Favorite Internal PSU form factor?

cleveland

Master of Cramming
Sep 8, 2016
418
222
It is a big factor in entire PSU efficiency, since dc-dc is only half of PSU a PC would need.

If it is 12v dc-dc PSU, it passes the 12v along with minimal losses 🙂

If it is 19v ac-dc high efficiency, a 19v -> 12v dc-dc is needed and that takes a chunk out of the efficiency.

For example:
19v AC-DC 94% efficiency
19v->12v DC-DC 95% efficiency

Both very efficient, high quality - 94-95% is spectacular.

But combine them together... 94%*95% gives overall efficiency around 89%. Good but no longer impressive. A 91% 12v AC-DC + DC-DC may not look as impressive on paper but would come out a little more efficient since the DC-DC only needs to do conversions for minor rails like 5v, 3.3v, etc
You forgot to factor the voltage drop under load.

12v, as you stated, is optimal if there's no resistance, impedance and inductance in the circuit, which is the exact oposite of what happens to be in a computer board.

19v normally is the way to go of most manufacturers because of that. if there's a verifiable current drop (on heavy loads there'll always be), it'll still leave the voltage above the 12v needed for the computer to work properly.
 

NateDawg72

Airflow Optimizer
Aug 11, 2016
359
271
It doesn't feel like we are talking about the same thing.
12v, as you stated, is optimal if there's no resistance, impedance and inductance in the circuit, which is the exact oposite of what happens to be in a computer board.
I didn't forget to factor it in. I don't think anyone here has ever had an issue with that with the power supplies we use. They are made for full load at 12v. I'm not saying voltage drop doesn't happen, it definitely does - It's just that it is already accounted for and within spec. SFX, Flex ATX, PicoPSU, Meanwell... All 12v. Can you show me a case where someone had an issue from voltage drop, rather than something like tripping OCP? Hasn't been an issue for Guryhwa's 12v bricks either.
19v normally is the way to go of most manufacturers because of that. if there's a verifiable current drop (on heavy loads there'll always be), it'll still leave the voltage above the 12v needed for the computer to work properly.
Most manufacturers? What manufacturers? I'm not talking about laptop power bricks, I'm talking about Meanwell, Artesyn, HDPlex, MiniBox, etc. 12V solutions from all of them exception being HDPlex. Laptop brick voltage being 18-20v has a lot to do with batteries.

@Marzipan
For 19V, HDPlex is the way to go for AC-DC & DC-DC units

For 12V AC-DC, MiniBox bricks for <=160w PCs, and Guryhwa's custom bricks (built to order). After that things are more DIY: there is Meanwell EPP/RPS series (ignore EPS from that link), Artesyn CPS250 series.
For good DC-DC units there is MiniBox Pico PSU, TheHack's direct plug unit, Guryhwa's archdaemon (built to order).
 
Last edited:

k0n

Caliper Novice
Jul 3, 2019
25
15
I hope for fanless FlexATX PSUs. 250/300W would be enough for me (single 8 PIN). The HDPLEX 400W can handle up to 300W fanless... so this should be possible.
 

Choidebu

SFF Guru
Aug 16, 2017
1,037
1,000
ummm...12v ac-dc bricks have platinum level efficiency...but what does that have to do with dc-cd PSU?
As @NateDawg72 has explained more in depth, I just like to add:

"DC DC" is just a moniker. They're not standardised so no official name available.

What any PSU does is convert external energy to whatever the system needs.

Recently dc-dc means a combo of ac-dc brick and dc-atx board (convert dc to atx spec voltages etc). It might come from ac-dc --> dc-atx shortened, but ac-atx can just mean normal flex/tfx/sfx/atx, so dc-dc it is.

I used to call it a brick + dpcb (direct plug circuit board), but in light of these newer higher wattage dc-atx unit that won't fit in dpcb format (yeah right tell that to @guryhwa ), they don't plug directly anymore so the name fails.
 

cleveland

Master of Cramming
Sep 8, 2016
418
222
I'm not talking about laptop power bricks, I'm talking about Meanwell, Artesyn, HDPlex, MiniBox, etc.
Oh, I see... you're talking about the ones who don't dictate the industries standards. Got it, dude. good luck on your journey. :)
 

Choidebu

SFF Guru
Aug 16, 2017
1,037
1,000
Oh, I see... you're talking about the ones who don't dictate the industries standards. Got it, dude. good luck on your journey. :)
To make things clear...
Here is Intel Desktop Form Factor Power Supply Design Guide, rev. 2 June 2018, Section 3.2.1 DC Voltage Regulation Requirement, page 14:



It's +/-5% on 12V. Furthermore, on load additional +/-5% is permitted (if I'm reading that correctly).

Voltage drop does exist, but let's have perspective here...

