報告_OCP2026_Samsung_SSD兩相浸沒相容性_20261015

PDF 原檔：報告_OCP2026_Samsung_SSD兩相浸沒相容性_20261015_original.pdf

講者公開簡報（OCP 官方 Recorded Content 頁提供下載）。文字以 PyMuPDF 逐頁抽取，### p<n> 為頁碼；圖表內容見原檔。

圖片清單（已驗證 2026-10-19）

原始內容

p1

Youngnam Kim Samsung Electronics Material Compatibility and Long-Term Hardware Integrity of Enterprise SSDs in Two-Phase Immersion Cooling Cooling Environments Immersion Project

p2

2Phase Immersion Ecosystem Chemours Coolant Opteon 2P50: Next-gen 2-Phase dielectric coolant Liquid Stack Infrastructure 2Phase Immersion tanks for production DCs Samsung Device High-Performance Enterprise SSD Development PKI Operation Immersion Cooling Infrastructure Operations Partners Roles

p3

Long-term Reliability Test Environment Item Spec. Status Samsung High-performance Gen4 & Gen5 Enterprise SSDs Active Liquid Stack 48U 2Phase Immersion Cooling Tank Active ＞15,000 Hours (Ongoing) Stable Chemours Specialized 2Phase Dielectric Coolant Opteon 2P50 Stable 0% Failure Rate (Under JESD219) Excellent Operational Status Summary 48U 2Phase-Immersion Cooling Tank

p4

Mechanical Reliability Paradigm Shift SSD Operating Temperature (To) : Air vs Immersion Key Observations - Under JESD219 (SSD Endurance Workloads) 2-Phase immersion cooling dramatically reduces ΔT vs air cooling. - Spatial/temporal uniformity across SSD components is significantly improved. ※ This work focuses on thermo-mechanical behavior; signal integrity is out of scope. Implication Lower ΔT reduces mechanical strain, setting the stage for chemical compatibility to become the new reliability bottleneck in Immersion Cooling. 20 30 40 50 60 70 80 90 0 15 30 45 60 90 120 150 180 210 240 To (°C) Time (min) Air-cooled Two-phase immersion New bottleneck: chemical compatibility ΔTa ΔTi Reduced ΔT

p5

Material Compatibility Test Methodology Details OCP Guide For 2PIC Samsung Spec For 2PIC Target Material Material + Product Test Method Soxhlet 1) Soxhlet 2) Soak Test Temp Boiling Point (30~60℃) 1) Boiling Point - Soxhlet 2) 80℃- Soak Test Duration 48hrs 1) 72hrs Soxhlet 2) 336hrs Soak Criteria (Acceptable) △Weight < 2% △Extracts < 2% △Weight < 2% △Extracts < 2% Operating Test Pass Methodology Material Compatibility Samsung Spec Sampling Pre Test Soak & Soxhlet Test Soak - Cleaning Soxhlet - Drying Evaluation

p6

Material Compatibility Test Result - Materials : From EP SSDs, -Temperature : 49℃(B/P) - Coolant : Chemours Opteon 2P50, -Duration : 72hrs Materials Appearance Mass Change 72hrs Result Capacitors No Change △Mass < 2% PASS Labels No Change △Mass < 2% PASS EMCs No Change △Mass < 2% PASS PCB No Change △Mass < 2% PASS TIM No Change △Mass < 2% PASS Soxhlet Test Result Mass Change after Soxhlet Test(%) -2.5% -2.0% -1.5% -1.0% -0.5% 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% EMC-1 EMC-2 PCB Cap-1 Cap-2 Cap-3 Cap-4 TIM Label-1Label-2Label-3 Soxhlet Test Me(%) Ma(%) USL LSL

p7

Material Compatibility Test Result Materials Appearance Mass Change 2weeks 4weeks Result Capacitors No Change △Mass < 2% △Mass < 2% PASS Labels No Change △Mass < 2% △Mass < 2% PASS EMCs No Change △Mass < 2% △Mass < 2% PASS Underfills No Change △Mass < 2% △Mass < 2% PASS PCBs No Change △Mass < 2% △Mass < 2% PASS S/Pastes No Change △Mass < 2% △Mass < 2% PASS TIMs No Change △Mass < 2% △Mass < 2% PASS *Products No Change △Mass < 2% △Mass < 2% PASS (Working) - Materials : From Samsung EP SSDs, - Temperature : 80℃ - Coolant : Chemours Opteon 2P50, - Duration : 2wk, 4wk Soak Test Result Mass Change after Soak Test(%) -2.5% -2.0% -1.5% -1.0% -0.5% 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% Soak Test 2weeks 4weeks USL LSL

p8

■ Coolant Contamination Risk - Burn-out failure observed on high-power IC and fine-pitch BGAs. Contamination-Induced Failure PCBA Cleaning Process Observed SSD-Level Risks Solder joint surface contamination occured PCB Coolant Evaporation Contaminated Coolant Supply Short Dendrite growth 1) Contamination - Flux residues and PCBA-origin ions migrate into the coolant. 2) Deposition - Contaminants condense on high-power PKGs at boiling sites. 3) Failure - Dendrite growth bridges adjacent balls → short-circuit → burn-out. Dendrite Growth ■ Failure Mechanism PCB PCB ■ Removes SMT-origin ionic residues before SSD enters the immersion cooling tank ■Aqueous wash → DI rinse → controlled dry (verified per IPC-TM-650) ■Coolant purity preserved : 0% failure across 15,000+

p9

■ Coolant Contamination Risk - Burn-out failure observed on high-power IC and fine-pitch BGAs. Contamination-Induced Failure PCBA Cleaning Process Observed SSD-Level Risks Solder joint surface contamination occured PCB Coolant Evaporation Contaminated Coolant Supply Short Dendrite growth 1) Contamination - Flux residues and PCBA-origin ions migrate into the coolant. 2) Deposition - Contaminants condense on high-power PKGs at boiling sites. 3) Failure - Dendrite growth bridges adjacent balls → short-circuit → burn-out. Dendrite Growth ■ Failure Mechanism PCB PCB ■ Removes SMT-origin ionic residues before SSD enters the immersion cooling tank ■Aqueous wash → DI rinse → controlled dry (verified per IPC-TM-650) ■Coolant purity preserved : 0% failure across 15,000+

p10

■ One. Immersion cooling dramatically reduces thermo-mechanical stress. ■ Two. Samsung SSD materials are compatible with Chemours Opteon 2P50. ■ Three. The remaining risk is cleanliness, and we control it before the product goes int the tank To Summarize Call to Action Immersion Ecosystem partners ■ One. Let’s extend the OCP material test guide to the product level. ■ Two. Let’s add a PCBA cleaning specification to the guide. ■ Three. Please join the OCP Immersion project and work with us.

p11

■ One. Immersion cooling dramatically reduces thermo-mechanical stress. ■ Two. Samsung SSD materials are compatible with Chemours Opteon 2P50. ■ Three. The remaining risk is cleanliness, and we control it before the product goes int the tank To Summarize Call to Action ■ One. Let’s extend the OCP material test guide to the product level. ■ Two. Let’s add a PCBA cleaning specification to the guide. ■ Three. Please join the OCP Immersion project and work with us. Immersion Ecosystem partners

p12

Thank You! For questions or further discussion yn2.kim@Samsung.com