• Home
  • blog
  • Tiny, Silent and Efficient: The picoPSU

Tiny, Silent and Efficient: The picoPSU

blog image

The tiny picoPSU is the most unconventional ATX12V power supply we’ve seen. By paying attention to how power supplies are actually used, not how they’re specified, the creative guys at Mini-box have come up with an amazing tool for custom builders, silent PC enthusiasts, and SFF modders. It’s not for everyone, but the picoPSU is right up our alley: Very efficient, innovative, and silent.

May 10, 2006 by Devon
Cooke
with Mike Chin

Product
picoPSU-120
12V, 120 Watt DC-DC ATX power supply
Manufacturer
Mini-box.com
Market Price
~US$50 + a 12V power brick

A 120 watt ATX power supply? Surely you jest. How can a 120 watt power
supply hold its own in a world where there are power supplies that put out nearly
ten times
that amount of power?

Here’s the secret: High powered systems may be as popular as ever, but the industry’s growing interest in performance-per-watt has people thinking about power efficiency.
Besides, just because a system is powerful doesn’t mean it has to be power-hungry.
We’ve known for ages that
even powerful systems rarely draw more than 200 watts, and the proliferation
of low power, high performancce Athlon 64 chips and even lower power Pentium M, Core Duo, and Turion
64 chips have made it easier than ever to build a system that peaks below 100
watts.

For a modest system, a 120W power supply like the picoPSU is an elegant
alternative to a conventional power supply. What makes
the picoPSU different from any other power supply? A wise man once said “A
picture is worth a thousand words”. Let’s take his advice:


Where’s the rest?

No, that’s not just one of the cables. That’s the whole darn thing. As its
name suggests, the picoPSU is tiny — so tiny that 70 picoPSUs would fit
inside the casing of a normal ATX power supply. The advantages of the picoPSU
over a conventional power supply is obvious:

  • It will fit in just about any case — even the smallest Small
    Form Factor systems with proprietary power supplies
  • It’s fanless, which means no noise
  • Cable clutter is minimal

The picoPSU is so simple that it’s amazing that someone hasn’t done this before.
Strictly speaking, that’s not quite true, someone has done this before,
but that someone is the same company. A year ago, Mini-box introduced a product
called the
PW-200M
, which is more or less the same thing in a bulkier package and still available. Compared
to its predecessor, the picoPSU is much smaller. Because its footprint is no bigger than the ATX connector itself, it’s difficult to imagine any motherboard or system that the picoPSU would not fit into.

The secret to the picoPSU is that it’s really only half a power supply. The
bulkiest part of the power conversion, 120V or 240VAC to 12VDC, is
taken care of by an external power brick which feeds the picoPSU. Because
most of the power required by a modern computer is in 12V form, all
the picoPSU has to do is pass most of the power straight through, taking only
what it needs to power the +5V and +3.3V lines. And, as we
noted in our article on power distribution
, the total power requirements
on these two lines is rarely more than 20~30 watts.

The engineers at Mini-box know this, and have taken the lesson to heart. The picoPSU can deliver only 6A on each of the +5V and +3.3V lines. But in most current systems
that’s enough.

FEATURE HIGHLIGHTS

Feature Highlights of the picoPSU-120 from Mini-box
(from
Mini-box.com)
FEATURE & BRIEF COMMENT
Very small, can build ultra-compact PC enclosures and slim server
enclosures.
Ideal for replacing noisy
power supplies in small form factor systems.
Fits any motherboard equipped
with a 20 or 24pin ATX connector.
20-pin by default, but
most 24-pin motherboards work fine without the extra four pins. An adapter
is included for “true” 24-pin systems.
100% silent, fanless, no moving parts.
Sounds good… real good.
Operates at only 12V.
Requires
an external power brick.
Highly efficient design, does not produce a lot of heat.
Most of the heat is generated
in the power brick, which is located outside the case.
Ideal for low power
Intel Core Duo as well as all VIA C3/C7 motherboards.
Just don’t try to run
any recent gaming cards on it. Intel’s Netburst chips are out too.

