Are PicoPSU-like supplies the best option at low loads?

PSUs: The source of DC power for all components in the PC & often a big noise source.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee, Devonavar

quest_for_silence
Posts: 5275
Joined: Wed Jun 13, 2007 10:12 am
Location: ITALY

Are PicoPSU-like supplies the best option at low loads?

Post by quest_for_silence » Wed Oct 01, 2014 3:42 am

I started to read this german review:

http://www.computerbase.de/2013-04/stre ... t/drucken/

obviously with the help of Google Translator (I don't speak german: too bad Google Translator loose some graphs).

Initially I was mainly interested in the relevant thermo camera shots, but from what I understood from that Pfab's article, some common wisdoms about those kind of DC-DC converters didn't seem to me quite exact, particularly about their efficiency and power ratings.

What do you think about?
Regards,
Luca

Support SPCR, use these links when you buy: NCIX, Amazon and Newegg

edh
Posts: 1617
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Are PicoPSU-like supplies the best option at low loads?

Post by edh » Wed Oct 01, 2014 4:06 am

About what I'd expect. Power quality is not great with a power brick and a DC-DC PSU. Efficiency I wasn't expecting it to be on top but it is really not great at that either! Maybe at single digit wattages they might still be in the lead.

Perhaps this is because although when PicoPSU designs came out over 5 years ago they were near the top. With time the designs haven't changed much whereas ATX PSUs have changed enormously. One of the advantages of DC-DC PSUs is that the conversion to 5V and 3.3V all happens as far down the line as possible. In effect a lot of ATX PSUs have also copied this design with having a separate board inside handling 12V to 5V and 3.3V.

DC-DC boards like this still have a huge advantage on size for building really compact systems. However, with continued efficiency improvements in standard PSUs their are now far better SFX sized units than was the case 5 years ago. I don't think it'll be too many years until we will see a passive SFX PSU and then making a smaller system will be far easier.
Raijintek Metis: i5-6400, Scythe Ninja 4, Asus Z170i Pro Gaming, 8Gb RAM, GTX 950, Crucial 250Gb M.2, Seasonic X-400

xan_user
*Lifetime Patron*
Posts: 2269
Joined: Sun May 21, 2006 9:09 am
Location: Northern California.

Re: Are PicoPSU-like supplies the best option at low loads?

Post by xan_user » Wed Oct 01, 2014 6:32 am

ive never really thought of them as more efficient than traditional. - I prefer pico because it places some of the PSU heat outside the case, its tiny, and it has no fan. :wink:
Help SPCR keep the lights on, use these links when you buy: NCIX, Amazon and Newegg

Abula
Posts: 3611
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Abula » Wed Oct 01, 2014 7:17 am

Todays PC mainly use 12V, so im PSU like picoPSU, they are dependant on the brick itself in terms of efficiency, there are different models but the ones that i have used are just pass throughs on the 12v, but where they are really good is in very low watt scenarios, where it does make a difference with standard psu.
GameMi - Intel i9 9900K + Gigabyte Z390 AM + Asus RTX2080Ti Strix | MiniMi - Intel i7 8700K + ASRock Z370ITX + MSI RTX2080Ti
StreaMi - Intel i9 7940X + AsRock X299 Taichi + Asus GTX1080Ti Strix | WorkMi - AMD 1950X + AsRock X399 Taichi + MSI GTX1080
ServeMi - Intel Xeon E3-1230 + Supermicro X9SCM-F + 144TB | StoreMi - Intel Xeon E3 1240v5 + Supermicro X11SSM-F + 42TB | CamMi - Intel i7 4770K + MSI H81i + Blue Iris

quest_for_silence
Posts: 5275
Joined: Wed Jun 13, 2007 10:12 am
Location: ITALY

Re: Are PicoPSU-like supplies the best option at low loads?

Post by quest_for_silence » Fri Oct 03, 2014 1:01 am

Abula wrote:but where they are really good is in very low watt scenarios, where it does make a difference with standard psu.
Maybe.

Probably at around 10-15W the efficiency may be 10% higher, I don't know: what I think is that a 10% higher efficiency means about 3W at 15W DC, but you will have a 100-110°C heat source near the mobo PCB, a below average DC quality & a lower regulation (which mean higher VRM temp), and often a sub-par efficiency above a 25-30W power draw, as any modern 300-350W unit will have a comparable or better efficiency from 20-25W, with usually far better vreg and ripple, while running cooler.

If you give a look to the 350W Seasonic SPCR review, at 230V you can expect an about 80% efficiency at 20-22W and >90% at 65W (the tested 150 Pico/Nano was 75% at 15W, 78% at 25W, about 79% at 30W-75W), and even the Seasonic G-360 (which is not particularly efficient at low loads) have an about 73-75% efficiency around 20-25W and around 87% at about 60W. Whether you pick a more modest unit, like the BeQuiet L8 300W 80Plus Bronze, then it showed a 74% efficiency at 25W (1,3W more than the Pico/Nano 150), and an 84% efficiency at 60W (almost 5W less than the Pico/Nano 150), and also the still 80Plus Bronze certified Silverstone ST-30SF did 80% at the 25W level, 82% at 30W, 86% at 60W (while staying fanless up to 180W).

