Seasonic Platinum Fanless 520W PSU

The title says it all: A fanless 80 PLUS Platinum 92% efficient power supply from Seasonic, the company which pioneered super-quiet power supplies. Can this product meet the expectations of its eagerly awaiting fans? ADDENDUM, 5 Feb 2013: The Third Sample.

This Seasonic fanless power supply became inevitable when the 80 PLUS Platinum
efficiency class was created. The Seasonic 400W
and 460W X-series models were the first fanless PSUs from Seasonic, released
two years ago in Q4 of 2010. Those 90% efficient PSUs swept all other fanless
models in every measure of performance. The only serious competition was the
fanless Platinum class Kingwin
Stryker STR-500, which appeared almost immediately after the X-series fanless
and surpassed the Seasonic X series in efficiency. With greater brand presence
and probably better global distribution, the Seasonic X-series models have become
the defacto choice for fanless PSUs, but Kingwin had already thrown down the
gauntlet with the Stryker STR-500.

The Platinum Series Fanless 520W is Seasonic’s answer to that challenge, albeit
a long time in the making. Seasonic has had over two years of continuous production
experience with the extensive 80 PLUS Gold X-series to bring to their new line,
so consumer expectations of the new Platinum models has got to be high. The
first model was the 1000W, announced with the 860W; 660W and 760W models are
coming, but the 520W fanless is actually the second Seasonic Platinum to make
it to market. The Platinum Fanless 520W is a natural evolution of the X-400
and X-460 which preceded it. It shares the same casing, completely modular cables,
and similar internal layout. New 460W and 400W Platinum models will displace
the earier X-series models. (Note: Most these products are already selling at
Newegg at time of posting, Dec 11.)

Long time SPCR readers need little introduction to Seasonic. The core story
of this Taiwanese company is that it started as an audio hifi amplifier company
in the 1970s, then expanded into making switching power supplies mostly for
Apple computers, became a well-respected OEM/ODM PSU manufacturer along the
way, and about a decade ago, began marketing and selling PSUs under its own
brand name as well. They attracted SPCR’s attention with a then-unique approach
to PSU cooling, which kept the internal fan running at very slow speed except
at higher loads. Seasonic PSUs were the quietest out-of-the-box when I first
started reviewing them, and they have always been at the top of our Recommended
PSU list.

The retail package is the same size as the X series fanless, and its graphics are in lighter shades, but still classy as per Seasonic norm. If you have a chance to handle the box in a store, you’ll find a massive amount of data and information all over it. It is obviously designed to persuade you to to purchase; too bad so few of us actually buy components in a brick-and-mortar store these days. Well packed in shock-resistant closed cell foam, with cables in a double-section pouch and the PSU in a velour draw-string bag. Here’s all the contents of the box. Note that except for the main ATX cable, all the output cables are of the flat data-cable style. The prominent caution label about PSU installation position used in the X-series fanless models has been retained.

DETAILS

The casing is unchanged from the X-series fanless models, as far as I could
tell. Like all the X-series PSUs, instead of the conventional double U-shaped
parts that clamshell together, the main casing is a 5-sided box, to which the
top perforated cover is secured with four small screws. It remains about the
most mechanically robust PSU casing I’ve come across. There are hex-hole vents
on all sides except the bottom. The top and back panels are almost entirely
vented.

Technically, Seasonic says the 520W uses full-bridge topology on the primary
and an LLC resonant controller for increased efficiency. The previous Gold-level
X-400 and X-460 used a half-bridge topology. The MOSFETs for +12V regulation
have also been moved from the trace side of the primary PCB to a secondary plug-in
board for better cooling. The secondary side is said to be a synchronous design.
The DC-DC converters are on the modular daughter PCB with the output connectors
solder directly on it. This improves efficiency and eliminates losses to connecting
wires. This aspect is unchanged from the X-series.

A completely modular cable setup. Inside, compared to the X-400 and X-460 fanless models, the 520W’s heatsinks have been beefed up, though they are still small compared to earlier, lower efficiency PSUs. Seasonic’s usual high standards of workmanship, layout and parts quality apply. As with the X series, the main PCB contains the circuitry to filter and convert the AC voltage to 12VDC, and the conversion from 12VDC to the lower DC lines is done in the PCB on the output panel, to reduce internal soldering/wiring and shorten the paths for lower voltages for minimal loss. Seasonic compares its patented output board with integrated low voltage line VRM against traditional modular design.