SPECIFICATIONS

OUTPUT SPECIFICATIONS: picoPSU from Mini-box (from
the picoPSU manual)
Input
12V DC
Input Current
2A~10A
DC Output
+3.3V
+5V
+12V
-12V
+5VSB
Maximum Continuous Output Current
6A
6A
7A
0.05A
1.5A
Peak Output Current
8A
8A
10A
0.1A
2.0A
Maximum Combined
120W
Peak Combined
140W
At max load, forced air ventilation is required. For fanless operation
de-rate the output of the 3.3 and 5V rails by ~20%. Peak load should not
exceed 60 seconds.

True to its name, the power output numbers are also tiny. Even so, neither
the +3.3V nor the +5V line is ever likely to be overloaded. Only the +12V line
looks like it might be a little on the weak side. But, keep in mind that the
+12V power is supplied from the external power brick, so the 7A rating is
probably only tentative. The “true” +12V capacity is probably
the rating of the brick itself, minus whatever power the picoPSU uses to supply
the other voltages.

Even if 7A is the maximum (a reasonable estimate for a 120W power brick),
there’s still 84 watts available for continuous output. Thanks to the
ever-decreasing power consumption of AMD’s processors, it’s quite easy to build
a fully-featured desktop system within an 84 watt envelope. With an especially
cool processor, it may even be possible to jam in a mid-range graphics card.
Even if it jumps above 84 watts every once and a while, there’s enough headroom
for peaks that it’s unlikely to fail immediately.

CLAIMED EFFICIENCY: picoPSU from Mini-box
+5V Load
+5V Efficiency
+3.3V Load
+3.3V Efficiency
1A
86%
1A
85%
3A
94%
3A
93%
5A
96%
5A
94%
8A
93%
8A
91%

Mini-box provides detailed efficiency specifications for the picoPSU. In the
context of how efficiency is usually measured, the figures are quite unbelievable
— 96% efficiency? Impossible!

It’s important to understand just how efficiency is measured for the picoPSU.
The picoPSU is a DC to DC power supply, so efficiency is calculated as
DC watts out ÷ DC watts in × 100%. The power lost in the AC to
DC conversion is not counted. Naturally, efficiency seems higher. On the other
hand, a convention power supply converts AC to DC, and therefore the
efficiency includes those losses, calculated as AC watts in ÷ DC watts
out × 100%.

Although the power brick is sold separately, Mini-box supplied two power bricks
to test alongside the picoPSU, one
rated for 80W
, the other for either
110W or 120W
(depending on which numbers you believe). Aside from their ratings,
the main differences between the two appeared to be size and cooling. The larger
of the two power supplies had cooling vents at either end and a tiny cooling
fan that turned on under heavy load.


Two external power bricks.


Cooling vents at the end of the 110W unit are evidence of active cooling.

Both bricks are from EDac.
They feature active power factor correction and come with short-circuit,
over-voltage, and over-current protection. In addition, the picoPSU itself has
over-voltage protection if the input voltage rises above 13~13.5V. The 120W
brick boasts of efficiency above 85%, so we hope to see it do well on our test
bench. The bricks also feature numerous logos from various international certification
agencies. There is no reason to expect that the use of an external brick is
any less safe than a conventional power supply.

The power bricks let down our expectations in only one respect,
voltage regulation. They are rated to maintain +12V ±10%, which
is greater than the ±5% tolerance of the ATX12V specification. As the power bricks
are likely to be used at close to full load, voltage stability could be a concern.
Mini-box wisely rates the +12V voltage regulation as dependent on the “switched
input”.

Regardless of the potential for fluctuations on the +12V line, the picoPSU
has a specified tolerance of just 1.5% for both the +5V and +3.3V lines.


Both bricks are made by EDac and are well certified.

OVERVIEW

There’s only room for a very simple regulation circuit on the picoPSU. Two parallel
circuit boards are used. Approximately half of each board is devoted
the generating +5V, while the other half contains similar circuitry for the
+3.3V line.