Obviously you can spare a lot of room with a DC-DC converter (over an ATX unit), but IMO at expense of case temperature and overall reliability, so that's neither a free lunch, nor a perfect trade-off, probably better suited just for BayTrail/Silvermont/Kabini/Temash (with not too many hard drives, I would add), where their fanless nature may let them shine over fanned PSUs (but not over hybrid ones, like the quoted Silverstone ST-30SF or the Seasonic Gold-rated TFX and 1U units): but IMO that could be not enough for any Haswell/Kaveri setup (or for rig more than three 3.5" disks), particularly when a good but traditional PSU will cost a fraction of the cost (typically 1/2 to 1/4 of a very good Pico-combo).
Regards,
Luca

Support SPCR, use these links when you buy: NCIX, Amazon and Newegg

Vicotnik
*Lifetime Patron*
Posts: 1816
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Vicotnik » Fri Oct 03, 2014 2:24 am

For a modest system, used mainly for light stuff lite office tasks and web browsing, not unlike the main system in my signature, picoPSU is king. For a one-system-for-all-tasks including gaming, not so much. Although a picoPSU will work in many of those systems also, but will not be optimal.

So, best option at low loads in a modest system? Yes, I think so. Pain in the ass to find a good 12V brick though.
Main: ASRock B85M-ITX | i3-4330 | 16GB DDR3 | Intel 730 240GB | HDPLEX H1-S | picoPSU | No moving parts | Idle 13.9W
HTPC: ASRock H81M-ITX | Pentium G3420 | 4GB DDR3 | Intel 535 120GB | HDPLEX H1-S | picoPSU | No moving parts | Idle 11.2W
Gaming: Intel DH77EB | i5-3570K | GTX 1060 6GB | 16GB DDR3 | TJ08-E | RM750X
Server: ASRock N3150-ITX | ~30TB | G-360 | Idle ~25W

quest_for_silence
Posts: 5275
Joined: Wed Jun 13, 2007 10:12 am
Location: ITALY

Re: Are PicoPSU-like supplies the best option at low loads?

Post by quest_for_silence » Fri Oct 03, 2014 3:15 am

Vicotnik wrote:not unlike the main system in my signature, picoPSU is king.

As already said, I may agree but just when you use a very small enclosure (and dumping a considerable amount of heat nearby the mobo PCB), because just 3W of wasted power don't worth all the hassle.

By the way Vicotnik, as said many times to several our co-forumers, I am still very dubious/cautious about idle measurement as your main system's one (thirteen w at the wall...): to the best of my knowledge, portable, consumer level instrumentation just doesn't have the necessary reliability and precision to do that, so any choice made just upon those is likely rather groundless (IMHO, of course).

Vicotnik wrote:Pain in the ass to find a good 12V brick though.

Which, when supplied by a reputable and knowledgeable source, it often costs almost as much as a Silverstone ST-30SF here in Italy (40 euros).
Last edited by quest_for_silence on Fri Oct 03, 2014 3:16 am, edited 1 time in total.
Regards,
Luca

Support SPCR, use these links when you buy: NCIX, Amazon and Newegg

edh
Posts: 1617
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Are PicoPSU-like supplies the best option at low loads?

Post by edh » Fri Oct 03, 2014 3:16 am

quest_for_silence wrote:[If you give a look to the 350W Seasonic SPCR review, at 230V you can expect an about 80% efficiency at 20-22W and >90% at 65W (the tested 150 Pico/Nano was 75% at 15W, 78% at 25W, about 79% at 30W-75W), and even the Seasonic G-360 (which is not particularly efficient at low loads) have an about 73-75% efficiency around 20-25W and around 87% at about 60W. Whether you pick a more modest unit, like the BeQuiet L8 300W 80Plus Bronze, then it showed a 74% efficiency at 25W (1,3W more than the Pico/Nano 150), and an 84% efficiency at 60W (almost 5W less than the Pico/Nano 150), and also the still 80Plus Bronze certified Silverstone ST-30SF did 80% at the 25W level, 82% at 30W, 86% at 60W (while staying fanless up to 180W).
The 80+ Titanium spec introduces an efficiency requirement for 10% load. Previously there has been no requirement at this low load but now it will be 90% efficiency. 80+ Titanium takeup is likely to be slower given the diminishing returns and increased investment required with each iteration. Currently only Corsair and Super Flower have 80+ Titanium consumer offerings abnd as ever they start at the insane 1.5kW size.
Raijintek Metis: i5-6400, Scythe Ninja 4, Asus Z170i Pro Gaming, 8Gb RAM, GTX 950, Crucial 250Gb M.2, Seasonic X-400

fuzzymath10
Posts: 69
Joined: Tue May 03, 2011 5:52 pm
Location: Toronto

Re: Are PicoPSU-like supplies the best option at low loads?

