Tech Field Day was a great success for the Phison team, who shared the company’s latest AI and data storage innovations through a series of talks to a panel of invited industry guests.
Now you can catch up on everything Phison shared with on-demand presentation videos from Tech Field Day. Dive into the innovations behind Pascari 企业 SSD 和 爱DAPTIV+, and see how we’re shaping the future of storage and AI.
Affordable on-premises LLM training and inference with Phison aiDAPTIV+
Brian Cox, Phison Director of Solution and Product Marketing, explained how Phison aiDAPTIV+ 技术 can significantly reduce the cost of training AI models on-premises. The solution works by using flash memory as a cache to offload LLMs from the GPU cards onto much higher-capacity SSDs. This enables organizations to handle larger miodels and process and train AI with fewer expensive GPU cards. It also makes on-premises AI projects more accessible to organizations that couldn’t afford the million-dollar setups of larger companies.
“This has completely changed the economics of GPU processing. Now you can load the largest models and scale that easily as the industry continues to move forward. You are able to do that and afford it with flash. So, it’s no longer just a domain of Fortune 500 companies, but small, medium businesses, state and local government, universities can now afford to have their own AI infrastructure to run their own models, to apply their research. It democratizes AI for all of society.”
Watch the Presentation
GPU memory offload for LLM fine-tuning and inference with Phison aiDAPTIV+
Phison CTO Sebastien Jean dove deeper into aiDAPTIV+ 技术 and the enormous memory capacity that fine-tuning requires—up to 20 more than the memory you need to run a large language model. He also explained the memory management middleware and all-in-one AI software toolset components of the aiDAPTIV+ solution and how they work together to give an SSD the ability to act as an extended memory pool.
“We’re already seeing [LLM training] cases where you have 605 billion parameters, 671, and 2 trillion is on the horizon. And we’re seeing trends where even at the high end, SSDs are becoming part of this flow. What we’re doing with our solution is making sure that we can continue to provide support for these larger models, so we’re actually making higher and higher density SSDs.”
Watch the Presentation
Innovation in the enterprise SSD market driven by Phison
Michael Wu, Phison General Manager and President (US), demonstrated how Phison has positioned itself to meet the enterprise market’s always-evolving needs. Today the company is at the vanguard of enterprise data storage with its Pascari Enterprise SSD portfolio. Wu says Phison’s commitment to R&D—which makes up 75% of its global workforce—as well as vertical integration of controller, firmware, and hardware design, keeps it at the forefront of data storage innovation.
“Every single part of the journey of every product segment, because of our dedicated engineering, we always shoot for the best path to market. We always have been the leader of every single segment we move into … We announced [the Pascari] brand last year, and then very quickly, within a year, we announced the largest enterprise drive. We’re ahead of many players right now.”
Watch the Presentation
The Pascari Enterprise SSD portfolio by Phison
Chris Ramseyer, Phison Director of Technical Marketing, showcased Phison’s Pascari Enterprise SSD portfolio. He focused primarily on the X-Series, which is designed with ultra-high performance in mind, and explained how Pascari is an example of innovation in the enterprise SSD market due to the products’ performance and reliability. Also discussed were the various form factors and benefits of the SSDs.
“The [Pascari] X200 is the fastest enterprise SSD available today. We’ve got independent testing that’s backing that up. We outrun everyone else. … [Also, the] new Pascari D205V is essentially a 128- terabyte drive. It’s a 400% increase from before. Rip out your 30 TB drive, put in the new 122 TB and you immediately get a 2.2 petabyte increase at the server. At the rack level, it’s almost 60. It’s going so fast.”
Watch the Presentation
简体中文 
English 
繁體中文 
日本語 
Deutsch 
한국어