Monday, March 27, 2006

PW200M installed

The PW200M DC/DC adaptor found it's place inside my Antec Aria Small Form Factor case. The Aria was designed to be a silent case with multi-layered noise-dampening side panels. This case was reviewed in detail a while back at SilentPCReview. Unfortunately the Aria's PSU fan gets too loud preventing it from being a very quiet system, even under moderate load. This makes it a perfect candidate to see what we can do by swapping the noisy PSU for a small and silent PW200M.

The nice large 120mm fan seen at the back of the case seems to suggest good airflow, but unfortunately the air must be drawn through the PSU from the cramped interior making it less effective.

Here you can see the PSU from the inside (right side of the picture) with a small row of intake grills where the air must pass through before leaving the case. There are no intake grills directly opposite the fan (which would make for a more direct airflow), but only on the four sides (top, bottom, left, and right).
Here's a top view with the optical and hard drive cage removed.

Let's compare the Aria's PSU to the PW200M. You can see a significant size reduction.

The layout of the ASUS A8NVM-CSM motherboard, had the RAM slots right next to ATX power plug, so an extension cable was used to install the PW200M up away from the board. This was actually an unused 24-20 pin convertor that came with a Seasonic power supply.

It installed easily and without issue. Here the PW200M is suspended only by the wires, but just for testing. I made sure it was not in contact with anything before powering on. I also added a 120mm fan where the old PSU was to provide some case airflow.

No problems, everything powered up on the first try and booted into Windows. I checked the PW200M expecting it to be hot like when I first tested in an Athlon XP system, but I was surprised it was only slightly warm. The PW200M was not even in the path of the 120mm fan.

I decided to fix the PW200M towards the rear of the case in the spot of the old PSU. There was the most space here, plus I figured the extra airflow from the 120mm fan wouldn't hurt. This required a longer ATX extension cable, which I had previously purchased at a computer accessories store. I made sure that the PW200M was secure and did not contact any other areas of the case.

This was done with some handy (recycled) twist-ties. The tie on the right is securing the ATX connector of the PW200M to the chassis, and the tie on the left just to keep the molex wires out of the way from the 120mm fan.

Here you can see the difference in space compare with the original Aria PSU on the left/top photo.

I took several readings at different CPU multiplier and Vcore settings (set using CrystalCPUID). All temperatures were measured by Speedfan in Windows, and the AC draw was measured as total system power consumption from the AC outlet (V x A) using a clamp meter. I also included some readings taken with an A64 3500+ CPU which I originally had in the Aria. Other components used were:
- Stock retail AMD CPU heatsink and fan (controlled by QFan)
- ASUS A8NVM-CSM motherboard (Nforce 6150 onboard VGA)
- 2 x 512MB Crucial Ballistix Tracers DDR400 RAM
- 250GB Hitachi T7K250 HDD

The Opteron 144 CPU was run at 9x multiplier and 1.0v Vcore and loaded under Prime95 for over 24 hours with each PSU. The temperatures (highlighted in blue) show that with the PW200M setup, all temperatures ran 2-4C cooler. The 120mm fan used is a CoolerMaster fan turned (nearly all the way) down to less than 700rpm. Although this was not a quiet fan, at such low revolutions it was significantly quieter than the Aria PSU; the noise of the HDD spin and Qfan controlled AMD heatsink now became apparent.

** All temperatures above were recorded with motherboard BIOS version 0601 which seems to overstate temperatures. Upgrading to a newer 0702 BIOS drops the CPU and NB temperature readings by 8C; SYS and HDD remain unchanged.

* A small side note, a comparsion can be seen between the A64 3500+ and Opteron 144 being run at the same 9x multiplier and 1.2v Vcore (highlighted in yellow).Despite the Opteron's larger cache, it draws around 5W less power than the A64. The CPU temps are similar, but other system temperatures are mixed (one higher, one lower), possibly the rearrangement of wiring when switching the CPUs had this effect.

The overall system power consumption was also reduced using the PW200M. Here is a comparison across various states and settings.

Under CPU load, the difference is 9-11W or 11-12%. Under idle, 11-12W or 17-21%. Even when powered off, there is a 6W reduction.