OUTPUT CABLES

As mentioned earlier, only the main ATX cable is "normal", nicely
sleeved. All the other output cables are flat data-style. This is the first
time such cables have appeared in a Seasonic PSU, although many products made
by Seasonic for other brands (such as Corsair and bequiet!) have featured them
for years. My guess is that there is some small savings in labor cost; the flat
cables do not require sleeving.

1 – Main ATX (24/20 pin), 60cm
1 – EPS12V/ATX12V (8/4 pin), 65cm
2 – PCIe (6/8 pin) connector, 60cm
1 – 4 SATA connectors, 70cm
1 – 2 SATA connectors, 40cm
1 – 3 4-pin peripheral connectors, 60cm
1 – 2 4-pin peripheral connectors, 40cm
1 – Floppy drive Y-adapter

The inclusion of only one pair of PCIe connectors seems indicative of
Seasonic’s typical conservatism. With improved energy efficiency in the latest
generation of GPUs, 520W will easily power a modern PC with a couple of mid/mid-high
performance cards. A pair of nVidia GTX670s, for example will pull a maximum
peak of 340W together. Current Intel socket 1155 CPUs don’t go much past 65W
in real apps, so the maximum peak demand for an Intel CPU SLI system using
the GTX670 would not run higher than 450W. Eager users can probably buy additional
cables with dual PCIe connectors, though.

TESTING

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

SPCR’s
PSU Test Platform V4.1. is the basic setup for the testing. It is a close simulation of
a moderate airflow mid-tower PC optimized for low noise. There is one major change: The primary testing is done with the PSU NOT inside the hotbox but atop it, out of the heat path. This is in recognition of several realities that prevail today:

• In SPCR’s earlier test platforms, the internal temperature varied proportionately
  with output load. The tested PSU was subject to this heat, and operating ambient
  temperature rose with increased load, reaching >40°C and often much
  higher at full power. This was a realistic simulation of a mid-tower PC case
  where the PSU is mounted conventionally at the top back portion of the case.
• The vast majority of "serious" PC cases for the home builder place no longer position the PSU at the top back corner. They put the PSU at the bottom/back corner, mostly out of the path of heat from the other components in the case. This design concept took root with the Antec P180 going back over 5 years, and dominates the DIY case arena. This means the PSU generally has to dissipate only its own heat.

Now, we’ve reversed our approach: The PSU is tested briefly in
the hotbox only to check what happens to noise, fan speed and temperatures when
it is used in an outmoded case design.

Acoustic measurements are performed in our own anechoic chamber with ambient level of 11 dBA or lower, with a PC-based spectrum analyzer comprised of SpectraPLUS software with ACO Pacific microphone and M-Audio digital audio interfaces.

REAL SYSTEM POWER NEEDS: While we test the PSU to full
output in order to verify the manufacturer’s claims, real desktop PCs simply
do not require anywhere near this level of power. The most pertinent range of
DC output power is between about 40W and 300W, because it is the power range
where most systems will be working most of the time. It is true that very elaborate
systems with the most power hungry dual video cards today might draw as much
as another 150~300W, but the total usually remains under 600W.

TEST RESULTS

The ambient temperature was ~22°, and the ambient noise level
was ~10 dBA.

1. EFFICIENCY — This is a measure of AC-to-DC
conversion efficiency. The ATX12V Power Supply Design Guide recommends 80% efficiency
or better at all output power loads. 80% efficiency means that to deliver 80W
DC output, a PSU draws 100W AC input, and 20W is lost as heat within the PSU.
Higher efficiency is preferred for reduced energy consumption and cooler operation.
It allows reduced cooling airflow, which translates to lower noise. The 80 PLUS
Platinum standard requires 90% efficiency at 20% load, 92% efficiency at 50%
of rated load, and 89% efficiency at full rated load.

At the super low 20W load, efficiency was already 72%, and it
rose quickly to 82% at just 40W. At 50% load, it easily exceeded the required
92% efficiency, and was still over 91% at full load.

There was no issue with crossloading. With over 90% of a 498W
load on 12V, naturally, efficiency improved over the standard loading.

2. VOLTAGE REGULATION refers to how stable the output voltages
are under various load conditions. The ATX12V Power Supply Design Guide calls
for the +12, +5V and +3.3V lines to be maintained within ±5%.

The critical 12V line was very tightly regulated. It started a
touch high at very low load, 12.2V (+1.6%) and dropped gradually as load was
increased, reaching a low of 12.13V (+1.1%) at full power. The 5V line was held
incredibly stable, starting at 5.04V (+0.8%) and dropping to 5.03V (+0.6%) at
full load. The 3.3V line simply did not budge from its initial reading of 3.34V
throughout the tests (+1.2%). This is the best voltage regulation I’ve encountered.

