PDF 原檔：報告_OCP2026_緯穎_單相轉兩相系統挑戰_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）
| 檔名 | 頁 | 分類 | 親眼所見內容 |
|---|---|---|---|
報告_OCP2026_緯穎_單相轉兩相系統挑戰_20261015_p5.png |
p5 | 真資料圖 | 兩相冷媒比較表：R134a／R1234yf／R515B／R1233zd(E) 的沸點、45°C 飽和壓力、液汽密度、潛熱、ASHRAE 分級、GWP（如 R1233zd(E) 潛熱 180、251 kPa、GWP 1；R515B 146、871 kPa、GWP 299） |
報告_OCP2026_緯穎_單相轉兩相系統挑戰_20261015_p6.png |
p6 | 真資料圖 | R134a（中壓）vs R1233zd(E)（低壓）同冷板、0.7 LPM 實測：最大熱阻（R134a 約 0.024～0.028 vs R1233zd(E) 約 0.027～0.033 °C/W；左側並列密度、潛熱、熱導率、黏度參數表）與壓降（R1233zd(E) 在約 2,300W 達約 42 kPa，R134a 同功率約 17 kPa） |
報告_OCP2026_緯穎_單相轉兩相系統挑戰_20261015_p7.png |
p7 | 真資料圖 | 系統壓力效應：假設系統 ΔP=100 kPa，R1233zd 飽和溫度由 50 降至 37°C（設施水需 29°C）；R1234yf 由 50 降至 46°C（設施水 38°C） |
報告_OCP2026_緯穎_單相轉兩相系統挑戰_20261015_p9.png |
p9 | 真資料圖 | PG25 vs R515B／R1233zd 結溫比較：3kW TTV（60×60 mm²）圖，及緯穎 1kW（120 W/cm²）／2kW（240 W/cm²）矽 TTV 結溫範圍，PG25 皆較低 |
報告_OCP2026_緯穎_單相轉兩相系統挑戰_20261015_p10.png |
p10 | 真資料圖 | 兩相流動架構：串聯（vapor quality 0→0.1→0.5，ΔP 10→20→60 kPa）與並聯（1500W／500W 負載不均造成流量分配不均） |
報告_OCP2026_緯穎_單相轉兩相系統挑戰_20261015_p11.png |
p11 | 真資料圖 | 兩相冷卻迴路方塊：流量限制器／調節器→冷板→回流管→冷凝器→儲液槽→泵→供液管；列出流動不穩、蒸氣體積大、泵汽蝕三項問題與對 CDU 體積的影響 |
原始內容
p1
Jake Hsieh Assistant Technical Manager, Wiwynn Yuzhou Lin Thermal Engineer, Wiwynn System-level Challenges and Considerations in Transitioning from Single-Phase to Two-Phase Cooling COOLING: COLD PLATE
p2
- Working Fluid • Refrigerant trade-offs drive system design decisions
- Flow Instability • Boiling introduces flow excursion and oscillation risks
- Flow Architecture • Series and parallel architectures impact performance and flow distribution
- Cooling Loop • Vapor management impacts piping, pumps, and loop integration Transitioning to Two-phase Cooling Requires More Than a Component Swap
p3
• Potentially higher heat transfer and better temperature uniformity • Potentially lower flow rate requirement and pumping power • Lower risk of damage due to leakage • Lower corrosion and bio-growth concerns Why Two-phase Cooling? Ref: Incropera, F.P. et al., Fundamentals of Heat and Mass Transfer Water as an example. Single-phase: Cp × ΔT (Cp = 4.18 kJ/kg·K, ΔT = 10°C); Two-phase: hfg (latent heat of vaporization) Phase Heat Transfer Process Heat Transfer Coefficient (W/m2∙K) Heat Absorption Capacity** (kJ/kg) Single: Liquid Forced Convection 100-40,000 41.8 Two: Liquid to Vapor Convective Boiling 2,500-100,000 2,257
p4
• Ideal refrigerant: zero ODP, GWP < 1, non-PFAS, ASHRAE A1 classification • Compatibility between the working fluid and wetted materials must be independently verified • Lower liquid-to-vapor density ratio may enhance flow stability and performance Two-phase Working Fluid Considerations Suitable Working Fluid For Two-phase Cooling Properties Latent Heat Density Boiling Point Viscosity Compatibility Quick Connector Hose Cold Plate Pump Sustainability and Safety ODP / GWP Flammability PFAS Toxicity
p5
• Working fluids can be broadly categorized into medium-pressure refrigerants (R134a, R1234yf, R515B) and low-pressure refrigerants (R1233zd(E)) • R515B and R1233zd(E) are key candidates in current data center two-phase cooling development and applications Common Two-Phase Cooling Working Fluids Refrigerant Boiling Point @ 1atm (℃) Saturated Pressure @ 45℃ (kPa) Liquid/Vapor Density @ 45℃ (kg/m3) Latent heat @ 45℃ (kJ/kg) ASHRAE Classification IPCC AR5 GWP R134a -26 1155 1127/57.4 158 A1 1300 R1234yf -29 1150 1013/66.2 128 A2L <1 R515B -19 871 1109/47.9 146 A1 299 R1233zd(E) 18 251 1213/13.5 180 A1 1
p6
• R134a showed better thermal performance and lower pressure drop than R1233zd(E) under the same primary-side conditions and with the same cold plate Medium vs. Low-Pressure Working Fluid Experimental Comparison Parameter R-134a (Medium-Pressure) R-1233zd(E) (Low-Pressure) Liquid/Vapor Density (kg/m³) 1127 / 57.4 1213 / 13.5 Latent Heat (kJ/kg) 158 180 Liquid Thermal Conductivity (mW/m·K) 72.6 76.9 Vapor Dynamic Viscosity (μPa·s) 12.6 11.0 0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 0.020 0.022 0.024 0.026 0.028 0.030 0.032 0.034 0.036 0 500 1000 1500 2000 2500 3000 Pressure drop (kPa) Max. Thermal Resistance (℃/W) Total power (W) R134a Rth R1233zd(E) Rth R134a ΔP R1233zd(E) ΔP Ref: AVC Experimental Results • Both at 0.7LPM