Even after 24+ hours of Prime95, the PW200M did not feel hot, but only slightly warm to the touch. I am not sure why it was significantly hotter when I first tested it in an Athlon XP system. Maybe because in that system it was flush on the motherboard, or possibly the different current draw on the various voltage rails had some effect?
As for the external AC/DC adaptor, this did get warm, warmer than the lower powered AC/DC brick of my ASUS Pundit, but not hot. This however is a good sign showing that much of the heat is being dissipated externally through the AC/DC brick and not inside the Aria.

Unfortunately my inaudible Pundit has really spoiled my ears, and although the Aria might now be considered a quiet system, it's still too loud in a very quiet room.
What the PW200M has done however, is remove the significant problem of a loud PSU, reduced the amount of heat in the case, improved case airflow, and made room for a larger CPU heatsink. This now allows further steps to be taken to make the Aria even more quiet.

Also I still plan to add a high powered Geforce 7800GT graphics card to the system. With an additional AC draw of less than 80W, this should still be well within the PW200M's capacity, however whether the improved airflow will allow such a card to be run quietly in the Aria remains a question.

Next PW200M and a 7800GT

Back to Pico PSU and PW200M DC-to-DC power supplies
Read more!

Thursday, March 23, 2006


Here is my attempt to organize and index my posts so they can be easily accessed from one page. This page will be updated as I put up more posts.

Small and Quiet PCs

ASUS Pundit P2-AE2 slim barebone

APPLE Mac Mini

Silent DC-to-DC Power Supply Solutions

- Inside an Antec Aria SFF Case

- Inside an AOpen XC Cube EZ18 SFF Barebone

Small and Quiet Prebuilt Systems

Touch Screens

Budget Systems



Read more!

Tuesday, March 21, 2006

Booting WinXP on the Intel Mac Mini

Booting Windows XP on the new Intel Macs, it was only a matter of time. Thanks to Blanka and Narf whose hard work has earned them the cash prize from Colin Nederkoorn' contest to be the first to provide such a working and replicable solution. The step-by-step guide has been posted at as well as on their wiki page. The instructions are clear and straightforward, however not all issues are resolved and certainly there is a lot more to be tested; also your success may depend on what version of the XP install disc you're using as not everyone has been able to get XP running. Here are my experiences trying to install and boot Windows XP on my Core Solo Mini...

Thanks to an iMac discussion on SPCR, forum poster Erssa kindly pointed out the success story and provided a link to I skimmed through the instructions briefly last night, and it seemed clear and complete enough to follow. At first I was a little hesitant to attempt this on my new Mini; I was enjoying the comfort of FrontRow too much and the thought of something going wrong plus this was still a beta solution told me I should wait. Well curiousity got the cat, and to me the thought of being able to run Win XP was too interesting to not have a go.

I won't go through any details as again the step-by-step instructions at the above links are very clear, however I did have to do some things differently so I will mention them here. I will also try to give an idea of how much time it took for those wanting to attempt the same.

The first step involved creating the installation CD for the Mac by patching your original Windows XP SP2 installation disk. I used a Volume License English version. It seems that your success may depend on what version (Retail/OEM, English/Non-English) as the instruction and patch files may need some adjustments accordingly.

As it was getting late, I only had time to create the CD that night. This part was so quick and easy, and led me to believe the rest of the installation would be the same. I was planning to take the morning off and stay home to finish the install, but something came up at work. Fortunately the Mini is so small and portable, it was a no brainer for me to bring it with me. Well it turns out that the rest of the install wasn't so quick and easy as I thought, needless to say I wasn't very productive work-wise today... Also unfortunately I did not have a camera with me, so you'll just have to bear with my text-only recount of the installation.

As XP needs to be installed in a seperate partition, the next step involved partitioning the hard disk. With OS X's disk utility, this meant erasing the HDD first, wiping out the previous OS X installation. It is possible some other third party partitioning tool will allow you to do so without erasing, but this has not been tested with this solution. Also booting XP from an external disk might be possible, but again no working solution yet that I am aware of. Well how long can a partition and OS X reinstall take...

1) Backup HDD to external HDD using Carbon Copy Cloner (optional)
This is not necessary, however I wanted to do this just in case something got messed up, I knew I could still get back into OS X. Afterall, I have less than 20G, how long could a backup take. I had an external firewire (prolific chip) HDD enclosure with a 40G 2.5" 5400rpm Seagate Momentus HDD (actually it was the drive I took out of my old G4 Mini). I'm not sure if it's because CCC runs under Rosetta, or maybe some indexing feature of OS X, but it took more than 45 minutes to complete which was slower than I expected.