Post by fuzzymath10 » Fri Oct 03, 2014 4:20 am

It's possible certain combinations work better than others too. I ended up with a wide-input pico psu and a universal laptop brick from cooler master (NA 90 I think).

http://www.jonnyguru.com/modules.php?na ... 5&reid=207

I have a Z77-N Wifi + i7 3770 + SSD and it idles around 25W. Not outstanding, but pretty good. I've even overclocked it a bit, to 1-core turbo at 4.2, and 2/3-core turbo at 4.1, and 4-core at 3.8.

It has only 6A on the 12V which can be a bit limiting. I messed with the clocks a lot to prevent crashing from PSU overloading.
Lone Industries L1: i7 3770, 2x4GB Mushkin Silverline, 512GB MX100, Dell U2412M + U2211H
Lian-Li A04: i5 2400, 2x4GB Mushkin Silverline, 160GB Intel 320, 4x3TB FlexRAID, all on Seasonic x460, hooked up to Sony 40" EX620 + KEF 2005.3 + Yamaha RX-V667

Pappnaas
Posts: 726
Joined: Mon May 16, 2011 11:23 am
Location: Germany

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Pappnaas » Fri Oct 03, 2014 4:27 am

Just a thought that occured to me:

Pico-PSU's reputation comes from the days before any PSU efficiency rating was invented. So compared to the best PSUs back then, their efficiency was unrivaled.

But todays gold/platinum/titanium rated PSUs seem to have caught up in low power efficiency, so it really boils down to the quality of the brick used to power the pico.

And yes, i second the thought that a kill-a-watt is not precise enough to catch todays low idle consumption, given that most cheap devices sport an error margin of +-5 V.

quest_for_silence
Posts: 5275
Joined: Wed Jun 13, 2007 10:12 am
Location: ITALY

Re: Are PicoPSU-like supplies the best option at low loads?

Post by quest_for_silence » Fri Oct 03, 2014 4:48 am

fuzzymath10 wrote:It's possible certain combinations work better than others too.

For sure, but at the same time I'm really amazed that a branded and expensive solution from Streacom doesn't look that good.

Anyway, I would stress it again, as those tiny PSU still have some appeal: when you want to use such a small enclosure like your Lone Industries L1 (which has a footprint just greater than a 140mm fan), or an heatsink case like the Vicotnik's HDPlex, I've no or little objections about a Pico-esque PSU, probably there's no room to put anything else.

Otherwise, there are better options around (IMO): my hope is that some reputable firm start to improve those converters and their bricks, as well as it happened to ATX PSUs in the last few years.
Regards,
Luca

Support SPCR, use these links when you buy: NCIX, Amazon and Newegg

Vicotnik
*Lifetime Patron*
Posts: 1816
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Vicotnik » Fri Oct 03, 2014 5:12 am

quest_for_silence wrote:Probably at around 10-15W the efficiency may be 10% higher, I don't know: what I think is that a 10% higher efficiency means about 3W at 15W DC, but you will have a 100-110°C heat source near the mobo PCB, a below average DC quality & a lower regulation (which mean higher VRM temp)
I've never actually thought much about the heat output of the picoPSU. I've never actually done any heat measurement, but we can do the math. Given the efficiency of the thing and the low power consumption of the system, how much heat are we talking about? 12V flows right through. Some wasted heat in converting 12V to 5V and 3.3V but we don't draw much on those rails. Bottom line; heat output should not be an issue.
quest_for_silence wrote:By the way Vicotnik, as said many times to several our co-forumers, I am still very dubious/cautious about idle measurement as your main system's one (thirteen w at the wall...): to the best of my knowledge, portable, consumer level instrumentation just doesn't have the necessary reliability and precision to do that, so any choice made just upon those is likely rather groundless (IMHO, of course).
You are right. The best we can do is to measure using a reasonably reliable meter and to specify the equipment used and how the test is done. In my case I'm using a NZR SEM16+. Not military grade for sure, but imo good enough for my needs. Seems to correctly measure the standby power of a NEXA switch at 0.2W AC and a LED lamp at about 0.8W AC.
By idle I mean display on, ethernet on but no activity.

