A NAS that idles at 40 W loads a 750 W power supply to about 5%, and at that load two units listed at the same $90 measure five points apart. The be quiet! Pure Power 13 M 750W averaged 84.78% across 20 to 80 W loads in the Hardware Busters 115 V runs (Amazon prices checked October 2026). The Corsair RM750e (2025) averaged 79.59%. So the best power supply for a NAS is the one that measures best at 20 to 80 W, not the one with the best badge at 50% load.
We ranked nine current ATX and SFX units by that light-load number. Then we priced the gap at the wall for a server that runs all year and worked out where a picoPSU should win. These power supplies were not on a meter in our lab. Every efficiency figure below is a Hardware Busters measurement read from its review charts, and every watt and dollar figure is our arithmetic on top of it.
Best power supply for a NAS: the measured ranking
Rows are sorted by average efficiency across 20, 40, 60 and 80 W, which is the band a 4 to 8 bay NAS actually lives in. The 2% column is the Intel test point for idle. It scales with the rating, so it means 11 W on a 550 W unit and 17 W on an 850 W unit.
| Power supply | Form factor, depth | 20-80 W average | 2% load | Standby draw | Warranty | Amazon, approx |
|---|---|---|---|---|---|---|
| be quiet! Pure Power 13 M 550W | ATX, 160 mm | 86.34% | 69.68% | 0.04 W | 10 years | lead-time only, $228 |
| MSI MAG A750GLS PCIE5 | ATX, 150 mm | 85.54% | 77.64% | 0.08 W | 10 years (per listing) | about $70, back-ordered |
| be quiet! Pure Power 13 M 750W | ATX, 160 mm | 84.78% | 72.12% | 0.06 W | 10 years | about $90 |
| Corsair SF750 (2024) | SFX, 100 mm | 84.13% | 76.42% | 0.05 W | 7 years | no listing found |
| Corsair SF850 (2024) | SFX, 100 mm | 84.06% | 72.66% | 0.05 W | 7 years | about $174 |
| Corsair RM750x (2024) | ATX, 160 mm | 82.98% | 74.00% | 0.02 W | 10 years | about $123 |
| Seasonic FOCUS GX-750 (ATX 3.1) | ATX, 140 mm | 81.63% | 70.3% | 0.06 W | 10 years | about $151 |
| Corsair RM650e (2025) | ATX, 140 mm | 80.79% | 62.61% | 0.05 W | 7 years | overpriced at $199 |
| Corsair RM750e (2025) | ATX, 140 mm | 79.59% | 64.13% | 0.05 W | 7 years | about $90 list, out of stock |
Prices move daily, so treat the last column as a band and check the live price before you buy. Read against what a home server actually does, the measurements point to five picks:
- Best low-idle ATX unit you can order now: be quiet! Pure Power 13 M 750W, with 84.78% at 20 to 80 W, a fan that stays under 10 dBA and a 10-year warranty, for about $90.
- Best measured, once it is back in stock: MSI MAG A750GLS PCIE5, first among the 750 W units on both light-load columns, back-ordered at about $70 and listed with a 10-year warranty.
- Best for a short ATX bay: Seasonic FOCUS GX-750, 140 mm deep with a fanless mode and 10 years of cover, at a price premium.
- Best SFX unit: Corsair SF850 (2024), with 84.06% at 20 to 80 W from a chassis only 100 mm deep.
- Best for SSD and one-drive servers: Mini-Box picoPSU-160-XT on a 12 V brick, which moves the AC stage out of the case.
We kept the Corsair RM750e in the table because it is a common NAS pick, and an earlier version of our TrueNAS SCALE storage server build named it too. It is a sound power supply, and the cheap copies on Amazon were Renewed listings when we checked. It also finished last on the 20-80 W average, so it suits a box that spends real time under load better than one that idles 20 hours a day.
How the numbers were taken
All efficiency data comes from Hardware Busters, where every unit goes through the same 115 V load sequence. We read each figure from the charts in that unit’s review. We kept the 115 V result because it is the harder case and the one North American readers see. Each column was taken this way:
- 20-80 W average: the Hardware Busters average efficiency across its 20 to 80 W load range, which is the closest published proxy for a NAS idling with spinning disks.
- 2% load: the idle point in Table 4-15 of the Intel ATX 3 power supply design guide for units above 500 W. The guide requires 60% there and recommends 70%.
- Standby draw: what the unit pulls from the wall with the system off, which matters for a server that sleeps on a schedule.
- Price and stock: Amazon catalog data on the day of writing. A unit with no buyable listing stays in the table for its data but gets no buy link.
Hardware Busters also argues that the old rule, big power supply means poor light-load efficiency, no longer holds. Modern controllers switch into burst mode at low output. The same author reports that some brands certify units with burst mode on and then ship them with it off. So we ranked on independent review measurements, not on the tier printed on the box, and that is how two 750 W units at the same price end up five points apart on the 20-80 W average.