2) Booting OS X from External HDD (optional)
Again not necessary, but just to confirm my backup works. In the System Preferences under the Startup Disk, simply select the disk to boot from. There have been some users having trouble with this, however it seems it's dependant on the type of firewire interface. Fortunately mine worked without any problems. It was quite slow though, over 1 minute to boot up compared with 25 seconds for the internal disk.

3) Erase and Partition the HDD
I split my HDD:
----Partition 1 = XP = 30G = MS-DOS
----Partition 2 = OS X = 25G = Mac OSX Ext (Journaled)

I decided to make the XP partition a little larger, figuring I have more XP software than OS X, plus OS X can read off the XP partition, but not vice-versa.

4) Reinstall OS X from the installation DVDs
Ooh boy, I remember I waited quite a while reinstalling OS X on my old G4 Mini, but didn't realize OS X install is really so slow...

First the installation utility does a check of the DVD. I assume this is not essential and there is even a SKIP button. I should have opted to skip, but wanting to be sure every i was dotted, I waited the 20 minutes for it to complete. Then the rest of the install (Disc 1 and 2) took nearly another hour. I suspect in the install screen there is probably some option to exclude programs you don't want installed and that might shorten the time somewhat. In total, my preparation took 2-1/2 hours, not an extremely long time and I did get some work done in between prompts, but certainly longer than I expected.

5) Finally, the XP install
Ok first the XOM.EFI file has to be copied to the OS X home directory and blessed, per the instructions. This nicely gives you a bootup menu to select between a crystal Apple Logo and a multi-color Windows Logo (I snapped these photos after I got my Mini home). In went the patched XP install CD I burned the night before...

A screen of text messages (white on black) from the XOM bootloader appeared while the optical drive chugged away. I was getting excited, especially after the several hours of prep work. Unforunately I didn't get past this screen and eventually the CD spun down and nothing happened.
Checking back on the onmac forums I found that a new XOM.EFI file had been posted specifically for the Mini with mention to solving freeze problem. Great, back in business. Unfortunately the same result again. Checking back on the forums again yielded different people trying different things with mixed success. This included typing g, ENTER, g, ENTER, or pressing F6, pressing ENTER, etc. What finally worked for me, thanks to some very useful advice from one of the posters (sorry can't find the thread to give proper credit) was holidng down the F6 key as soon as the text screen appeared (I actually started holding it down just before), and then to keep holding until the drive stops spinning. This time there was an additional line with "patching successful" or similar phrase. Then pressing ENTER after releasing the F6 key caused the CD to spin up again and successfully continue into Windows XP disk partition screen!!

For those of you having trouble, there might be some helpful tips that apply to your situation in these forums and

I selected Fat32 instead of NTFS as that way I could both read and write to my Windows partition while in OS X. I also opted for the slow format just to be sure which meant an extra 15 minute wait, but again just to be extra sure.

I did have the resolution set to 1280x1024 in OS X prior to booting the XP install. It seems this helps keep the resolution correct for the XOM.EFI and also XP install screens, some users who didn't set this first had problems. Consequently this also meant all the XP install screens looked beautiful in hi-res and not the crummy low-res (800x600?) that we usually find on PC installs.

With each passing step of the XP install, my heart rate climbed. I was really quite excited and anxious at this point, and things were moving along smoothly!! I had to record this, my super low-res phone cam would have to do. Unfortunately it got further clipped during the transfer to my PC, but you get the idea...

6) XP Installation continued
XP install successfully completed it's first part, and continued after rebooting from the HDD! In the instructions, it mentions install hanging at the end and needing to manually power down the Mac, but in my case with the Mini, XP completed installation normally and rebooted itself. Well between trying different XOM.EFI files, searching on the forums, reading through posts, (also took time to search for my CD key as I forgot to bring my original CD with me), I lost track of how long the install actually took. However it was about 3 hours from the time I first put in the XP CD until the install finished, but the actual install itself probably took less than 40 minutes.

7) XP Boot
The moment of truth... (again photos taken after I got home)

Selecting Windows from the boot screen...

The same text screen seen during the install (maybe some lines have changed)... Windows is loading... it must be...