The -12V and +5VSB lines are not immediately visible. It is plausible that
the standby line uses the same circuitry as the main +5V line. The appropriate
pins would be disabled when the power supply is not active.


Up close and personal.


+12V input is on the top left, while the four wires on the right carry +5V
and +12V.


Two PCBs provide all the power that is needed.


All hooked up and ready to go.

CABLES AND CONNECTORS

Numerous adapters offset the lack of cable sets.

As is fitting for such a tiny power supply, the number of cables is kept to
a minimum. In fact, aside from the main ATX connector, the only other output
plugs are two IDE drive connectors and a floppy connector, all on the same short
cable set. Another plug on a separate cable accepts the +12V input. The plug
is designed to be mounted on a metal panel. A PCI slot cover would be ideal,
but the appropriate metal bits were not included.

To make up for the lack of cables sets, a number of adapters are included so
that different kinds of devices can be connected. Most important of these is
an adapter for the +12V AUX plug found on nearly all mainstream motherboards.
A 20-pin to 24-pin adapter is also included, although most 24-pin boards will
function just fine without the extra four pins connected. Only systems with
a high-powered PCI Express graphics card are likely to need the extra four pins,
and this kind of system is likely to overload the picoPSU anyway. Surprisingly,
a PCI Express adapter is also included, although any card that requires the
external power connector is likely to be too much for the external power brick
to handle.

Another advantage of the adapters is their ability to extend the length of
the single cable set, which needs to be everywhere at once. Luckily, the ATX
header, which is where the cables start from, is a much more central location
than where the power supply is usually located, so the short cable was less
of an issue that it might have been.

Even with the numerous adapters, a couple of adapters are sorely
missed. One is an adapter for SATA drives. The other is a Y-splitter, so that the number of IDE
connectors could be increased. With one connector occupied by the +12V AUX plug
and the other by the hard drive, there were simply not enough plugs to go around
to add an optical drive. This is a serious oversight, as even a low powered
system is likely to require all three of these. Only VIA-based systems, which
do not require the +12V AUX connector are safe in this regard. Keep in mind that both SATA adapters and Y-splitter are available for purchase from most computer component stores.

TEST RESULTS


The picoPSU in action.

For a fuller understanding of ATX power supplies, please read the reference
article Power Supply Fundamentals & Recommended
Units
. Those who seek source materials can find Intel’s various PSU
design guides at Form
Factors
.

For a complete rundown of testing equipment and procedures, please refer to
SPCR’s PSU Test Platform
V.3
. The testing system is a close simulation of a moderate airflow
mid-tower PC optimized for low noise.

Because of the unique design of the picoPSU, no thermal testing was done. Not
only is our standard test bench ill-equipped to test a power supply on the actual
ATX header, but the levels of power are so low that not much heat is generated
anyway. Unlike a conventional power supply, the picoPSU plays no role in removing
heat from the system, so it would be unfair to subject it to the rigors of our
standard test. By the same token, noise levels also did not need to be tested,
with the exception of the fan in the larger of the two power bricks.

The picoPSU was run through two complete tests: One for each of the supplied power bricks.
By necessity, the measurements reflect the combined performance of the picoPSU
and the power brick. Testing them independently would have been far too impractical,
and would have been of limited use in the real world anyway.

Ambient conditions during testing were 20°C and 20 dBA, 121V/60Hz.

TEST RESULTS: picoPSU from Mini-box, 80W brick






DC Output Voltage (V) + Current (A)
Power Factor

Total Output

Input

Calculated Efficiency
+12V1
+5V
+3.3V
-12V
+5VSB
12.12V
1.05A
5.03V
1.06A
3.32V
1.04A
0.0
0.0
0.83
21.5W
27.1W
79.4%
11.94
2.05
5.01
2.11
3.32
1.03
0.0
0.5
0.92
41.0
49.0
83.6%
11.73
3.03
4.99
3.11
3.32
2.98
0.0
0.8
0.98
65.0
76.6
84.8%
11.65
4.00
4.98
3.12
3.32
2.01
0.0
0.8
0.99
72.8
86.0
84.7%
11.61
3.97
4.98
3.13
3.32
4.20
0.0
1.0
0.99
80.6
95.3
84.6%