We need a lab for central testing to do better than this. So take it for what it is.
Main: ASRock B85M-ITX | i3-4330 | 16GB DDR3 | Intel 730 240GB | HDPLEX H1-S | picoPSU | No moving parts | Idle 13.9W
HTPC: ASRock H81M-ITX | Pentium G3420 | 4GB DDR3 | Intel 535 120GB | HDPLEX H1-S | picoPSU | No moving parts | Idle 11.2W
Gaming: Intel DH77EB | i5-3570K | GTX 1060 6GB | 16GB DDR3 | TJ08-E | RM750X
Server: ASRock N3150-ITX | ~30TB | G-360 | Idle ~25W

quest_for_silence
Posts: 5275
Joined: Wed Jun 13, 2007 10:12 am
Location: ITALY

Re: Are PicoPSU-like supplies the best option at low loads?

Post by quest_for_silence » Fri Oct 03, 2014 5:52 am

Vicotnik wrote:I've never actually done any heat measurement, but we can do the math.

Well, no, I'm not just talking about wasted heat dumped inside the case, if you read the german review maybe you could take my point:

http://www.computerbase.de/2013-04/stre ... il-test/7/

So you likely should have a >100°C point on your mobo, probably two: or isn't it so?
Whether it were so, with reference to those, do you think they won't matter at all, at least reliability wise? And according to you, won't those heat up the nearby air at all?

Vicotnik wrote:I'm using a NZR SEM16+. Not military grade for sure, but imo good enough for my needs.

It looks like a nice piece of equipment. Unfortunately, as most manufacturers, they do not publish a complete datasheet: NZR talks of an about 1% precision, without stating at which power level it's referred, and which is the absolute minimum error in W, so that I just cannot check those important data.

Thanks for your clarifications, anyway, I appreciate them very much.
Last edited by quest_for_silence on Fri Oct 03, 2014 6:38 am, edited 1 time in total.
Regards,
Luca

Support SPCR, use these links when you buy: NCIX, Amazon and Newegg

Vicotnik
*Lifetime Patron*
Posts: 1816
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Vicotnik » Fri Oct 03, 2014 6:38 am

quest_for_silence wrote:Well, no, I'm not just talking about wasted heat dumped inside the case, if you read the german review maybe you could take my point:

http://www.computerbase.de/2013-04/stre ... il-test/7/

So you likely have a >100°C point on your mobo, probably two. With reference to those, do you think they won't matter at all, at least reliability wise? And according to you, won't those heat up the nearby air at all?
I take your point. But you are talking about full load, with a Streacom Nano150 that might be close to the VRM circuity. In my case it's mostly idle, with a picoPSU no where close to the VRMs.

I worry about the heat coming from my picoPSU about as much as I care about the heat coming of my SSD, my RAM or anything else in the box; in relation to how much heat the stuff gives out. The coils on my picoPSU are warm to the touch. I think they might heat up a bit during load but not to the point that I worry about the RAM stick that is quite close by.

Is that the barrel connector in the third picture? That is a weak spot on the picoPSU, replacing that weak barrel connector (usually specified for just 5A) is a good idea.
Main: ASRock B85M-ITX | i3-4330 | 16GB DDR3 | Intel 730 240GB | HDPLEX H1-S | picoPSU | No moving parts | Idle 13.9W
HTPC: ASRock H81M-ITX | Pentium G3420 | 4GB DDR3 | Intel 535 120GB | HDPLEX H1-S | picoPSU | No moving parts | Idle 11.2W
Gaming: Intel DH77EB | i5-3570K | GTX 1060 6GB | 16GB DDR3 | TJ08-E | RM750X
Server: ASRock N3150-ITX | ~30TB | G-360 | Idle ~25W

BillyBuerger
Patron of SPCR
Posts: 854
Joined: Fri Dec 27, 2002 1:49 pm
Location: Somerset, WI - USA
Contact:

Re: Are PicoPSU-like supplies the best option at low loads?

Post by BillyBuerger » Fri Oct 03, 2014 1:01 pm

Interesting article, but it doesn't jive with my experience with picoPSUs. It could be that this is just a bad/cheap product. I have a couple of different picoPSUs in both the 12V and wide input variations. I did some testing in systems between around 20-40W and the picoPSUs gave the best results compared to some good ATX PSUs. Unfortunately I never tested my Seasonic SS-460FL on the same systems. But a Seasonic SS-300SFD and SS-350ET both were behind the picos in general. Those are only 80+. A Dell RM-112 which is in the 80+ silver/gold range was very close to most of the picos but is also a very low wattage PSU at 220W compared to most ATX PSUs which are 400W+. Obviously the brick is very important. I have a 200W 12V brick from an old shuttle case that was basically tied with the SS-330SFD. But that same pico with a good 60W brick used 5W less and was a good bit ahead of the RM-112.

I prefer to see graphs of efficiency by wattage, not percent. As that gives you a better view of where the strong points of the different PSUs lie. A pico and an ATX PSU may both be 80% efficient at 20%, but on the 150W pico that's at 30W while a 400W ATX would be at 80W. Measure that ATX PSU at 30W and you won't get 80%. So the pico may not be great at higher loads but could still be better at low loads.