(ok this shot was taken after I had already booted XP several times and installed some stuff, but it was very exciting the first time)

Without any drivers installed, the boot time into XP is around 25 seconds (hey just like OS X!).

Here is a shot showing the XP partition from within OS X. OS X now recognizes the formerly named "XP" partition as "NO NAME", but I don't think I'll try to change it.

So What Now?
Well being a PC silencing enthusiast, let's see what some PC silencing tools can do.

Unfortunately v4.28 is unable to detect any temperature (other than HDD), fan speed, or voltage sensors. It does pick up FDC37n972 super i/o chip and has two speed controls, but changing their settings doesn't seem to have any effect.

CPUZ identifies the 65nm Yonah Intel Core and displays a nice Coro Solo inside logo. Also shows speed step (correct name?) automatically cycling the CPU from 6x (1Ghz) @1.004v to 9x (1.5Ghz) @1.404v under load.

CrystalCPUID seems to be working at least partially. The multiplier can be set between 6x and 9x and the respective frequency is verified in CPUZ. Voltage can also be set from 1.004v to 1.404v, also with the same being displayed in CPUZ, however I am not quite convinced that vcore is actually being changed. CPUZ is displaying the selected value verbatim and not a fluctuating reading normally found on AMD boards, casting doubt if this is the true vcore. Unfortunately without any voltage or temperature sensors, it is quite difficult to verify. Running maximum clock at minimum voltage 9x (1.5Ghz) @ 1.004v under Prime95 seems perfectly stable, which can be read either the chip undervolts perfectly at this level, or the vcore is not being set. There is a lot of hot air being blown out when running Prime95 but I could not notice any difference between the different voltages.

Here is a screenshot at idle with CPUZ, Speedfan, and CrystalCPUID open.

And under load with Prime95.

With Prime95 running, even with the lowest voltage setting there are no signs of instability (Prime was only run 30 minutes). Although a voltage lower than 1.004v can be selected in the listed options, CrystalCPUID doesn't let the set voltage go lower than 1.004v.

Well so far I don't notice any difference noise-wise between booting into XP or OS X. The fan seems to be operating the same in XP. Running Prime95 even after an hour, I don't notice the fan spinning up, but certainly the hot air coming out the back can be felt. The top of the case does not seem hot either. Possibly the fan is spinning up gradually, but it is hard to tell.

Despite people's different preference for XP or OS X, I think most people will appreciate the option to dual-boot.

This is just the first day with XP on my Mini so it's hard to say what issues will come up when I install more applications. It seems most of the drivers (lan, wifi, bluetooth, etc) are working, however people are having limited results notably with the video driver. This will probably be solved shortly, especially for the Mini since it uses Intel's own onboard GMA950 solution. Some minor problems I noticed so far (note that I have not installed any of the drivers, which may or may not correct this)

- occassionally the system boots into XP or OS X with power LED remaining off
- rebooting from OS X to XP with resolution lower than 1280x1024 causes XP to be improperly sized (you only see a corner of the XP desktop); rebooting from XP to OS X also seems to set resolution at 1024x768 even though it was set differently last time (this might be monitor related)
- XP shutdown, drive spins down and screen goes blue, but must power off manually
- Standy option is greyed out from within XP

These are only very minor points at this time, I'm sure there will be more interesting and exciting discoveries along the way...

Read more!

Friday, March 17, 2006

Pico PSU and PW200M DC-to-DC power supplies

DC-to-DC power supplies for PCs
Power supply solutions using a DC-to-DC adaptor have been around for a while, but traditionally they were used mainly in Mini-ITX cases and car PCs. I have been pondering the idea of using this in a SFF PC since a fellow SPCR forum member "lm" posted such an idea over a year ago. However articles, reviews, and even fellow user experiences remained scarce, and it wasn't clear how well they would work in a desktop system. On top of that they weren't readily available and cost more than good quality ATX power supplies.
However after my good experience with the ASUS Pundit's external power supply setup, allowing it to run virtually silent, I decided it was time to step up and give these DC/DC adaptors a try.

The Pico PSU and the PW200M are both DC/DC adaptors that come integrated on a small circuit board and can be plugged into practically any motherboard that utilizes a 20 pin ATX connector. These boards do need to be paired with an external "brick" AC/DC adaptor.

Let's have a quick look at these first...
Here is the Pico PSU. Yes that is the complete DC/DC adaptor. That's all that needs to go inside the PC.

