We all know a CPU requires a heatsink in order to stay at a safe temperature, but one supporting component that does not get the attention it deserves is the VRM. Your VRMs, whether single stage or multistage, might require a heat sink in order to stay within the right operating temperature.
In VRMs built from discrete components, you’ll have some difficulty if you do decide to use a VRM heatsink for temperature regulation. In VRM ASICs, the design of the chip itself determines whether a VRM heatsink is necessary. If either type of VRM is designed properly, you won’t need to include undue thermal management measures, unless of course you decide to overclock your GPU...
When to Use a VRM Heatsink
The subject of when to use a VRM heatsink is rather interesting. It depends on a number of factors, including the discrete components used in your board, the thermal conductivity of your PCB substrate, the presence of any other active cooling measures, and whether you are using a single-stage or multistage VRM. Let’s quickly remember how a VRM operates and where heat is generated.
A VRM is normally implemented as a buck converter, so it uses a PWM signal with low duty cycle to switch a power MOSFET. The output voltage is regulated between two levels with a large capacitor. A single-stage VRM contains two MOSFETs, while a multistage VRM contains two MOSFETs per stage. The MOSFETs in each stage switch in succession, which then modulates the level of regulation in an LC circuit at the output from the VRM.
If you look at a typical power MOSFET, you’ll see a large metal pad on the back of the component; this is the MOSFET’s heatsink. I’ve seen some designers attach custom-built heatsinks directly to these pads to provide heat dissipation.
Power MOSFET with built-in heatsink.
Once placed on the board, the power MOSFETs are angled face-down, meaning the MOSFET heatsinks sit directly against the board. Normally, these boards are soldered to a copper thermal pad, which allows heat to move to the backside of the board through thermal vias. Most of the heat will accumulate on the paddle, and placing the MOSFET paddles against the board means that the board itself acts like a heatsink. If you expose a computer motherboard and locate the VRM MOSFETs, try turning on the computer and wait for the system to heat up. You’ll find that the back of the board gets very hot.
MOSFETs in a VRM (see the right-hand side of this image). Note that the MOSFET heatsinks act like the VRM heatsink for thermal regulation.
In the example shown above, the top surface of each MOSFET is insulated, so very little heat flows away from the board and an additional heatsink on top of these components is not necessary. Some manufacturers will still add heatsinks on the component surface for decoration purposes. Some modern VRMs are designed so that the heat pads of the MOSFETS are on the top surface, which are primarily used in GPUs. On these, the heatsink is essential, as these VRMs do not use the motherboard as a heatsink.
Where to Put an Additional VRM Heatsink
If you do need an additional heatsink, where should it be placed? If the MOSFET pads sit against the board, and the back side of the board reaches an unacceptable temperature, then you would want to put a heatsink against the back side of the board. One way to do this is to extend the copper thermal pad for the MOSFET to the back side of the board. During assembly, you can place a thermal pad or other thermal interface material (TIM) directly on the copper. You then attach the heatsink directly to the thermal pad, as shown below. This will remove heat from the back of the board and dissipate it into the surrounding air.
VRM heatsink attached on the back side of a board.
Single-stage vs. Multistage VRM Heatsink and Temperature
Note that a multistage VRM has some advantages here beyond the power stability benefits. A multistage VRM already has lower ripple in the output voltage compared to a single-stage VRM running at the same frequency. If you wanted to have a single-stage VRM provide the same ripple as a multistage VRM, then the single-stage VRM will need to run at a higher PWM frequency. This will cause your single-stage VRM to reach a higher temperature than an equivalent multistage VRM with the same ripple.
If you do need to cool each stage in a multistage VRM together, you can use a shared heatsink that spans across all the components. This requires placing a pad on each stage (either on the back of the board or on the top of the component) and connecting a large heatsink across all stages on the VRM. This is used in VRMs for some high performance graphics cards and is definitely recommended if you are overclocking your system.
Finally, there are some digital and analog VRM controller ICs available on the market. Some of these ICs can support 6 stages in a multistage VRM, providing extremely smooth output voltage. These ICs also generate a significant amount of heat, and the heatsink mounting strategies outlined here apply equally to these ICs.
Building a layout with a VRM heatsink can be difficult, but you can easily layout a VRM and add any require heatsink when you use the right PCB design and analysis software. OrCAD PCB Designer and Cadence’s full suite of analysis tools can help you with thermal management in complex layouts to ensure your next product operates as designed.
If you’re looking to learn more about how Cadence has the solution for you, talk to us and our team of experts.