Step 1: Work out the idle and spin-up load
Size the power supply from two numbers. Idle draw sets the electricity bill, and the 12 V spin-up surge sets the minimum rating. The WD Red Plus product brief gives both for every capacity. The spread between a 4 TB and a 12 TB drive is larger than most builders expect:
| WD Red Plus model | 12 V peak (spin-up) | Idle | Read/write | Standby |
|---|---|---|---|---|
| WD40EFPX (4 TB) | 1.75 A | 3.1 W | 4.7 W | 0.3 W |
| WD80EFPX (8 TB) | 1.75 A | 3.4 W | 5.2 W | 0.4 W |
| WD120EFGX (12 TB) | 1.9 A | 6.1 W | 8.8 W | 0.3 W |
Four 12 TB drives idle at 24.4 W on their own. Add a board and boot SSD near 10 W, the target of our low-power home server build, and the box sits around 35 to 40 W DC. Four 4 TB drives bring the same box down to about 22 W. Both totals land inside the 20 to 80 W band, so that column is the one that predicts your bill.
Spin-up is the opposite problem. Eight 12 TB drives starting together pull 15.2 A from the 12 V rail, about 182 W before the CPU does anything. Four drives pull 7.6 A, about 91 W. A 750 W unit delivers nearly all of its rating on 12 V (Seasonic rates the FOCUS GX-750 at 62 A, or 744 W). Capacity is therefore never the limit for a NAS with a 65 W desktop CPU and no GPU, whose realistic peak stays well under 300 W.
Connectors are the real limit. Split the drives across at least two SATA cables so one daisy chain does not carry the whole surge, and count the SATA plugs against your bays. The Pure Power 13 M 750W and the RM750e ship 6 plugs, while the RM750x and the SF850 ship 8.
Step 2: Pick an ATX power supply by its 20 to 80 W result
Rank ATX candidates by the 20-80 W column first, then filter by depth and warranty. The 550 W be quiet! posted the best number of the nine we ranked at 86.34%. Its only Amazon listing was a lead-time offer at more than double the 750 W price, though, so the 750 W unit at 84.78% is the one you can actually order. The smaller unit would gain a NAS 1.6 points, worth about 0.9 W at a 40 W idle.
be quiet! Pure Power 13 M 750W
Verdict: for most ATX NAS cases, buy the be quiet! Pure Power 13 M 750W (about $90, check the live price). It measured 84.78% at 20 to 80 W and 72.12% at its 15 W test point, and its fan never passed 10 dBA in the review.
The cost is depth. At 160 mm it is 20 mm longer than the Seasonic and Corsair RMe units, which matters where the power supply sits next to a drive cage. Skip it if your case lists a 140 mm limit; our NAS case guide gives the clearance for each chassis.
MSI MAG A750GLS PCIE5
The MSI MAG A750GLS PCIE5 measured better still: 85.54% at 20 to 80 W and 77.64% at 2%, the top 750 W result on both charts in the Hardware Busters review.
It is 150 mm deep and Huntkey-built, with semi-passive fan control. Two things keep it below the be quiet! for now: Amazon listed it as back-ordered at about $70, and it ships only 5 SATA plugs. The listing gives a 10-year warranty, which the review does not state. If it is in stock when you buy and five drives are enough, it is the better-measured unit for less money.
Seasonic FOCUS GX-750 (ATX 3.1)
Choose the FOCUS GX-750 when the bay is short or the server lives in a quiet room. It costs the most of the buyable ATX picks:
Seasonic lists the 750 W model at 140 x 150 x 86 mm with a 135 mm fluid dynamic bearing fan. Its Digital Hybrid Fan Control stops the fan entirely at low load. That fanless mode is the main reason to pay about $151 (check the live price), because a NAS idling at 40 W never wakes the fan and the drives become the only sound. On efficiency it sits in the lower half at 81.63% and 70.3% at 2%, about 1.8 W more at the wall than the be quiet! at a 40 W idle. Skip it if the case takes a 160 mm unit, since the be quiet! is cheaper and more efficient.
Corsair RM750x (2024)
The Corsair RM750x (2024) sits between the be quiet! and the Seasonic at about $123. Its 82.98% average and 74.00% at 2% both beat the Seasonic, and its 0.02 W standby draw is the lowest measured here.
It also brings 8 SATA plugs and a 10-year warranty. At 160 mm deep it competes with the be quiet! rather than the Seasonic, and it loses that contest on price and on the 20 to 80 W number.
Step 3: Use an SFX unit only when the case forces it
Buy SFX when the chassis accepts nothing else, as several compact Jonsbo boxes in our NAS case roundup do. The efficiency penalty is smaller than it used to be. The Corsair SF850 (2024) averaged 84.06% at 20 to 80 W and the SF750 averaged 84.13%, within a point of the be quiet! 750 W ATX pick, from a chassis 100 mm deep.
Corsair SF850 (2024)
Our SFX pick is the Corsair SF850 (2024) at about $174 new (check the live price).
The SF750 measured better at the 2% point, 76.42% against 72.66%, but we found no buyable Amazon listing for it. Those two numbers are not like for like either, because the 2% point is 15 W on the SF750 and 17 W on the SF850. Both ship 8 SATA plugs and a 7-year warranty.