3. AC RIPPLE refers to unwanted "noise"
artifacts in the DC output of a switching power supply. It’s usually very high
in frequency (in the order of 100s of kHz). The peak-to-peak value is measured.
The ATX12V Guide allows up to 120mV (peak-to-peak) of AC ripple on the +12V
line and 50mV on the +5V and +3.3V lines. Ripple on all the lines was excellent
at all power levels, staying under 13mV on 12V at all power levels up to around
250W. Surprisingly, the highest ripple of 23mV was seen at mid-power, not at
full load where it dropped down nearly as low as the <250W range. This is
excellent.

4. POWER FACTOR is ideal when it measures 1.0. In the most
practical sense, PF is a measure of how "difficult" it is for the
electric utility to deliver the AC power into your power supply. High PF reduces
the AC current draw, which reduces stress on the electric wiring in your home
(and elsewhere up the line). It also means you can do with a smaller, cheaper
UPS backup; they are priced according to their VA (volt-ampere) rating. Power
factor, as with all Seasonics for years, is excellent, running 1.0 through most
of the loads.

5. LOW LOAD TESTING revealed no problems starting at very
low loads. Our sample had no issue starting up with no load, either, and the
power draw of 7.9W was very low. The 0.3W power draw in standby (power switch
on but computer off) is the best measured.

6. LOW & 240 VAC PERFORMANCE

The power supply was set to 428W load at various AC input voltages.
Most full-range input power supplies achieve 2~3% higher efficiency with 220~240
VAC, compared to 110~120 VAC. SPCR’s lab is equipped with a 240 VAC line, which
is used to check power supply efficiency for the benefit of those who live in
higher mains voltage regions. We also used a hefty variac to check the stability
of the PSU under brownout conditions where the AC line voltage drops from the
120V norm.

Efficiency improved to 95% at 240VAC. This is a new high; that
level of efficiency has never been reached by any PSU in 10 years of SPCR testing.
The sample passed the 100VAC minimum input load without any issues, with a 0.9%
drop in efficiency. Neither voltage regulation nor ripple changed appreciably
during these tests.

7. COOLING

The temperatures marked in the main test results table need some
explanation. Normally, for a fan-cooled PSU, a thermal sensor is moved around
a bit during warmup at ~200W load to find the "hot spot" on the exhaust
grill where is attached using friction. This provides consistent enough readings
of the exhaust air temperature. With a fanless PSU, there is no force airflow,
so where should the sensor go? For the Seasonic Platinum 520W, it was placed
on the grill atop what appeared to be one of the hotter spots in the PSU, directly
over the large transformer. But this data is probably not very useful. To be
honest, without an array of thermal sensors which can be placed at many spots
all around the interior, monitoring the temperature of a fanless PSU is futile.
In any case, the unit did not suffer any temperature overload shutdown during
the testing.

As mentioned earlier, the Seasonic X-400 fanless PSU, with much
smaller heatsinks, sailed unscathed through a 15-hour
torture test at full load in a 53~57°C hot box two years ago. I did
not run this sample in a test like that, but in normal use, the 520 Platinum
should hold up fine even in hot weather.

8. NOISE, PART 1

There was some trace of tonal noise at level too low for it to
be registered in the anechoic chamber with the microphone a meter away. It was
more than a single frequency peak, there being a bit of lower frequency buzz,
albeit very low in level, as well as higher frequency whine that was a bit more
audible. The main source of noise was the 5V Standby line. Placing a 0.4A load
on that line always provoked a 4.7 kHz tonal peak that proved to be audible
from about 2′ away in the anechoic chamber.

Outside the chamber, in a completely undamped and therefore reverberant quiet room (17~18 dBA), the audible distance was about the same; at a meter I could not hear it, but at 2′ I could. Changing the 5Vsb load to 1A made the
tone disappear altogether, but dropping it to 0.1A or 0.2A changed the frequency,
to around 6 KHz; this was discovered using the mic at 1′ distance with the spectrum
analyzer.

The 4.7 kHz whine, traced to 0.4A load on the 5Vsb line. Note the close approximately one foot distance of the microphone from the PSU.

The question is whether this noise is…

Atypical or typical? This is one the hardest questions for any product
reviewer to deal with: Whether the characteristic of the sample is analomous
or normal. If the items are inexpensive, like fans, then getting even a dozen
samples to check out is not difficult, and testing that many samples helps
answer the question of consistency and what is likely to be typical. Mind
you, when goods are mass produced in the thousands, 12 is still a paltry sampling
(as any statistician would tell you). But when it’s a high ticket item in
short supply, even a dozen samples are impossible. I asked Seasonic to send
me whatever other samples they can spare. They say they will try and get me
one in the next 10 days or so.

