Power Supply What the heck...

[Kmpkt]

Hey guys,

Quick question regarding AC-DC PSUs. In my previous iteration of my S4 build, I used a i3 4360 and GTX 950 Plugless (54W + 75W) which rarely if ever pulled more than about 140W from the wall. I went through two HDPlex AC-DC 160 units on this combo before swapping to a much larger external AC-DC (Dell 330W). I've now managed to have a Razer 150W AC-DC go on me running a 4690T + 1050ti (albeit consistently for a couple weeks) which never took more than 125W from the wall at any time under any load. Assuming these bricks are about 90% efficient, shouldn't they only be outputting between 110-120W at load?

Am I missing something here and killing my own bricks, or is this just bad luck? Once I switch back to my Dell 330W everything is perfectly fine, but I'd prefer not to use a 10lb megabrick if I don't need to. Right now I'm planning to put another 160W HDPlex AC-DC into the unit since Larry was kind enough to replace the ones that broke on me, but I don't want to keep frying units if I'm doing something to contribute to this (I really don't think I am). Also worth mentioning is that the two 160W HDplex units blew while using the Z4-ATX-200 while the most recent Razer unit blew while using the HDPlex 250W DC-DC. Cheers to all who can help.

 

[zhl146]

I don't have any great answers, but from what I can tell laptop bricks are notoriously unreliable. Even so, you must be having a string of bad luck to have so many go in such a short time span. I've used the pci 160XT + adaptec brick that @Josh | NFC recommends and that was golden and never gave me any issues. It's a very robust brick and I would recommend that you try that one for your needs as it is rated for 192 watts. It also is a 12V brick, so I think you'll get less heat generation inside your system as downvolting from 19 to 12 causes a good amount (I think).

 

[Stevo_]

I've used a good many bricks for both laptops and desktops over the years and never had one fail. That said, there is a healthy market in counterfeit bricks especially for Dell's stuff on ebay etc. Only other things I would check is the quality of the input AC power, things like AC motors get killed over time by high/low voltage probably not a big issue here. Purity of the sine wave possibly another issue, if the power source is generated from a solar etc using an inverter a distorted input could really screw with PFC correction and efficiency. Power factor of input another possible issue if it's completely whacked at the input already may have similar effects, power companies try to keep sane at their end but a place with lots of reactive or non-linear loads(switched PSUs) could be pretty far off.

 

[EdZ]

Hey guys,

Quick question regarding AC-DC PSUs. In my previous iteration of my S4 build, I used a i3 4360 and GTX 950 Plugless (54W + 75W) which rarely if ever pulled more than about 140W from the wall. I went through two HDPlex AC-DC 160 units on this combo before swapping to a much larger external AC-DC (Dell 330W). I've now managed to have a Razer 150W AC-DC go on me running a 4690T + 1050ti (albeit consistently for a couple weeks) which never took more than 125W from the wall at any time under any load. Assuming these bricks are about 90% efficient, shouldn't they only be outputting between 110-120W at load?

Am I missing something here and killing my own bricks, or is this just bad luck? Once I switch back to my Dell 330W everything is perfectly fine, but I'd prefer not to use a 10lb megabrick if I don't need to. Right now I'm planning to put another 160W HDPlex AC-DC into the unit since Larry was kind enough to replace the ones that broke on me, but I don't want to keep frying units if I'm doing something to contribute to this (I really don't think I am). Also worth mentioning is that the two 160W HDplex units blew while using the Z4-ATX-200 while the most recent Razer unit blew while using the HDPlex 250W DC-DC. Cheers to all who can help.