So yeah, there are some good ATX PSUs these days that make picoPSUs not as impressive as they once were. But they still can be a good choice for specific configurations. The 10% requirement on 80+ titanium sounds nice. That might help bring up some of the efficiency on the low end a little bit. As long as it works it's way into reasonable wattage PSUs.

CA_Steve
Moderator
Posts: 7444
Joined: Thu Oct 06, 2005 4:36 am
Location: Monterey Bay, CA

Re: Are PicoPSU-like supplies the best option at low loads?

Post by CA_Steve » Fri Oct 03, 2014 3:11 pm

I don't think we'll see titanium PSUs at reasonable wattages in the US. Maybe in EU with 230V.
1080p Gaming build: i5-4670K, Mugen 4, Asrock Z97 Anniversary, MSI GTX 1660 Gaming X, 8GB 1866 RAM, Samsung 860 Evo 500GB, Crucial MX100 256GB, WD Red 2TB, Samsung DVD burner, Fractal Define R4, Antec True Quiet 140 (2 front + rear) case fans, Seasonic SSR-550FX. 35W idle, 45-55W video streaming, 170-200W WoW, 318W stress test (Prime95 + Furmark)

Telstar
Posts: 140
Joined: Fri Oct 05, 2007 2:20 pm
Location: Italy

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Telstar » Fri Oct 03, 2014 4:41 pm

xan_user wrote:ive never really thought of them as more efficient than traditional. - I prefer pico because it places some of the PSU heat outside the case, its tiny, and it has no fan. :wink:
This is the exact reason why I used one in my fanless Music PC.
2600k@4,5GHz 1.25v | Corsair H100 w/ AP120 QE PWM| Asrock P67 Extreme4 | 8GB Corsair Vengeance 1866 CL9 | GTX970 Asus Strix | Sandisk Extreme II 480GB | WD RED 3TB x2 | intel i210-T1 LAN | Lian-Li T60 open case | Corsair HX850i | BenQ BL3200PT 1440p 32" | audio-gd DAC19DSP > DIY headamp > Audio Technica W1000X | 300Mbps FTTH

edh
Posts: 1617
Joined: Mon Mar 17, 2008 1:49 pm
Location: UK

Re: Are PicoPSU-like supplies the best option at low loads?

Post by edh » Sun Oct 05, 2014 10:19 am

CA_Steve wrote:I don't think we'll see titanium PSUs at reasonable wattages in the US. Maybe in EU with 230V.
With the way 80+ ratings are done, that won't happen. The only 230V ratings are for redundant PSUs and the % requirements are higher than for the 115V measurements. 80+ is entirely US based, hence why the ratings (with exception to redundant server PSUs) are always run on the kind of weak, watery stuff that you Americans call electricity. :wink:
Raijintek Metis: i5-6400, Scythe Ninja 4, Asus Z170i Pro Gaming, 8Gb RAM, GTX 950, Crucial 250Gb M.2, Seasonic X-400

Stevo_
Posts: 43
Joined: Fri Oct 17, 2014 2:46 pm

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Stevo_ » Wed Oct 22, 2014 9:25 am

I think the Johnny Guru article above captures the real issue, on a 12V DCDCPSU running off of a 12V brick the 12V power droop at higher loads becomes an issue, you really have to spec a capable brick. As the article indicates, this is less of a problem with a 19V based brick/DC-DC system.

I use several HD PLEX 19V DC-DCs pico style PSUs with 19V bricks, one on a 3770K/750 Ti system and I can run Prime95 + Kombustor all day without issue(using their 160W/200W peak model) 330W Dell 19V brick. Other system is a J2900(10W TDP) based fanless system using a 30W Toshiba brick.

Vicotnik
*Lifetime Patron*
Posts: 1816
Joined: Thu Feb 13, 2003 6:53 am
Location: Sweden

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Vicotnik » Wed Oct 22, 2014 10:08 am

Stevo_ wrote:I think the Johnny Guru article above captures the real issue, on a 12V DCDCPSU running off of a 12V brick the 12V power droop at higher loads becomes an issue, you really have to spec a capable brick. As the article indicates, this is less of a problem with a 19V based brick/DC-DC system.
That is not a problem with a good 12V brick. Unfortunately they use crappy EDAC bricks for 12V and a better FSP brick for 19V. Their conclusion might have been different with a better 12V brick.