Audible in a PC in normal use? I have assume that few users would
use a fanless PSU in PC with lots of other noisy components. They’d use it
in a quiet PC. So would this level of tonal noise be audible in a PC with
broadband noise of less than 20 [email protected], in a relatively quiet room? My guess
is… no for most users most of the time, but yes for some users, some of
the time. This is assuming a fairly select, aurally sensitive group of users
— which fanless PC components always attract. The frequency of the noise
is not high enough to be affected by the age-related decline in high frequency
perception, either.

There is no simple answer to these questions. I need to examine more samples,
and hear back from end-users in the field. At time of writing, the first small
shipment of Seasonic Platinum 520 PSUs were already sold out at Newegg. Some
SPCR readers must be among them; please tell us about your experience on the
discussion thread for this article.

Discussions with Seasonic

I discussed my findings with Seasonic, and they immediately arranged to have
the sample flown back to their Taiwan headquarters overnight for their engineering
team to examine. The acoustic phenomenon I described had not been enountered
before. All the early samples for reviewers had been double-checked for normal
operation before being shipped. They needed to find out what was different with
this unit and if had changed, how and in what way. In the meanwhile, they asked
me the favor of not publishing my review until they could at least get a handle
on the problem. My review would probably have been the first out, and I appreciated
that a negative one would have hurt, so I chose to be helpful. I wanted them
to solve the problem anyway, it was certainly in the best interest of the Silent
Computing Community. That was in early November.

9. NOISE, PART 2

Here we are a month later with sample #2 which just arrived late Friday. The
feedback from Seasonic on the problem with my original sample was somewhat murky.
It might be a language issue, as Seasonic’s engineering team does not have much
English, and I have no Mandarin. There seemed to be two primary messages:

• First, the acoustic noise was caused mostly by new components used in achieving
  the higher efficiency, particularly for the <1W standby that’s stipulated
  by the European Union ErP Directive (2009/125/EC) to reduce phantom power
  consumption. It’s not clear whether Seasonic has developed a solution or whether
  my first sample was just particularly bad.
• Secondly, they have a tough time measuring or hearing this noise because
  it is at such a low level, and Taipei is a noisy place. I did visit Seasonic
  headquarters in one of my trips to Computex, and while it is not a particularly
  noisy building, it is in Taipei, on the Northern side, and nowhere in Taipei
  is very quiet. There are no homes with the kind of quiet I have in my house,
  for example, except in Gigabyte’s underground anechoic chamber. Hmmm… perhaps
  Seasonic should knock at Gigabyte’s chamber door?

So in the end, the news from Seasonic engineering on the issue of the first
sample noise is unclear and inconclusive. I will assume, for the time being,
that it was an anomaly, Seasonic is aware there many be other samples with the
same issue, and they are working on resolving the issue (if there is one which
makes some samples noisier than others). At least I have another sample to consider.

So what of the second sample?

The first thing I did was to examine it physically. Check: No change from the
previous sample, and no, it is not the same one.

Secondly, I hooked it up on the power load tester in the anechoic chamber and
powered it up with a ~120W load, a mix of the various lines, including the 5Vsb,
of course.

My intial observations, a foot away from the PSU:

• No high pitched whine. Good!
• But wait, there is a trace of a lower pitched tone. Not really a whine,
  that term suggests something higher pitched. This is more in the midband,
  more like a buzz. Move back a couple feet and it’s gone. Move back close,
  and it’s still there. Change the load settings, and there are some very subtle
  changes in pitch, affected most by the load on the 5Vsb line. Power up the
  acoustic measurements computer, calibrate the mic and take some quick measurements.
  From a meter, nothing registers above the normal ambient. Move the mic to
  1′ from the PSSU, and a sharp spike shows up in the spectrum. See it captured
  below.

There are two tonal peaks in the second Seasonc 520W Platinum sample: One at just under 500Hz, another smaller one at about 1.3kHz. Note the level of each — about -7 dBA, and -10 dBA.