Miscellaneous Results:

  • Power Draw in Standby: 1.5W
  • Power Factor in Standby: ~0.30
  • Power Draw with no load applied: 2.5W
  • Power Factor with no load applied: ~0.36

NOTE: The current and voltage for -12V and +5VSB
lines is not measured but based on switch settings of the DBS-2100 PS
Loader. It is a tiny portion of the total, and potential errors arising
from inaccuracies on these lines is <1W.

* * *

TEST RESULTS: picoPSU from Mini-box, 120W brick






DC Output Voltage (V) + Current (A)
Power Factor

Total Output

Input

Calculated Efficiency
+12V1
+5V
+3.3V
-12V
+5VSB
12.32V
1.07A
5.03V
1.06A
3.32V
1.05A
0.0
0.0
0.96
22.0W
28.3W
77.7%
12.18
2.11
5.01
1.05
3.32
2.04
0.0
0.5
0.98
40.2
47.0
85.6%
12.00
3.11
4.99
3.14
3.32
2.01
0.0
0.8
0.99
63.7
73.1
87.1%
11.80
5.13
4.99
3.15
3.31
3.02
0.0
1.1
0.99
91.7
105.3
87.1%
11.64
6.05
4.94
5.04
3.30
5.14
0.0
1.5
1.00
119.8
143.0
83.8%

Miscellaneous Results:

  • Power Draw in Standby: 3.7W
  • Power Factor in Standby: 0.56
  • Power Draw with no load applied: 4.7W
  • Power Factor with no load applied: 0.60

NOTE: The current and voltage for -12V and +5VSB
lines is not measured but based on switch settings of the DBS-2100 PS
Loader. It is a tiny portion of the total, and potential errors arising
from inaccuracies on these lines is <1W.

ANALYSIS

1. VOLTAGE REGULATION

The picoPSU itself provided excellent voltage regulation, staying well within
the ±1.5% specified. In fact, the +3.3V line did not appear to fluctuate
at all — it stayed stuck at +3.32V throughout the test. However, the +12V line from the power bricks sagged significantly
under heavy load, dropping by almost 0.5V. Even so, it always stayed within
the ±5% tolerance specified by ATX12V. The voltage regulation with the 120W brick was more or less the same as with
the 80W brick, with a little less sag on the 12V line at loads >60W. Perhaps because of the higher amount of power being drawn, the
two lines controlled by the picoPSU did show slightly more variation when delivering >80W with the larger power brick. Even so, neither of the lower voltage lines were ever in danger of
exceeding the tight ±1.5% tolerance.

Overclockers would probably be scared off by the weak voltage
regulation on the +12V line, but for the rest
of us the voltages are good enough. Understandably, the drop was most significant
as the maximum load was approached, so the solution is simple: Either build
a more efficient system, or use the larger 120W brick.

2. EFFICIENCY was excellent with both the 80W brick and the 110W brick — much higher than any other power
supply we’ve measured at these low output power levels. Part of this is by design —
the low capacity of the brick and the picoPSU mean that efficiency is optimized
over a much smaller output range than conventional PSUs, which are rated well over 300W these days. However, there is more at work here. The efficiency wasn’t just high
for the load, it was high, period. In fact, only two other power supplies
we’ve tested have had higher efficiency than 84.8% at any level. That the picoPSU
+ 80W power brick was able to maintain this through a large part of its output range is impressive.

The 87.1% peak efficiency
makes the picoPSU + 120W brick the most efficient power supply we have
tested by almost two percentage points. Even better, this efficiency was maintained
between 65W and 90W — precisely what most minimalist AMD-based systems demand when
under heavy CPU load. This will make a very real difference in power draw, as
very few conventional power supplies reach even 80% at such low loads.