It is indeed very tiny and the "Pico" name is well earned. It's not much larger than just the motherboard connector of a normal ATX power supply!!
And here is it's older brother, the PW200M. These are two seperate and independent DC/DC adaptors. The PW200M is slightly larger, but still very compact. We'll discuss their differences later.

In case the molex connectors don't give you a good enough idea about it's size, here they have their photo taken with a traditional ATX sized power supply.

So why would anyone want to install one of these inside their PC instead of a normal ATX power supply? Well of course we mean someone who wants a Small and Quiet PC...

1) Both the AC/DC as well as the DC/DC adaptors can be fanless, meaning they generate no noise.

2) In general such a setup will be more energy efficient especially at low power consumption levels where most silent-oriented PCs will tend to be. This also means the less energy that is wasted, the less heat that will be created. This should mean the system does not need as much cooling and fans can be run quieter.

3) The DC/DC conversion part should be more efficient compared with the AC/DC conversion, which also means most of the heat generated will be outside of the system again reducing the amount of cooling required inside the system. The claimed efficiency of the PicoPSU and PW200M is "greater than 95% at full load" whereas even the highest quality ATX PSU is rarely able to boast more than 85% (please note though, 95% only applies at full load and the AC/DC efficiency must also be considered when talking about overall efficiency).

4) With their tiny size, they only take up minimal space inside the PC. This becomes important for Small Form Factor PCs where space is already so tight and often airflow is not so good. Poor airflow will mean fans have to work harder. Removing the stock PSU should allow for much better exhausting of heat, keeping overall system cooler which again should allow it to run quieter.

So what's the catch, why aren't these setups more widely used? Well besides the ASUS Pundit P2-AE2 and the Mac Mini, there are actually other barebones/SFF that come with this type of setup. Shuttle has their Zen ST62K and more recent SD11G5, both very quiet SFF. Dell also runs their Optiplex ultra-small-form-factors like the SX260/270/280 or their current GX620 to name a few. ECS also did the same with their EZ-Buddie. However none of these manufacturers or other major brand are offering any DIY solutions you can apply to your home PC. Some minuses to consider.

1) Most people would choose to have their PC in one package and no external brick sitting on the floor.

2) Most of these DC/DC adaptors tend to be limited in wattage, 120W for the Pico and 200W for the PW200M. I am not aware of any larger DC/DC adaptor readily available. Whereas for ATX PSUs these are commonly in the 300-600W range, and some even going up to 1KW (in case you need to power all the PCs in your neighborhood :)
However what you can power with 120W or even 200W may surprise some of you; more on that later.

3) There is limited availability and costs tend to be higher. I was only able to find them in specialty Mini-ITX/Car PC retailers or otherwise resort to eBay. The Pico is generally priced around $45 and the PW200M can be found for $40. However adding in the AC/DC adaptor, depending on the wattage and where you get it from, could add another $35-80. Unlike lower wattage AC/DC bricks, high wattage bricks are a lot less common, especially ones that supply the 12v required by these two DC/DC boards. This compared to a high quality quiet ATX PSU like the Seasonic S12 330W which is sold for around $60.

4) Possibly other issues of electrical noise, durability?

Frankly for a regular sized PC, it's hard to justify the external PSU solution over a high quality quiet ATX PSU which can run very quietly with good airflow. For a SFF system however, the external PSU should have significant advantages... let's find out...


Here is a closeup of the Pico PSU. The black and white wires is the 12v DC input. The other set of four wires provides the 12v/5v to the P4, molex, and floppy power connectors.
The two yellow squares and the tiny grey one on the back are metallic casings, possibly for isolating electric noise assuming this is a switched mode convertor, or to serve as a heatsink for dissipating heat? Again any engineers reading this are welcome to enlighten us by adding your comments.
From the side we see that the Pico is made of two double-sided circuit boards sandwiched together. Some ICs can be seen in between.

Here is the back side of the PW200M where it has the same three square metallic casings. In addition we see a series of capacitors not found on the Pico (CapXon 390uf 35v and 1000uf 10v).

