Conduction-Cooled vs. Liquid Cooling for High-Efficiency AC/DC Converters: Lessons from the Field

The Cooling Problem I Didn't See Coming

Back in 2019—or maybe early 2020, I'd have to check my notes—I spec'd a 19 inch rack power supply for a home electricity storage demo system. High efficiency AC DC converter, rated at 95%. Looked great on paper. What I didn't account for? Heat management. Specifically, whether a conduction-cooled plate would handle the thermal load or if we needed something more aggressive.

That mistake cost us about $4,200 in rework and delayed the project by three weeks. The vendor's standard PSU liquid cooling option was $300 more. I skipped it to save budget. Big mistake. Simple.

I'm not a thermal engineer, so I can't speak to the fluid dynamics of coolant flow or fin optimization. What I can tell you, from a system integration perspective, is how to evaluate these two approaches before you commit. And I've documented 11 similar errors over the past five years, totaling roughly $28,000 in wasted spend. This checklist exists because I earned it.

What We're Comparing: Conduction-Cooled Plate vs. PSU Liquid Cooling

Both of these technologies are used to manage heat in high-efficiency AC/DC converters and power storage devices. But they approach the problem differently.

Here's the core framework:

• Conduction-cooled plate: Heat transfers through a metal plate (usually aluminum or copper) that contacts the heat-generating components. Passive, no moving parts, no fluid.
• PSU liquid cooling: A closed-loop system with coolant, a pump, and a radiator. Active, more complex, but typically handles higher thermal loads.

The question most people ask: Which one is better? The real question: Which one is better for your specific power storage device and deployment scenario?

Dimension 1: Thermal Performance Under Load

This is where I made my first mistake. I assumed a conduction-cooled plate would handle a 3kW continuous load on a 19-inch rack PSU. It didn't. After two hours of runtime at 85% load, the baseplate temperature hit 78°C. The converter derated to 70% capacity. The home electricity storage system couldn't meet its peak demand spec.

Liquid cooling, on the other hand, kept the same unit at 52°C under identical conditions. The difference wasn't marginal—it was functional. If your AC/DC converter runs above 70% load for more than 30 minutes, conduction cooling probably isn't enough.

But here's the counterintuitive part: for intermittent loads—like a power storage device that only cycles during grid outages—a well-designed conduction-cooled plate is often sufficient. I've got a system running since 2021 that proves it. No pump, no fluid, no maintenance. Reliability wins.

Dimension 2: Space and Form Factor Constraints

In a crowded 19 inch rack power supply setup—I'm talking 6U with battery management, inverter, and monitoring—conduction-cooled plates have a real advantage. They're thin. They don't need a radiator or pump block. You can mount them directly to the chassis wall.

Liquid cooling adds volume. Depending on the system, you need:

• A coolant reservoir (typically 0.5–1.5 liters)
• A pump (with its own power draw)
• A radiator (needs airflow)
• Hoses and fittings

That extra footprint can push you from 4U to 5U or force you into a wider cabinet. In one of my 2022 projects, I had to redesign the entire rack layout because I originally spec'd liquid cooling for all three PSUs and ran out of vertical space. Conduction cooling saved the day for two of the three units.

Dimension 3: Total Cost of Ownership (TCO)—The Real Surprise

This dimension is where my thinking changed completely. Conventional wisdom says conduction cooling is cheaper. It's passive, no pump, no fluid, no maintenance. Right?

Sort of. Let's break down the costs I've tracked across 7 different deployments:

Initial hardware cost:

• Conduction-cooled plate: typically $50–$150 per unit
• PSU liquid cooling: typically $300–$800 per unit

Installation and integration:

• Conduction: low—mount the plate, apply thermal paste, attach. One hour labor.
• Liquid: higher—route hoses, fill coolant, test for leaks, ensure pump orientation. 3–5 hours.

Operational costs over 5 years:

• Conduction: zero. No moving parts. No maintenance.
• Liquid: pump replacement ($80–$200) at year 3–4. Coolant replacement. Potential leak remediation.

Here's the conclusion that surprised me: For low-duty-cycle applications (less than 1,000 runtime hours per year), conduction cooling has a lower 5-year TCO. But for continuous-duty applications (like a UPS in a data center running 8,760 hours per year), liquid cooling's thermal performance prevents derating—and that derating actually costs more in lost capacity than the premium for liquid cooling.

I calculated this after the third rejection in Q1 2024. The numbers don't lie.

Dimension 4: Reliability and Risk—Lessons I Wish I'd Learned Earlier

Dodged a bullet in 2023 when I insisted on a test-run for a liquid-cooled PSU before shipping to a utility customer. The pump failed after 72 hours. Turned out the coolant had air pockets from improper filling. If I remember correctly, the manufacturer's installation guide was outdated—didn't specify the bleed procedure. We caught it because we tested. Others might not have.

Conduction cooling is almost impossible to screw up. You attach it. It works. Or it doesn't, and you see it immediately (thermal paste gap, mounting pressure).

But liquid cooling has failure modes you might not think about:

• Pump bearing wear (audible noise after 18 months)
• Coolant degradation (reduces thermal transfer over time)
• Micro-leaks at fittings (slow, invisible, catastrophic over months)
• User error on maintenance (wrong coolant type, overfilling)

For home electricity storage or remote installations where service access is limited, conduction cooling wins every time. For utility or data center contracts with onsite techs, liquid cooling is manageable.

When to Choose Conduction-Cooled Plate

Based on the mistakes I've made (and the ones I've seen others make), choose conduction cooling when:

• Your AC/DC converter runs at <70% continuous load
• Duty cycle is intermittent (under 4 hours per day)
• Space in the 19-inch rack is tight
• Remote or unattended operation (no service visits)
• Budget is constrained—initial cost, not just unit price

When to Choose PSU Liquid Cooling

Go liquid when:

• Continuous high load (above 70%) for hours at a time
• Peak power demands that push the converter to spec limits
• You need consistent performance regardless of ambient temperature
• On-site maintenance staff is available
• Total system TCO analysis shows liquid cooling prevents costly derating events

Final Thought: The $300 Mistake

The $300 I saved by skipping liquid cooling in 2019 turned into $4,200 in rework. The lesson wasn't that liquid cooling is better. The lesson was that I didn't ask the right questions about thermal load and duty cycle before making the decision.

Now I have a checklist. Every new power storage device or 19-inch rack PSU I evaluate gets the same treatment: load profile, duty cycle, service access, and TCO. I wish someone had handed me this framework back in 2017. Period.