Motherboard VRM explained

The VRM is the part of the motherboard that feeds the processor. It matters a great deal if you run a high-core-count chip flat out for long periods or overclock, and hardly at all if you are fitting a 65 W processor for gaming. Most of what board makers print about it, phase counts above all, tells you less than it appears to. This guide explains what the parts do, how to read the numbers and how to tell whether you need to care.

What a VRM does

Your power supply delivers 12 V to the board through the 8-pin CPU power connector near the socket. A processor runs at roughly 1 V and can draw well over a hundred amps at that voltage. The voltage regulator module converts one into the other, and it has to do so very precisely while the load swings from idle to full in a fraction of a second.

The main parts, which you can see around the socket on any board:

• The controller, a small chip that decides moment by moment how much power to send.
• Power stages (MOSFETs), the switches that do the conversion. On all but the cheapest boards each is a single integrated package containing the switching transistors and their driver, often described as DrMOS or a smart power stage.
• Chokes, the grey or black cubes in a row beside the socket, which smooth the current.
• Capacitors, the small cylinders, which smooth the voltage.
• Heatsinks over the power stages, because conversion is never perfectly efficient and the loss comes out as heat.

What a phase is

One power stage with its choke is one phase. A board has many, working in turn, so that each carries a share of the current and gets a rest between pulses. More phases of the same quality mean less heat in each one, a smoother output and more total current before anything reaches its limit. That is the sound engineering reason for high phase counts. It is also why a phase count on its own, without knowing the quality of each phase, is not a measure of anything.

Reading "14+2+1"

Makers quote power delivery as two or three numbers added together. The first is the number of phases feeding the processor cores, and it is the one that matters. The second feeds the processor's integrated graphics or, on AMD, the SoC rail. A third, where present, feeds a minor rail. So a "14+2+1" board has fourteen phases for the cores. Only compare the first number between boards, and only between boards for the same socket.

Why the count can mislead

• Doubled and teamed phases. A controller has a fixed number of outputs, often fewer than the advertised phase count. Makers reach higher numbers either with doubler chips that split one control signal in two, or by driving two power stages from one signal, which some makers call teamed power stages. A board sold as 16 phases may be eight control signals each running a pair of power stages. That is not a defect, and a well-executed teamed design performs very well, but it does mean "16 phases" on one board and "16 phases" on another are not the same claim.
• Amp ratings. Power stages are sold by current rating: 50 A, 60 A, 90 A and so on. Marketing multiplies the rating by the phase count to produce a very large total. The rating is a best-case figure for a cool component. A real power stage under a heatsink inside a warm case delivers a good deal less, and no desktop processor draws anything like the headline total in any case.
• Different parts, same number. Twelve phases of discrete low-cost MOSFETs with no heatsink and twelve integrated 90 A power stages under a finned heatsink are both "12 phases".
• Heatsinks that are mostly decoration. A solid block of aluminium with a large logo has less surface area than a plainer finned one. Mass helps for short bursts; surface area and airflow are what matter over a long render.

When the VRM matters

It matters in proportion to how much power the processor draws and for how long.

• High-core-count processors under sustained load. A Ryzen 9 or a Core i9 rendering, encoding or compiling for an hour keeps the VRM at its maximum continuously. Unlocked Intel Core i7 and i9 chips of the 12th to 14th generations can draw well over 200 W when the board lifts their power limits, which many boards do by default.
• Overclocking, which raises both voltage and current beyond what the chip draws at stock.
• Poor airflow around the socket. A liquid cooler removes the fan that would otherwise blow across the VRM heatsinks. In a case with little front-to-back airflow, VRM temperatures can be noticeably higher with a liquid cooler than with a tower air cooler.

When a VRM overheats, nothing dramatic happens. The board tells the processor to slow down, and performance under load drops until temperatures recover. You would see it as clock speeds sagging during long jobs. Run hot for years, the components also age faster.

When it does not

A 65 W Ryzen processor is limited to 88 W at the socket at stock settings, and the 105 W class to 142 W. Any board with a heatsink on its power stages handles that without effort. Gaming loads are also much lighter than rendering loads: a processor that can draw 200 W in a benchmark typically draws far less in a game. If your build is a mid-range processor and a graphics card for gaming, a sound mainstream B650, B850 or B760 board is enough, and money spent on a heavier VRM buys nothing you will notice.

One Intel detail: processors sold as 65 W parts are allowed to boost well above that figure for short periods, and some boards let them do so indefinitely. A very basic board may hold such a chip closer to its base power to protect its own VRM, which costs a little performance in long all-core work and nothing in games.

Where cheap boards fall short

The real risk is at the bottom of the range. Entry-level boards on chipsets such as A620, A520 and H610 sometimes have a handful of phases and no heatsink at all, because they are designed around 65 W processors. The socket will physically accept a top-end chip, and the board may even list it as supported, but it will not sustain its full performance. If you plan a high-end processor now or as a later upgrade, this is the reason to avoid the cheapest board on the platform. Our guide to how much to spend covers where that line sits.

How to judge a board's VRM

1. Look at the photos. Are there heatsinks over the components on both the left and top sides of the socket? Are they finned or a plain block? No heatsink means a board for low-power processors only.
2. Read the specification for the power stage type and rating, not just the count. Integrated power stages with a stated rating are a better sign than a bare phase count with no detail.
3. Check the maker's CPU support list. Some budget boards list high-end processors with a note about reduced performance or a power limit.
4. Find an independent review that measured VRM temperature under a long load with your class of processor. This is the only real test, and it regularly shows cheaper boards beating dearer ones.
5. Ignore the second 8-pin power connector as a quality signal. One 8-pin connector supplies enough for any mainstream desktop processor at stock; the second is there for extreme overclocking and can be left unplugged otherwise.

VRM and open-box boards

A VRM fault is one of the few problems on a returned board that a visual inspection will not find. It shows up as crashes or throttling under sustained load, not at the desktop. That is why our used motherboard checklist includes a stress test inside the returns window, and why a board that has plainly been run hard, with discoloured areas around the socket, should go straight back. For the wider question of which features deserve your money, see how to choose a motherboard.