Ironically, even this combination would probably fail the 80
Plus certification program
, which requires that a power supply be at least
80% efficient at 20%, 50% and 100% load. Ordinarily, the 20% point is the crucial
one: It’s the most important in real terms, and it’s the hardest target to achieve, especially for lower powered PSUs. The picoPSU + 120W brick would also fail at the 20% test point; 20%
of 120W is 24W, where the efficiency was ~78%.

3. POWER FACTOR

This result seems mostly dependent on the power bricks. Although EDac advertises active power factor correction for the 80W brick,
it measured quite low until the output reached at least 40W. At
20W output, the power factor was 0.83 — higher than is achievable with
passive power factor correction, but much less than we’ve seen from other forms
of active power factor correction. Above 40W, the power factor was always at
the expected 0.9 and above.

The 120W brick was much better. PF was well above
0.90 even at 20W, and even touched 1.00 during the last test point.
In standby mode, it maintained a respectable 0.60.

4. FAN & FAN CONTROLLER

The 80W unit doesn’t have a fan, so this section applies only to the 120W unit. It’s the one downside to the 120W brick — the built-in cooling fan.
The fan turned on promptly when we boosted the output load to 90W, and it was loud. The 36 dBA@1m that we measured was far too loud
to be acceptable in a quiet system. The fan speed increased slightly when full
load was reached, but by then it was already too loud to consider.

Because of the fan, the 120W brick is best treated as a 90W unit that can sustain
short peaks of 120W. Any sustained load above 90W would be unacceptable to users
of quiet systems.

MP3: EDac
120W power brick @ 90W output: 36 dBA@1m

COMPARISON

To illustrate just how efficient the picoPSU / EDac power bricks are, one need
only look at the best of the best that we’ve previously tested.
Take careful note of where the peak efficiency is reached, and keep in mind
that a system without an external graphics card typically requires between 65~90W
under load. Toss in a mid-powered graphics card, and that number jumps to 90W~120W.

Efficiency Comparison: picoPSU vs. Past Efficiency
Champs
Target Output Power
40W
65W
90W
Peak Efficiency
Load @ Peak
picoPSU + 80W power brick
83.6%
84.8%
84.6%
84.8%
65W
picoPSU + 120W power brick
85.6%
87.1%
87.1%
87.1%
65~90W
Seasonic SS-300SFD 80 Plus
81.5%
83.6%
84.6%
85.2%
200W
Seasonic 400HT Active PFC F3 80 Plus
76.6%
81.5%
82.8%
85.3%
150~200W
FSP Zen FSP300-60GNF
76.6%
80.4%
83.0%
84.6%
150~200W
Silentmaxx Fanless 400 Watt MX460-PFL01
70.6%
76.6%
79.6%
84.1%
200W

Only one of the conventional power supplies manages to match the efficiency
of the picoPSU anywhere: The Seasonic SS-300SFD and the picoPSU / 80W brick
both reach 84.6% efficiency at 90W. The 120W brick is unchallenged;
its peak efficiency higher than any other power supply and it is right where
most systems hover.

IN-SYSTEM TEST

The picoPSU needs to prove
itself outside the test conditions in an actual system. Things tend to go wrong in less controlled circumstances
— temperatures are higher, the loads are dynamic, and there’s always the
risk that a stray peak might burn something out. In addition, using the picoPSU
removes both a source of heat and a source of airflow, so some real-world thermal
data could be very useful.

The picoPSU was installed in a low-powered HTPC setup, and we ran a basic before/after
test to check for differences in temperatures, stability, and power consumption.
The following components were used:

  • Zalman HD160 HTPC case
  • AMD A64 3000+, Venice Rev. JH8-E3. TDP is 44.1W, and TCaseMax is
    57°C
  • DFI RS482 motherboard
  • Onboard Video w/ ATI 200 chipset
  • 512 MB OCZ PC4000 DDR RAM, one stick
  • Zalman 7700CU heatsink modded with a Nexus 120mm fan @ 12V
  • 2 x 80mm Zalman case fans @ 5V
  • Samsung Spinpoint P120 250GB hard drive
  • Samsung CD-RW drive

Baseline results were measured using a
recently-reviewed Seasonic S12-330 Rev. A3
power supply.