• Some things to note about this spectrum capture:
  • The ambient noise was higher because it’s been poring hard all day in
    Vancouver, and the drain pipes make a bit of a racket when it’s raining
    this hard. No way to damp that out. Hence the blue line is higher than
    usual; the red ambient reference was captured on a quiet evening.
  • The two spikes are in the midband, which might make them more audible,
    but they are also much lower in level than the 4.7kHz tone of the original,
    which was at about +1 dBA with the mic at the same distance. These are
    -8 and -11 dB lower in level.
  • After an hour running at the same load setting, the noise did not change.
  • Outside the chamber, the audibility of the buzz was limited to perhaps
    1.5′, at least for me, in this environment (and yes, with the slightly
    higher ambient cause by the rain’s pitter patter.)

In the end, I have to conclude that if this second sample is more representative,
Seasonic doesn’t really have a problem, although a small handful of highly sensitive
people in the world might disagree. Do I think the first sample is an anomaly?
Yes. It was much more readily audible and annoying than this one is, and no
one in their right mind at Seasonic would have sent me or anyone related to
SPCR a product with extraneous noise.

Spot checks of performance at various load points was done on the second sample,
and the results were within 1W (or better than 1%) of the readings on the first;
i.e., within my test equipment’s margin or error. The amazing voltage regulation
was the same.

COMPARISONS

The comparison table below shows the SPL versus Power Load data on PSUs tested
in ambient room temperature, typically 20~24°C. By SPL at 1m, it tops the table;
the question whether the tonal noise in my sample is typical.

The comparison table below shows the SPL versus Power Load data on all the
PSUs tested in the hotbox. It’s difficult to rank them, as the measured SPL
varies differently with power load. The units which are quietest at minimum
load are not always the quietest at midload (100W~300W), which may make them
louder in actual use. Then there’s the noise level at 400W and up, which will
determine the quietest PSUs for high power gaming rigs, during actual gaming.
Again the Seasonic 520 Platinum tops the table.

Caution: Please keep in mind that
the data in the above table is specific to the conditions of our test setup.
Change the cooling configuration, the ambient temperature and any number of
other factors, and you could change the point at which the fans start speeding
up, as well as the rate of the rise in speed. The baseline SPL is accurate,
however, probably to within 1 dBA.

CONCLUSIONS

The Seasonic Platinum Fanless 520W delivers the most exacting
performance ever measured on the SPCR PSU testing rig. Its voltage regulation
is astonishing, especially on the 5V and 3.3V lines, which remained essentially
unchanged from minimal 20W load to maximum 520W load. The 12V line, with a drop
of 0.07V between minumum and maximum load, is merely excellent, but also unmatched
by any other PSU tested. Noise and ripple were also superbly low, with the oddity
of it being slightly higher at mid-power than maximum. Efficiency was the highest
I’ve measured in any PSU thus far. The 95% efficiency seen at 400W load with
240VAC input means this PSU produces no more heat (21W) at that power level
than many a low power optimized mobile CPU. The luxuriant packaging, excellent
build and parts quality, 7-year warranty and Seasonic’s long history of quality
products all combine to affirm my own high expections of this remarkable product.

The fly in the ointment is that trace amount of high frequency
whine in the first sample at certain loads on the 5V Standby line. Too low to
be measured at 1m distance in an 11 dBA ambient anechoic chamber, but probably
audible to users with highly sensitive hearing in very quiet rooms who want
to place a small, near-silent PC on the desktop. This is the single biggest
challenge for makers of fanless PSUs: The absence of masking broadband noise
from a fan lays bare any trace of any other noise, while attracting the most
aurally sensitive customers. It leaves Seasonic between a rock and a hard place.

The second sample did exhibit some trace tonal buzz, but at a
much lower level than the whine of the original, and otherwise performed just
as beautifully as the first. So while there’s a bit of a question about which
sample is exhibiting the more normal behavior, the latter is more likely, and
this trace level of buzz is likely to be insignificant for the vast majority
of users.

It is not possible to bar the Seasonic Platinum 520W from an Editor’s
Choice award, given all the other strengths of this outstanding power supply,
but I will keep attuned to the web for genuine reports of buzzing or whining
from actual users. To end on a positive note, the $149 suggested retail price
is really a pittance for such outstanding performance, which was not available
at any price a few short years ago.

* * *

Much thanks to Seasonic
USA for the review samples.

Seasonic Platinum Fanless 520W wins the SPCR’s Editor’s Choice Award

* * *

SPCR Articles of Related Interest:
Power Supply Fundamentals

Recommended Power Supplies
SPCR PSU Test Rig
Fanless PSU Torture
Test Roundup
Fanless
PSUs: Kingwin Stryker STR-500 & Silverstone ST50NF
Seasonic X-400 Fanless
PSU
Enermax Platimax 600W
Seasonic G360

* * *

Discuss
this article in the SPCR Forums.