With a beefy 12V brick like the Dell DA-2 and any picoPSU (like the 90W model) I've powered systems with discrete graphics cards without issue (once the barrel connector was replaced). I've had problems with the one EDAC adapter I've used though.
Main: ASRock B85M-ITX | i3-4330 | 16GB DDR3 | Intel 730 240GB | HDPLEX H1-S | picoPSU | No moving parts | Idle 13.9W
HTPC: ASRock H81M-ITX | Pentium G3420 | 4GB DDR3 | Intel 535 120GB | HDPLEX H1-S | picoPSU | No moving parts | Idle 11.2W
Gaming: Intel DH77EB | i5-3570K | GTX 1060 6GB | 16GB DDR3 | TJ08-E | RM750X
Server: ASRock N3150-ITX | ~30TB | G-360 | Idle ~25W

MikeC
Site Admin
Posts: 12283
Joined: Sun Aug 11, 2002 3:26 pm
Location: Vancouver, BC, Canada
Contact:

Re: Are PicoPSU-like supplies the best option at low loads?

Post by MikeC » Wed Oct 22, 2014 12:10 pm

A pertinent fact is that, afaik, the "Class V" rating is the current top efficiency standard for for power bricks, and this is only 87% or better. (International Efficiency Marking Protocol for External Power Supplies @ energystar.gov) A Class VI is mentioned in that protocol but only reserved for future use.

This doesn't mean that all Class V adapters have only 87% efficiency; they may offer higher. Without testing of a variety of models, we can only guess at how much higher the efficiency might go. What is obvious is that mfgs have no incentive (or benefit) to achieve better than 87% without a broadly accepted standard/certification tag that they can proudly wave.

Hence, a PicoPSU's very best efficiency will be no better than 87%. Losses in further conversion to 5 and 3.3V will be on top of that. Peak efficiency for a 120W is 96% at 5A for 5V, 94% at 5A at 3.3V (both probably higher than normal for a current PC); this doesn't include the 12V conversion loss. When this is accounted for, the overall efficiency for 5/3.3V is ~83%. At the more likely 2A on both 5/3.3V lines (see Power Distribution in Three PCs (2012) ), Pico120 efficiency drops below 93%; let's say 92%. Then overall efficiency drops to 80%.

Still, since the lower voltage lines represent only about 10% of total load, these losses aren't that significant. The most pertinent point is that actual picoPSU efficiency is always going to be lower than 87%.

A 300W 80+ Gold PSU must do 87% at 60W to pass; this is probably at least 2% better than any combination of Pico + AC/DC adapter, which represents >2.5W at the AC input.

At 30W, that 300W Gold will probably do no better than say 80%. A Pico might still pull off 84%, if the 12V adapter power curve is right. This difference is 2W at the wall.

IMO, these power differences are not enough to quibble over unless you're working on a truly starvation energy diet.

So... yes, the Pico definitely has a place for low power systems for Watt counters and to keep system size & internal heat to a minimum, but above 40~50W load, a modest 300W 80+ Gold PSU is more efficient. Only when we can get AC/DC bricks with >90% efficiency will the Pico remain competitive to say 100W or higher.

In the context of this discussion, the truly neat low power solutions are conversion-on-board devices like those on NUC, BRIX, ZBOX & Thin m-ITX boards that take a standard 19VDC laptop brick. The 19V bricks will hit >90% efficiency before the 12VDC ones, their sheer numbers would result in much greater overall energy saving for the world (than improving 12VDC adapter efficiency, tho perhaps the 2 will benefit equally from the same improvements).
Mike Chin, SPCR Editor/Publisher
Support SPCR by buying your gear through this link: Amazon

Stevo_
Posts: 43
Joined: Fri Oct 17, 2014 2:46 pm

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Stevo_ » Wed Oct 22, 2014 12:26 pm

Vicotnik wrote:
Stevo_ wrote:I think the Johnny Guru article above captures the real issue, on a 12V DCDCPSU running off of a 12V brick the 12V power droop at higher loads becomes an issue, you really have to spec a capable brick. As the article indicates, this is less of a problem with a 19V based brick/DC-DC system.
That is not a problem with a good 12V brick. Unfortunately they use crappy EDAC bricks for 12V and a better FSP brick for 19V. Their conclusion might have been different with a better 12V brick.

With a beefy 12V brick like the Dell DA-2 and any picoPSU (like the 90W model) I've powered systems with discrete graphics cards without issue (once the barrel connector was replaced). I've had problems with the one EDAC adapter I've used though.
Quoted but not read apparently. Of course, not everyone is up to hacking the Dell DA-2 connector either.
Last edited by Stevo_ on Wed Oct 22, 2014 1:02 pm, edited 1 time in total.

quest_for_silence
Posts: 5275
Joined: Wed Jun 13, 2007 10:12 am
Location: ITALY

Re: Are PicoPSU-like supplies the best option at low loads?

Post by quest_for_silence » Wed Oct 22, 2014 1:02 pm

MikeC wrote:At 30W, that 300W Gold will probably do no better than say 80%. A Pico might still pull off 84%, if the 12V adapter power curve is right. This difference is 2W at the wall.

IMO, these power differences are not enough to quibble over unless you're working on a truly starvation energy diet.

Just a side note: it depends of the specific PSU topology, as differences can be either even more subtle, or on the contrary noticeably wider.