IN-SYSTEM TEST

PSU
Idle
CPU Load
(
CPUBurn)
CPU Load
Temperature
Seasonic S-12 330
47.5W
80.0W
46°C
picoPSU +
120W Power Brick
39.5W
72.6W
45°C

The system remained perfectly stable throughout the test; there was never any reason
to suspect that it might be overheating or overloading. However, we did run
out of connectors to power everything: The need to use one of the Molex connectors
to power the +12V AUX cable meant we did not have a spare connector for the
optical drive. We also used an adapter for our SATA drive, as it did
not accept the older IDE-style connector.

The most significant difference with the picoPSU was a drop in power consumption.
The AC power consumption dropped by approximately 8W both at idle and under
load. At idle, that difference was nearly 20% — a huge amount considering
that the electrical load did not change. Even under load, the difference was
still 10%.

On top of that, the ambient temperature in the system also dropped slightly,
no doubt because the power supply had no thermal impact at all. Any power lost in the AC-to-DC conversion was lost outside of the
case, where it could not affect the system temperature. In other words, the
use of the picoPSU actually improved the thermal environment inside the case.

Finally, the picoPSU made the system a bit quieter, as the 120mm fan in the PSU was eliminated as a source of noise. With the exhaust and heatsink fans, and the HDD still making noise, the drop in noise was subtle, not much more than 1 dBA, but it was audible.

CONCLUSIONS

It’s hard not to be delighted with the picoPSU. It’s quiet (fanless, in fact), efficient,
small, cool… the list goes on. Although it is targeted at systems using VIA’s
embedded EPIA technology, it can easily be used in a desktop system with a low-powered
Athlon 64, to say nothing of a laptop processor such as a Core Duo or a Turion
64.

Its only real disadvantage is its size: 120W isn’t quite enough for a
system with a decent (read: power-hungry) graphics card. Its limited cable length
and connectors may also cause difficulties. But these are small gripes about
what is otherwise an excellent power supply.

With the right power brick, the picoPSU has proved itself to be the most efficient
power supply that you can buy right now. This isn’t just a couple of percent at a power
level that your system will never even reach. The benefits of the efficiency
were seen in an actual system against a known good contender.
The net decrease in power draw of 20% at idle is amazing considering that the
load did not change.

Of course, with the right power brick is a key part of the efficiency
results. +12V power bricks that can output 120W capacity are hard to come by,
and their efficiency is virtually unknown. Until (and if) detailed information
about other models on the market can be found, it’s probably safest to stick
with the EDac bricks that we tested. Unfortunately, the 120W brick has its own
drawback: The cooling fan that comes on at 90W load. Hopefully, there are other 12VDC power bricks
of this capacity that don’t have a cooling fan.

Andrei Bulucea, the creator of the picoPSU, says that sales are about evenly split between companies and individual end users. The companies that buy the picoATX use it for specialized commercial and industrial applications where miniaturization and low power consumption are absolutely critical — digital display signs, PoS systems, digital video & sSecurity, and so on. Individual end users are generally hobbyists (like many SPCR forum readers) who use the picoPSU for special projects such as tiny mini-ITX PCs. We think it could quickly become a SPCR modder’s weapon of choice.

All in all, the picoPSU is inventive, practical, and even affordable. (Why
do those things so rarely go together?) In time, we hope to see many similar
products — or perhaps an industry-wide shift to using single-rail power
supplies.

* * *

Much thanks to Mini-box.com
for the opportunity to examine this power supply.

*

SPCR Articles of Related Interest:
Power Supply Fundamentals & Recommended
Units

Power Distribution within Six PCs
Seasonic SS-400HT power supply, 80 Plus version

Seasonic SS-300SFD 80 Plus: Little Big PSU

* * *

Discuss
this article in the SPCR Forums.

Leave a Comment

Your email address will not be published. Required fields are marked *