sustainable, AI-Ready Cloud Data Centers with Arm Neoverse” forum suggests that AI is everywhere, because the gains AI brings to productivity is palpable.
The Arm executive then presented Arm KlediAI, which is the company’s solution to handle the myriad of device types, neural networks and inference engines. KlediAI is a broad program of software that accelerates AI alongside software community engagements.
The storage that is Arm Kleidi Libraries allows developers to transparently access the AI powers of the pervasive Arm CPU which also operates the bulk of the world AI inference workloads from to cloud to edge.
Seagate’s Global VP of Brand Strategy Rosalina Hiu’s “AI Storage: Where the Future is Read and Written” forum shared with the audience how AI storage, specifically hard drives, shape the future of AI.
Because the TCO of acquisition of hard drives against SSDs used for storage including AI is six to one, 90% of data of companies are actually stored in hard drives rather than SSDs, according to Liu, which makes possessing hard drives and acquiring them at large spaces and low cost has never been more vital.
Seagate’s solution to the demand of hard drives is the Mozaic 3+ showcased by Hiu during her forum. The tech incorporates the company’s implementation of HAMR to deliver mass-capacity storage at areal densities of 3 TB per disk and beyond, which enables it to power the company’s flagship Exos 30 TB+ series that have the capacity points of 30 TB and beyond.
“Maximizing AI Workloads with Cloud Native Processors”forum presented by Ampere’s Chief Product Officer Jeff Wittich, who said power consumption remains the most critical constraint to continue the expansion of compute. Moreover, emerging workloads like generative AI only bring to surface the fact that the industry is in need of solutions that meet the demands in a sustainable manner, which was why Wittich heavily promoted his company’s Ampere One CPUs.
Ampere One is described by the company as its newest and most efficient Cloud Native Processors with up to 192 high performance cores that makes the tech the perfect choice for compute environments requiring the highest level of consistent and predictable performance at scale.
The final COMPUTEX forum of June 4 was delivered by Phison CEO K.S. Pua, who spoke on “The Challenges and Opportunities in the Implementation of Generative AI Applications.”
Pua revealed that while the world has been fascinated by generative AI, the service might not be as profitable or sustainable in the long time.
The answer to the concern, Pua said, lies in the widespread adoption of pre-trained generative AI models for the sake of affordability.
The Phison CEO introduced the aiDAPTIV+, a hybrid software / hardware AI fine-tuning end-to-end business solution for organizations to train large data models without additional staff and infrastructure. The platform is also able to scale linearly with data training and time requirements to consolidate efficiency to display results.
With the plethora of solutions aimed at resolving the issues that are stalling the progress of the widespread implementation of AI, the forums provided its audience with the insight and opportunity to fully capitalize on the potential of generative AI and advance their businesses.