At the 25W level, with 230V mains, an average DC-DC converter like the 75W LC-Power LC75ITX (12V power brick) is around 80%: as comparison, the gold-rated 300-350W Seasonic platform seen in their 1U ant TFX units is around 75%.
But if we disregard the 300W level, we can note that a gold-rated Seasonic G 550W shows a 70-71% efficiency, while a still gold-rated Cooler Master V550S has an about 82% efficiency, as well as the elder Super Flower Golden Green 550W is still capable of a respectable 81%. There are similar oddities even among the lineup of a same brand, so that at the 25W power level the bronze-rated Silverstone ST30SF 300W hold on 79-80%, while the gold-rated ST45SF-G offers a more modest 75% efficiency.

From an end-user perspective, with reference to Pico and the likes what is really disappointing, IMO, it's the absolute lack of information about power bricks quality/efficiency in the different markets, so that it's often wearying to sort out and then actually find out the right brick.

I would add to the deciding factors, along with efficiency and dimensions, also price.

Btw, thanks for your detailled and helpful post.
Last edited by quest_for_silence on Thu Oct 23, 2014 11:54 am, edited 1 time in total.
Regards,
Luca

Support SPCR, use these links when you buy: NCIX, Amazon and Newegg

bastiaan
Posts: 98
Joined: Fri Oct 19, 2012 4:23 am

Re: Are PicoPSU-like supplies the best option at low loads?

Post by bastiaan » Thu Oct 23, 2014 12:41 am

edh wrote:
CA_Steve wrote:I don't think we'll see titanium PSUs at reasonable wattages in the US. Maybe in EU with 230V.
With the way 80+ ratings are done, that won't happen. The only 230V ratings are for redundant PSUs and the % requirements are higher than for the 115V measurements. 80+ is entirely US based, hence why the ratings (with exception to redundant server PSUs) are always run on the kind of weak, watery stuff that you Americans call electricity. :wink:
This has recently changed. If you visit the 80 Plus website, you'll find they now have a category called '230V EU Internal' for non-redudant PSUs. Presumably this is intended to cater to manufacturers aiming at the 230V markets, who can now get their products certified for 230V without having to worry about 115V efficiency.

Only a few PSUs are certified at the moment. Unfortunately, they do not seem to test PSUs that are electrically compatible with both grids for both certification programmes.

Pappnaas
Posts: 726
Joined: Mon May 16, 2011 11:23 am
Location: Germany

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Pappnaas » Thu Oct 23, 2014 4:16 am

I bet this whole efficiency thing will change as soon as people in the US are forced to pay as much for electricity as we in Europe are paying now.

Looking at fuel pricing and the impact it had on car sales in the US, I can sense a similar development.

CA_Steve
Moderator
Posts: 7444
Joined: Thu Oct 06, 2005 4:36 am
Location: Monterey Bay, CA

Re: Are PicoPSU-like supplies the best option at low loads?

Post by CA_Steve » Thu Oct 23, 2014 6:38 am

How much are you paying for a kWh? Policies in the US vary state to state and utility to utility. In CA, utilities are tasked to provide tier pricing - the more you use, the more you pay per kWh. Mine starts at $0.136/kWh and progresses up to $0.36/kWh. Conversely, I remember looking at my mom's power bill (midwest state) with a flat rate of $0.08/kWh. Never going to get much conservation going there.
1080p Gaming build: i5-4670K, Mugen 4, Asrock Z97 Anniversary, MSI GTX 1660 Gaming X, 8GB 1866 RAM, Samsung 860 Evo 500GB, Crucial MX100 256GB, WD Red 2TB, Samsung DVD burner, Fractal Define R4, Antec True Quiet 140 (2 front + rear) case fans, Seasonic SSR-550FX. 35W idle, 45-55W video streaming, 170-200W WoW, 318W stress test (Prime95 + Furmark)

quest_for_silence
Posts: 5275
Joined: Wed Jun 13, 2007 10:12 am
Location: ITALY

Re: Are PicoPSU-like supplies the best option at low loads?

Post by quest_for_silence » Thu Oct 23, 2014 11:38 am

CA_Steve wrote:How much are you paying for a kWh?

Here it's about $0.33/kWh on average.
Regards,
Luca

Support SPCR, use these links when you buy: NCIX, Amazon and Newegg

Stevo_
Posts: 43
Joined: Fri Oct 17, 2014 2:46 pm

Re: Are PicoPSU-like supplies the best option at low loads?

Post by Stevo_ » Fri Oct 24, 2014 7:45 am

quest_for_silence wrote:
Abula wrote:but where they are really good is in very low watt scenarios, where it does make a difference with standard psu.
Maybe.
.....