It's the difference between average power draw, which is within the capacity of your bricks, and burst load, which can peak at quite a bit higher.
Modern CPUs and GPUs will change their power draw very rapidly because they can switch on and off parts of themselves extremely rapidly to save on power, as well as rapidly changing clock speed. The goal is to draw a lot of power and do a lot of work as quickly as possible, then turn off unused areas and clock back down to save power. The faster you can get the work done, the lower the average power draw. But to do this, that inital 'sprint'# draws a lot more power than the average. You can see this with very high resolution power testing, e.g. Toms Hardware's graphs of the R9 Nano's power consumption, which has an average draw of 186W, but has peaks that last for a few milliseconds of 300W-400W! For a (competently built) ATX compliant power supply, you only need to worry about average load, because the PSUs are designed to tolerate burst loads.
The problem comes in that laptop power bricks are designed only to charge an internal laptop battery, a nice steady load. They are not designed to tolerate bursts of high power, because they will be connected to a laptop that contains power circuitry that smooths power spikes. When your desktop components spike up to a high power level, the laptop brick will trip its overcurrent protection circuit (or let out the magic blue smoke....) because it is not designed to handle that load. While a DC-DC converter can be designed to tolerate bursty power loads, I am not aware of any designed to actively buffer power draw spikes to the degree that modern components produce. Instead, they 'pass it on' by exhibiting a similarly bursty power draw to whatever is powering them: the AC-DC brick.

 

[iFreilicht]

Top-notch explanation. You can imagine the reason why your 330W Dell brick isn't having any issues because it can tolerate an average load above the peak load your system might be exhibiting.

I guess you could hack something together to smooth out the spikes with a super-capacitor, but that doesn't sound like fun to me.

 

[Josh | NFC]

Top-notch explanation. You can imagine the reason why your 330W Dell brick isn't having any issues because it can tolerate an average load above the peak load your system might be exhibiting.

I guess you could hack something together to smooth out the spikes with a super-capacitor, but that doesn't sound like fun to me.

I call dibs on Super-Capacitor for a band name!

 

[iFreilicht]

I call dibs on Super-Capacitor for a band name!

And their live-long rivals, the Ultra-Resistors.

 

[ColorZepppelin]

I was planning a build with a thin-itx and GPU powered directly off of 12V power brick. It seems that might not be a great idea in some cases where the hardware requires those high bursts of power going over the rated DTP of the brick. I imagine mostly the problems would come from GPUs.
It would be nice if there was that sort of small battery pack (capacitors?) that would go between the power brick and the hardware.

 

[iFreilicht]

It would be nice if there was that sort of small battery pack (capacitors?) that would go between the power brick and the hardware.

Doesn't that defeat the whole point of the exercise? Power bricks are a compromise between volume reduction and usability, but putting a power brick between your power brick and your pc just so it runs well seems absurd. I'd much rather someone create a good brick that's up to such high peaks.

 

[Phuncz]

I support solving the problem above solving the symptom. The problem is indeed the power bricks not being made for this PC usage.

 

[EdZ]

I support solving the problem above solving the symptom. The problem is indeed the power bricks not being made for this PC usage.

It does, however, alleviate the symptoms in a manner that is immediately implementable using existing parts, rather than requiring new power brick designs (in low volumes, so at high cost).
Whether it's a good way to alleviate the symptoms depends on whether a laptop brick + 'power conditioner' is lower in volume than an external SFX (or similar) PSU.

 

[Kmpkt]

Could you use a small battery on a load sharing device like this to perform the power conditioning?

http://www.mini-box.com/Y-PWR-Hot-Swap-Load-Sharing-Controller

Or any of the other units offered on Pico Box (as there are a few).

 

[Kmpkt]

Also I've been talking about this issue with Larry at HDPlex and he's taking this spiking into account for the 250W AC-DC as he wants it to be able to run a 1080/1080ti in an eGPU application.

 

[ColorZepppelin]

I'd much rather someone create a good brick that's up to such high peaks.

Indeed. Theoretically what would it take for someone to modify a brick (internally) rated at a certain modest wattage to handle these extra bursts. Adding some kind of small supercapacitor?

Could you use a small battery on a load sharing device like this to perform the power conditioning?

So simply you connect a 12V battery to the one 4pin input and a 12V AC/DC power brick to the other 4pin input?

 

[Kmpkt]

I was thinking that you'd have a battery on one and the AC-DC on the other and then a lead from the AC-DC BEFORE the splitter to the input on the battery?

 

[iFreilicht]

Indeed. Theoretically what would it take for someone to modify a brick (internally) rated at a certain modest wattage to handle these extra bursts. Adding some kind of small supercapacitor?

That could work, but I think the more durable solution would be to just take a FlexATX/1U PSU, put a shell around it and modify its connectors. Most of these bricks are also potted (filled with resin) and tightly packed, so it's not like you could just put a few additional components in there.