Here is the external brick AC/DC adaptor that I ordered with the Pico PSU. It's rated output is 12v / 8.5A which is 102W, but it claims a maximum 110W and was the largest wattage brick available from the store where I placed my order. Unfortunately it arrived defective. When I plugged it into the AC outlet, it immediately started ticking and the LED began pulsating. Measuring the output voltage showed a cyclic fluctuation from 10.2-12.2v.
The international freight cost was about the same as the cost of the adaptor so it really made no sense to return it for a replacement. I tried to see if it could be repaired locally, but unfortunately in our everything is disposable society, I couldn't find anywhere that repairs AC/DC adaptors. Manufacturing costs have come down so much that the replacement cost is less than the labor cost to repair it... My own attempts to repair it were thwarted as I couldn't even get the case open despite having removed all screws and trying to pry open the seams.

Well I'll say again, high wattage 12v adaptors are extremely hard to find. Replacement AC/DC adaptors are plentiful for notebooks, but these usually stay below 80W, and for the higher wattage adaptors that I did find, were usally more than 17-19v. In the end I found a second hand Lite-On brick rated at 12V 12.5A (150W) for a pricey $50. Anxious to test out my DC/DC adaptors, I bit my lip and took out my wallet.
Both DC/DC adaptors have a simple round 5mm DC input connector with a + and - connection. Unfortunately the LiteOn brick had a 4 pin din plug. I had no luck trying to find the pin out on the internet for this adaptor, but it seemed that some video equipment used a similar power connection. Then I read some AC/DC adaptors actually output two seperate rails, maybe that was the reason for 4 pins instead of just a simple two? This never occurred to me before, but if the 150W were split on two rails, then just taking one rail might leave me with 75W or less, which would not be enough power for the PC I was planning. So opening up the brick was in order, that plus I hadn't the slightest clue regarding the pin out.

Please remember opening up any electronic equipment especially power supplies can be dangerous even if it is unplugged there may still be residual charges. Having said that, the LiteOn brick opened up easily after the four screws were removed.

Fortunately my fear about seperate rails was unfounded; here we can see the red and white wires are soldered together to one connection, with the black, brown, and shielding wire soldered together to another connection.

The solder points can be seen better from the underside.
Ok let's put back together the brick and start splicing the wires.

Hurray a 12.2V reading! A quick check with the voltmeter reconfirms our understanding is correct. The red and white wires both correspond to +12v and the black, brown, shielding wires to ground (please be careful that other models/manufacturers could be wired/colored differently).
Off comes the old head and we splice in a round 5mm connector salvaged off some dead brick at the same second hand store.
No soldering yet, just some electricians tape for now so we can get to testing. Notice how much thicker the wire is coming from the LiteOn brick, I plan to purchase a connector and solder it on directly later, but for testing this should not be a problem.

Finally we get to hook up the Pico PSU with a 12V input! A green LED on the front lights up to let you know it's receiving power.

Using a paper clip to simulate the PC ON switch by shorting the ATX connector's PowerOn pin. A nice little red LED on the back let's you know you have powered on.

Without any load, checking the various rails all seemed quite reasonable. The -12v was a little loose, but probably still acceptable.

Pico PSU (12.2v, 5.0v, 3.3v, -13.3v)
PW200M (12.2v, 4.9v, 3.25v, -14.3v)

I was however getting a very odd reading when the DC/DC boards were in power off mode. The 12v rail was reading 3.2v-4.5v instead of the expected 0v. This had me concerned and a little hesitant to test them in a real system. Maybe this was only very minor current leaking through (as the 12v rail is actually passed in from the 12v input)? Well I did have an old system I was ready to give away and wouldn't cry (not too much anyways) if something happened to it. Specs:

Athlon XP-M 2600+
256MB DDR333 RAM
120GB Hitachi 7200RPM 3.5" HDD
GeForce Ti4200 VGA card

First let's check the system power draw using a regular ATX power supply. Unfortunately my Seasonic PSUs (which SPCR has tested efficiency curves) were not available and I had to use a Tagan TG380-U01. The maximum AC power draw measured was 150W. The Tagan's efficiency at load if I remember correctly is only 70%, and probably 65% or less at lower loads. That would mean the equivalent of around 100W DC draw. As the Pico is rated at 120W, this might be pushing it especially when the load balancing between the different voltage rails cannot be checked. So I decided to drop the processor speed to a safer lower consumption level. Initially it was set at 12.5x166 or 2.08Ghz, dropping it to 8x166 1.3ghz gave a more comfortable peak of 130W AC/90W DC (Just a note, despite the XPM mobile processor being able to run at very low voltages, the A7V333 motherboard I used did not allow voltages lower than 1.6v hence the higher power consumption).