Obviously you can spare a lot of room with a DC-DC converter (over an ATX unit), but IMO at expense of case temperature and overall reliability, so that's neither a free lunch, nor a perfect trade-off, probably better suited just for BayTrail/Silvermont/Kabini/Temash (with not too many hard drives, I would add), where their fanless nature may let them shine over fanned PSUs (but not over hybrid ones, like the quoted Silverstone ST-30SF or the Seasonic Gold-rated TFX and 1U units): but IMO that could be not enough for any Haswell/Kaveri setup (or for rig more than three 3.5" disks), particularly when a good but traditional PSU will cost a fraction of the cost (typically 1/2 to 1/4 of a very good Pico-combo).
I think you have to expand your view from the article, there are other DC-DC supplies out there that don't run blazing hot, HD-Plex for one, after using a pcioPSU style Z3 (http://www.pico-box.com/z3-atx-200.html) which couldn't be touched at any load whereas my HD Plex units(have 150W, 160W, and 250W units) run nice and cool. These also feature heatsinks on their MOSFETs and don't plug directly into the mobo unlike the mini-box and pico-box units so when you do crank up the load the heat can be transferred somewhere else besides open air and mobo. Not sure there's been any reliabilty ratings done but PSUs overall seem to be a weak point all over, in any form factor. Only component I've ever had to RMA so far was an Enermax PSU.

Cost numbers from (1/2 to 1/4) are true for lower down SFX but my Dell 330W brick ($80) plus HD-Plex 160W ($45) was about even with some Silverstone SFX PSUs and less than others of theirs (YMMV of course especially in the EU). Even the 250w ($80) would just bring it about even to a 600W SS unless they've dropped in price. Bricks can be had for much less at lesser power and good quality units too (Delta, FSP & others). As a contrast my latest Bay Trail build uses a 30W Toshiba brick had for $12. A 12V combo for a Bay Trail setup could be put together for well less than an ST30. SFX and ATX PSUs impose too many sacrifices in PC size for me to consider anymore and are usually operating at the very low-end of their efficiency curve unless you're mining, folding, or gaming. Even running Prime + Kombustor on my 3770K/750Ti system power tops out ~130W HD Plex only warm to the touch(330W/160W above combo) the real heat at that point is from GPU and CPU, still enough left to OC some. Going by alll the complaints on [H] forum about the Silverstone's SFX PSUs I'll take my silent and cool brick/DC-DC combo over one of those anyday. The 250W HD Plex can power a fairly serious gaiming rig.

RIchardF
Posts: 42
Joined: Tue Oct 21, 2014 9:46 am

Re: Are PicoPSU-like supplies the best option at low loads?

Post by RIchardF » Mon Dec 15, 2014 8:14 am

Pappnaas wrote:I bet this whole efficiency thing will change as soon as people in the US are forced to pay as much for electricity as we in Europe are paying now.

Looking at fuel pricing and the impact it had on car sales in the US, I can sense a similar development.
We bought a VW Passat TDI and diesel is routinely $1.10 higher per gallon except in the warm months. Someone also did math that said it would take ten years of quite a bit of driving to recoup the extra upfront cost of the TDI upgrade (ignoring VW's questionable reliability, although this Passat isn't exactly a Mexican Beetle).

But, I like the car anyway. It's comfortable thanks to the softer suspension and wider wheelbase. Those features were absolute requirements because I have a bad back. It's also nice to not have to refuel as often thanks to the big tank and the good efficiency (except for stop/go short trips). The big question is now about nanoparticles created by the emissions system.

Anyway... Aside from efficiency, what I'm concerned with is ripple. Some of those tests were not very impressive at all.

aquelito
Posts: 14
Joined: Thu Feb 17, 2011 3:19 am

Re: Are PicoPSU-like supplies the best option at low loads?

Post by aquelito » Sun Feb 21, 2016 4:13 am

Stevo_ wrote:after using a pcioPSU style Z3 (http://www.pico-box.com/z3-atx-200.html) which couldn't be touched at any load whereas my HD Plex units(have 150W, 160W, and 250W units) run nice and cool. These also feature heatsinks on their MOSFETs and don't plug directly into the mobo unlike the mini-box and pico-box units so when you do crank up the load the heat can be transferred somewhere else besides open air and mobo.
Hi, please allow me to continue this thread !

I just bought a Z3-ATX-200 PSU. Do you think it would be possible to use a 24pin ATX extension cable to :

- avoid having to plug the PSU straight into the mobo,
- place the PSU wherever you want, that means into direct airflow.

Using small heatsinks on the PSU wouldn't hurt either right ?
Image

To sum up, improved cooling is going to help improve the Z3-ATX-200 performances (less heat = less energy deperdition ?) ?

Last thing :

The Z3-ATX-200 has 5.5 mm barrel connector. My brick (HP 230W) has a 7.4 DC connector. Using that kind of adapter is "dangerous" ?
http://www.amazon.com/7-4x5-0mm-5-5x2-5 ... B00G700QQ0

Thanks a lot

Post Reply