p7
20 30 40 50 60 100 200 300 Sat. temperature (℃) Pressure (kPa) R1233zd Assumed system ΔP=100kPa Sat. T drops due to ΔP 50→37℃ Facility water temp. 29℃ 20 30 40 50 60 800 900 1000 1100 1200 1300 1400 Sat. temperature (℃) Pressure (kPa) R1234yf Sat. T drops due to ΔP 50→46℃ Facility water temp. 38℃ Assumed system ΔP=100kPa Low-pressure Refrigerant • More sensitive to pressure drop, requiring stricter facility water conditions • Lower component pressure rating required Technical Consideration 1: System Pressure Effect Medium-pressure Refrigerant • More tolerant of pressure drop, enabling easier facility water integration • Higher component pressure rating required
p8
• Two-phase flow instability may occur across multiple scales, from cold plate channels to trays and racks • Potential mitigation approaches include inlet flow restrictors, expanded outlet channels/piping, and optimized manifold designs Technical Consideration 2: Flow Instability Ref: 1. L. E. O'Neill and I. Mudawar, Review of two-phase flow instabilities in macro- and micro-channel systems, International Journal of Heat and Mass Transfer 157 (2020) 119738 2. AVC Experimental Study Ledinegg Instability Negative P-Q curve → Flow excursion Elongated Bubble Oscillation Confined bubble growth → Flow Oscillation
p9
• PG25 shows better thermal performance than R515B and R1233zd(E) on both platforms, with the same PG25-designed cold plate and primary-side conditions for each comparison Further Cold Plate Optimization May Improve Two-Phase Performance 64 66 68 70 72 74 76 78 1.0 2.0 3.0 4.0 5.0 Tjunction (℃) Flow Rate (LPM) R515B PG25 Ref: AVC Experimental Results • 3kW TTV (60 × 60mm2) • 25°C PG25 inlet temperature Ref: Wiwynn Experimental Results • Silicon TTV (26 × 32mm²), 9 Tjunction measurement points recorded • 30°C PG25 inlet temperature, 1.5 LPM/kW Min Ave Max Min Ave Max Min Ave Max Min Ave Max 50 60 70 80 90 100 Tjunction (℃) R1233zd PG25 1kW (120W/cm2) 2kW (240W/cm2)
p10
In Series High pressure drop & Uneven cold plate performance ΔP=10kPa ΔP=60kPa 𝜒= 0.1 χ = 0.5 ΔP=20kPa 𝜒= 0 𝜒=vapor quality In Parallel Flow maldistribution under non-uniform heat loads 1500W 500W higher 𝜒 higher ΔP lower 𝜒 lower ΔP Higher Flowrate Lower Flowrate Technical Consideration 3: Flow Architecture (Flow Boiling) Heat (W) ΔP (psi)
p11
Technical Consideration 4: Two-phase Cooling Loop 1. Flow instability and non-uniform distribution Flow restrictors/regulators may be required → Higher pump head requirement 2. High vapor volume Larger return piping to reduce flow velocity & pressure drop → Larger manifold size/CDU footprint 3. Pump cavitation A low-NPSHr pump and large receiver may be required → Larger CDU footprint Pump Supply Piping Flow Restrictor/ Regulator Cold Plate Return Piping Receiver (Buffer Tank) Condenser *NPSHr: Net Positive Suction Head Required
p12
Problem to solve • Two-Phase optimized cold plate development • System-level experiments to identify and mitigate flow instabilities in real applications • Standardization and ecosystem interoperability Get involved in Cooling Environments – Cold Plate Sub-project • Every second Wednesday of the month from 8 to 9AM PT • https://www.opencompute.org/w/index.php?title=Cooling_Environments/Cold_Plate Call to Action
p13
Thank You! Jake Hsieh Jake_Hsieh@wiwynn.com https://www.wiwynn.com Webb Chen webb_chen@avc.co https://www.avc.co/en-us Yuzhou Lin Yuzhou_Lin@wiwynn.com https://www.wiwynn.com Special Thanks to Webb Chen from AVC for providing the test data for this work Wiwynn Whitepaper