Noise is not the trade-off. In the Hardware Busters standard load test the SF850 fan sat at 0 RPM at 10, 20 and 30% load (85 to 255 W) and only started at 40% (340 W), so a NAS idling at 40 W keeps it passive. Price and cover are the trade-off: the SF850 costs about $84 more than the be quiet! ATX unit, measures slightly lower and carries 7 years of warranty against 10. Skip SFX if your case takes ATX.
Step 4: Drop to a picoPSU for SSD and one-drive builds
Below about 20 W DC, a picoPSU on an external 12 V brick should beat the ATX numbers above, and it suits only small builds. The 2% column shows why we expect that, although nobody has published a measurement of the whole picoPSU and brick chain at idle. At 15 W the best unit in the table wastes 4.3 W and the worst wastes 8.4 W, so on a 10 W mini server the power supply loss can approach the load itself.
Mini-Box picoPSU-160-XT
Plug the picoPSU-160-XT straight onto the 24-pin board socket and feed it regulated 12 V, so the AC conversion happens in a fanless brick outside the case:
The Mini-Box manual sets limits per rail. The 12 V rail is a switched passthrough from the brick rather than a converter, so the brick rating is what the drives and the CPU really draw from:
| Rail | Continuous | Peak |
|---|---|---|
| 12 V (passthrough) | 8 A | 15 A |
| 5 V | 8 A | 10 A |
| 3.3 V | 8 A | 10 A |
| 5 VSB | 1.5 A | 2 A |
Combined output must stay under 200 W, and peaks may last 60 seconds. At full load the 5 V and 3.3 V rails need forced air, or derating until the board runs under 65 °C. The limit that decides NAS use is smaller. The drive harness carries one SATA and one Molex plug and is rated for 3 A of ground return. A single WD Red Plus pulls 1.75 to 1.9 A on 12 V at spin-up, so the harness supports one 3.5-inch drive within its rating, not four. We therefore recommend the picoPSU-160-XT (about $50, check the live price) for SSD-only servers, router boxes and one-drive backup targets. Anything with a drive cage gets an ATX unit from Step 2.
Pair it with the Mini-Box 192 W 12 V adapter, rated Level VI with at least 88% average efficiency. Level VI under 10 CFR 430.32 also caps a 49 to 250 W adapter at 0.21 W with nothing attached. Neither the rule nor Mini-Box publishes efficiency at a 10 W load, so real brick loss at idle is the one number here nobody has measured. Our sub-10 W home server build walks through a picoPSU install on a board that reaches deep package C-states.
Step 5: Convert the efficiency gap into watts and dollars
Wall draw is DC load divided by efficiency. For each unit, the table applies the 20-80 W average to a 40 W DC load, which is a 4-bay NAS with 12 TB drives. It applies the 2% figure to the 2% load point. The 40 W case is priced at $0.20 per kWh over 8,760 hours a year. A single 40 W point is an approximation of a 20 to 80 W average, so read differences under 0.5 W as noise.
| Power supply | Wall draw at 40 W DC | Loss at 40 W | Cost per year at 40 W | 2% point: DC load, wall draw |
|---|---|---|---|---|
| be quiet! Pure Power 13 M 550W | 46.3 W | 6.3 W | $81 | 11 W, 15.8 W |
| MSI MAG A750GLS PCIE5 | 46.8 W | 6.8 W | $82 | 15 W, 19.3 W |
| be quiet! Pure Power 13 M 750W | 47.2 W | 7.2 W | $83 | 15 W, 20.8 W |
| Corsair SF750 (2024) | 47.5 W | 7.5 W | $83 | 15 W, 19.6 W |
| Corsair SF850 (2024) | 47.6 W | 7.6 W | $83 | 17 W, 23.4 W |
| Corsair RM750x (2024) | 48.2 W | 8.2 W | $84 | 15 W, 20.3 W |
| Seasonic FOCUS GX-750 | 49.0 W | 9.0 W | $86 | 15 W, 21.3 W |
| Corsair RM650e (2025) | 49.5 W | 9.5 W | $87 | 13 W, 20.8 W |
| Corsair RM750e (2025) | 50.3 W | 10.3 W | $88 | 15 W, 23.4 W |
From best to worst, the spread is 4 W at the wall, about 35 kWh or $7 a year. The be quiet! 750 W saves about $5 a year over the RM750e at the same $90 list price, and it carries 10 years of warranty against 7. At equal cost, that makes the RM750e the harder unit to justify for a box that idles all day. SFX units sit in the same band as the good ATX units, so an SFX case costs nothing in efficiency, only in purchase price.
Next to the drives, the power supply stops being the main lever. Four WD120EFGX drives idle at 24.4 W and draw 1.2 W in standby. Letting them spin down in quiet hours saves about 23 W while they sleep, several times the whole spread in this table. If you are still choosing parts, the idle column in our NAS hard drive comparison and the boards in our NAS motherboard guide move the bill further than any power supply here. Our DIY NAS vs prebuilt numbers show where those savings land against a store-bought box. Buy the power supply with the best 20 to 80 W number your case fits, then spend the effort on spin-down and C-states.