Online calculator for voltage drop across cable shows..
12V, 5A over single, 20AWG, 50cm stranded copper cable drops 0.17V (1.39%) to 11.83V at the other end.

What we do here is well within spec.

Edit: I tried 19V, 3.16A (~60W just like above), and voltage drop is 0.11V or 0.55%.
The difference is quite substantial, more than double the loss percentage. BUT, 0.84% really nothing compared to another 5 to 7% loss in stepping down.
 
Last edited:

Choidebu

SFF Guru
Aug 16, 2017
1,037
1,000
FWIW, ripple =/= tolerance. Just in case someone out there citing atx's <100mV ripple spec.
 

NateDawg72

Airflow Optimizer
Aug 11, 2016
359
271
Oh, I see... you're talking about the ones who don't dictate the industries standards. Got it, dude. good luck on your journey. :)
For the most part I'm talking about the ones that have detailed data sheets and test reports. They are awesome since you get to see the characteristics of the PSU at full load.

I've already completed my journey so thank you 🙂 ryzen 2600 and gtx 1080 run great on an EPP-400. But if you want, we can go warn all the S4 mini hdplex users how they are making a mistake, or tell jhack the meanwell power supplies he sells have too much voltage drop (despite the datasheet and test report saying otherwise!)
 

cleveland

Master of Cramming
Sep 8, 2016
418
222
For the most part I'm talking about the ones that have detailed data sheets and test reports. They are awesome since you get to see the characteristics of the PSU at full load.

I've already completed my journey so thank you 🙂 ryzen 2600 and gtx 1080 run great on an EPP-400. But if you want, we can go warn all the S4 mini hdplex users how they are making a mistake, or tell jhack the meanwell power supplies he sells have too much voltage drop (despite the datasheet and test report saying otherwise!)
looks like someone's been skipping physics classes...
 
  • Angry
Reactions: Choidebu

Choidebu

SFF Guru
Aug 16, 2017
1,037
1,000
I really don't see what we're arguing here. ATX(read: common) psus outputs 12V (along with some other voltages) and you're fine with it, but not with ac-dc psu doing 'just' the 12V?

What? What impedance? Capacitance? You need to start talking specifics.

Was it about 19V vs 12V?
I'm pretty sure 12V _is_ more efficient although not marginally. Just like gold vs platinum. Voltage drop of 19V compared to 12V is negligible compared to heat wasted in the 19V input dc-atx step. Tell me about it I've got a dynamo mini.

I do love that thing anyway, it takes 25V (!? Yes!) from a meanwell EPP-200-24, AND 16-24V battery! Let's see 12V proponents find a buck regulator that can also cleanup ripple, and see how big heatsink on it!
 
Last edited:
  • Like
Reactions: NateDawg72

Stevo_

Airflow Optimizer
Jul 2, 2015
333
200
19v normally is the way to go of most manufacturers because of that. if there's a verifiable current drop (on heavy loads there'll always be), it'll still leave the voltage above the 12v needed for the computer to work properly.
Those 19V DC-DC PSUs that run from bricks are typically buck or even buck-boost when they have a wide input range to deliver the 12V(which is the main power rail) so they have quite a bite more heat dissipation involved. They can get quite hot under not a lot of load compared to the 12V versions which rely on a very low resistance FET to switch the 12V thru and filter it. The mini-box 160W-XT runs very cool, the HDPlex 160W(and higher) units OTOH not so much and tend to cook your RAM(contact them typically) as they are very thick with heatsinks to dissipate the added heat as well.

As long as the 12V brick or something like a Meanwell has sufficient power and feedback the switching supply can adapt for any over/under it senses. In either case these are not open loop supplies feeding the 12V DC-DC PSU where line drop is an issue.
 

Choidebu

SFF Guru
Aug 16, 2017
1,037
1,000
@Stevo_ you sound like the guy to ask this: I've been following Gan FETs news these days. While I'm not up to reading papers on it, I keep finding that most of its uses is at the high voltage step - from 110/250 to 48 or 32 or whatever. At the latter stage they still use current FETs, why is this the case?
 

Stevo_

Airflow Optimizer
Jul 2, 2015
333
200
@Stevo_ you sound like the guy to ask this: I've been following Gan FETs news these days. While I'm not up to reading papers on it, I keep finding that most of its uses is at the high voltage step - from 110/250 to 48 or 32 or whatever. At the latter stage they still use current FETs, why is this the case?
Not really my specialty other than basic EE stuff from way back in college, been wondering myself seems like only low voltage GaN is in laptop and below chargers maybe a current limitation. I'm in digital ASIC design at Luxtera/Cisco so like you mostly what I read on that front. Even in my area, you get insulated from the dirty details of 7nm finfets etc unlike the analog designers.
 
  • Like
Reactions: Choidebu