Phison Electronics Corp. · TPEX:8299
Phison Electronics — FMS 2023
18 slides from Phison Electronics's conference presentation, Aug 7, 2023
Slide 1: Welcome to FMS 2023 Media Presentation
the Possibilities Welcome to FMS 2023 Media Presentation Presented by Chris Ramseyer Director of Technical Marketing C O N F I D E N T I A L FMS 2023
Slide 2: Agenda
AGENDA Corporate Update What is Imagin+? Product Showcase X2 – Enterprise Gen5 E26 “Max14um” 14GB/s PS5027-E27T E18 pSLC C O N F I D E N T I A L FMS 2023
Slide 3: Corporate Update
The Possibilities 1 Corporate Update C O N F I D E N T I A L FMS 2023
Slide 4: Quick Facts About Phison
Quick Facts About Phison Experience 23+ $2B 600M Years 2022 Revenue Annual Controller Shipments 3800 +HC Business Categories Client | Embedded | Enterprise 16 EB Automotive | Gaming | Design Service Annual NAND Consumption 75% R&D Power IC | Signal IC Publicly listed 8299.TWO IT Press Tour 2023
Slide 5: Phison's World-Wide Influence
Phison’s World-Wide Influence Powered by Phison Phison 1 out of every 4 SSD 26% 2022 World-Wide shipped with Phison inside Client SSD Unit Shipments 244 MILLION TOTAL Other 74% C O N F I D E N T I A L FMS 2023
Slide 6: What is IMAGIN+
The Possibilities 2 What is IMAGIN+ C O N F I D E N T I A L FMS 2023
Slide 7: Design Services Empowering Innovation
Design Services Empowering Innovation Industry Collaborative Efficient Flow for Leading Technology Resources Sharing New Product Introduction (NPI) Enhanced Streamlined Accelerated Competitiveness Engineering Processes Time-to-market ✓ Industry leading NAND technology ✓ Best-in-class IP, ASIC, FW ✓
Slide 8: Process - Easy as 1,2,3
Process - Easy as 1,2,3 Complete NDA, discuss specs and objectives. Step 1 Brainstorming, specs Develop customer advantages with IMAGIN+ Phison provides cost estimates, program schedule, Step 2 Business plan review and payment milestones Document scope Step 3 Execute MOU or SOW, launch project! of w
Slide 9: Product Showcase FMS 2023
The Possibilities 3 Product Showcase FMS 2023 C O N F I D E N T I A L FMS 2023
Slide 10: PS5302-X2 Gen 5 Enterprise SSD
PS5302-X2 Gen 5 Enterprise SSD PS5302-X2 SSD Designed to be the best of PCIe Gen5 performance, features, endurance, and economics for enterprise applications. Phison is dedicated to developing advanced SSD technology to provide the industry with increased density, higher performance and power effici
Slide 11: PS5026-E26 Max14um Gen5 SSD
PS5026-E26 Max14um Gen5 SSD PS5026-E26 Max14um SSD E26 tuned to maximize the full bandwidth of PCIe Gen5. E26 was made for 2400MT/s NAND. The first consumer SSD to surpass 1,000 MB/s in UL PCMark 10 Storage, Quick, Data Tests and UL 3DMark Storage Test. On display with Frore Systems AirJet solid-sta
Slide 12: PS5027-E27T Gen4 DRAMless for Client
PS5027-E27T Gen4 DRAMless for Client PS5027-E27T SSD The E27T delivers on the promise of low-power PCIe Gen4 performance in mobile devices for gaming and general PC computing. Put the power of 7,000 MB/s in your notebook. On display enabling both Micron B58R 2400MT/s and Kioxia BiCS6 2400MT/s NAND i
Slide 13: PS5018-E18 pSLC High Endurance Champion
PS5018-E18 pSLC High Endurance Champion PS5018-E18 pSLC SSD Take your 100% write workload to another level with exceptional performance and up to 43 drive writes per day (DWPD). On display with the Apex Systems X21 add-in card that enables 21 M.2 2280 SSDs in a single PCIe Gen4 x16 slot delivering o
Slide 14: Rapid Growth of AI Model Parameters
Rapid Growth of AI Model Parameters • More computation and memory are required as model parameters increase significantly • GPU memory size cannot scale with AI model, especially for the Large Language Model (LLM) • High DRAM costs limit memory capacity expansion • The limited DRAM capacity becomes
Slide 15: Phison aiDAPTIV+ service solution expands NAND storage in AI market
Phison 'aiDAPTIV+' service solution expands NAND storage in the AI application market Current Phison’s aiDAPTIV+ AI Computing Architecture AI Computing Architecture Millions of GPU Cards Few GPU Cards + SSDs Large AI Model Large AI Model • Minimize the numbers of GPU cards and save cost significantl
Slide 16: The current market segmentation of AI computing architectures
The current market segmentation of AI computing architectures Large Cloud Service Providers • Super fast performance • Millions of users • Open for worldwide • More budget Phison Focus Small/Medium Organizations • Moderate performance • Hundreds of users • Private use • Customization required • Less
Slide 17: LLM Training Benchmark
LLM Training Benchmark Current Architecture Phison aiDAPTIV+ Entire LA LB LC ∙∙∙ LN L… L… ∙∙∙ LNx LLM Model SSD • 8 GPUs are required to run the specified • It is possible to train models with fewer GPU cards model size under the given testing conditions • Memory swap Optimizer withLA’Phison states
Slide 18: Thank You!
Thank You! C O N F I D E N T I A L FMS 2023