Ok, let's start with the higher rated PW200M. It's that tiny green card located in the middle of the photo. It's probably clearer in the closeup.

There was no problem fitting this board directly onto the ATX power connector of the motherboard. However it should be noted that the board covers the release clip making it inaccesible for removing. Fortunately there is a tiny hole in the PW200M that allows a thin rod to be inserted down to pry open the release clip.

It was with a mixture of anxiety and excitement that I turned on the power button, but fortunately everything went smoothly and the system booted up without a hitch! This was only a first test, so nothing extensive was done. The 12/5/3.3v rails all seemed reasonable both via the voltmeter and motherboard sensors. In addition I was happy to confirm that once the PW200M was connected to motherboard, there was no more voltage measurable on the 12v rail when the system was off.

Next up was the Pico PSU. Here it is installed, it really is very tiny! Look Ma, no PSU!

No issues here either, the system seemed to run normally during the 20-30 minutes of my testing.

So far there didn't seem to be any stability issues, but of course the system was not run for too long. Both the Pico PSU and PW200M do get quite hot however. Within a minute of powering up, the square metallic casings get burning hot to touch, and after 20-30 minutes of use, the overall adaptor is too hot to grip for unplugging. As these are often used in fanless or low airflow Mini-ITX systems, I suspect it should not be an issue and the specs rate operating temperature up to 85C.
Overall power efficiency is better than the Tagan ATX power supply by 15-20% based on the measured AC draw. If our guesstimate of the 65% is correct for the Tagan, that would mean the DC/DC+AC/DC solutions are running at around 75-80% overall efficiency.


Well the next step is to put them into their intended small form factors for some real testing. How much will they improve a SFF based on noise, temperature, and energy efficiency? Also how much can these DC/DC boards really power, and will they endure such use?

The PicoPSU is intended for my slightly dated HTPC. It's an AOpen XC Cube EZ18, running an undervolted XPM 2200+ chip, 1G RAM, nforce2 board with onboard VGA, 2.5" HDD, and a combo-optical drive. The maximum AC draw was only 87W last time I checked, so the Pico should be able to handle it comfortably.

And for the PW200M, how about an Antec Aria SFF case with an Athlon 3500+ Socket 939 CPU, 1G RAM, Asus A8NVM-CSM Nforce 6150 motherboard with onboard VGA, 250GB 7200RPM 3.5" HDD, a DVD optical drive, and let's add a 7800GT high end VGA card.
No I'm not kidding, even with the 7800GT (not pictured here) the maximum AC draw running Prime95 together with 3DMark is 183W. Assuming the Antec PSU is running at 75% efficiency, we're looking at less than 140W DC.

SilentPCReview did a detailed analysis of various systems' loading on their PSUs and a very similarly speced system running a more power hungry 6800GT measured a total DC power draw of 125W under CPU load (no 3D loading). As the article points out, most of the loading is on the 12v rail, which they measured at 7.8A. The PW200M's 12v rail is rated at max load of 12A with a peak load of 13.5A. With my 7800GT the difference in total AC draw between CPU load (no 3D) and CPU+3D load is about 52W, and if we assume this is all on the 12v rail, even at 80% efficiency this translates to 3.3A + 7.8A = 11.1A still within the rated 12A. However whether or not running a system so close to it's limits is wise or not is another matter.

Only time (and testing) will tell...
Also I need to get a second AC/DC adaptor, and preferably one close to 200W. So far I've only managed to find the 220W Delta 12V brick used in Shuttle's SD11G5, but with a $90 price tag, I'm still searching for other options.

I will post back after I've had a chance to run these PSUs more...

NOTE: For the defective AC/DC adaptor, upon notifying the online store they offered to immediately send me a replacement. However as the two-way international freight cost would be more than the adaptor's retail price, I figured I would try to get it fixed locally. Since that didn't work out, the store issued me a full refund on their own initiative without asking me to send back the defective brick. I do not have any affiliation with this store (other than having placed on order from them), but would like to add that their customer service has been very professional; they responded quickly to all my questions pre and post sales, shipment was sent immediately upon payment on the exact date they informed me (a Saturday), and resolved the defective item in a more than satisfactory manner. Thank you Carl at .

Goto PW200M installed

Goto PicoPSU Installed

Read more!