Announcement

Collapse
No announcement yet.

The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

    #21
    Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

    a $15 power supply thats not a steaming pile.
    now label it properly and its actually a good one for the average joes pc replacement market.
    now to find the real world load death point ....

    Comment


      #22
      Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

      Originally posted by dmill89 View Post
      FYI: The protection IC in this PSU is a Weltrend wt751002 which appears to provide over voltage protection and under voltage protection but no over current protection (though even if I used this in a PC it wouldn't be in an application with enough load where OCP would be needed)
      That's to make sure it can provide 875w
      "The one who says it cannot be done should never interrupt the one who is doing it."

      Comment


        #23
        Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

        Originally posted by Per Hansson View Post
        That's to make sure it can provide 875w
        And a glowing report!
        PeteS in CA

        Power Supplies should be boring: No loud noises, no bright flashes, and no bad smells.
        ****************************
        To kill personal responsibility, initiative or success, punish it by taxing it. To encourage irresponsibility, improvidence, dependence and failure, reward it by subsidizing it.
        ****************************

        Comment


          #24
          Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

          Originally posted by ChaosLegionnaire View Post
          if any nub or *gasp* even a nub engineer on another forum says all caps are created equal, i'd show them this forum topic! that'll blow 'em away and teach 'em good!
          Admittedly the increase in capacitance probably had more effect than the better quality caps (the Changx caps probably wouldn't last but for now only tested 0.02-0.03 ohm higher ESR than the Chemi-Con/Nichicon caps that replaced them with the exception of the little 470uf caps on the 5vsb/-12v rails, granted this was with a cheepo tester, though the Japanese caps do have less voltage loss)

          Changx GR 1000uf 16v:


          Chemi-Con KYB 2200uf 16v:


          Chang x GR 1000uf 10v:


          Chemi-Con KY 2200uf 10v:


          Changx GR 470uf 16v:


          Nichicon HE 1000uf 16v:



          Originally posted by ChaosLegionnaire View Post
          compete with even high quality units from seasonic and superflower.
          And just for fun here's a Seasonic X560 under the same load for comparison:

          12V:


          5V:


          3.3V:
          Attached Files
          Last edited by dmill89; 10-11-2018, 05:20 PM.

          Comment


            #25
            Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

            Originally posted by kc8adu View Post
            a $15 power supply thats not a steaming pile.
            now label it properly and its actually a good one for the average joes pc replacement market.
            +10

            Definitely a decent power supply now that you can use in any average-powered rig (i.e. nothing with a big power-hungry GPU over 80 Watts.)

            And yes, probably a good idea to re-label it anyways, in case you ever forget you recapped it and wonder what this PSU is doing in your stash.

            In my case, I had a crappy J&C 235W PSU based on some Meico gutless crap. I gutted the original PSU a long time ago for components. But its case became a new home to a homeless HiPro 250W PCB that someone removed from its original case for whatever reason and left for recycling.
            Oh well, I'm still using that PSU. It looks completely ghetto. In fact, I know I posted it somewhere in the Ghetto mod thread. And to make sure the label wasn't confusing, I just took a Sharpie / permanent marker and crossed out the original J&C model number and ratings, then wrote down what PSU was actually inside + its ratings.

            Comment


              #26
              Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

              Originally posted by momaka View Post
              And yes, probably a good idea to re-label it anyways, in case you ever forget you recapped it and wonder what this PSU is doing in your stash.
              Done, this seems realistic-ish:
              Attached Files

              Comment


                #27
                Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                ^ Nice!
                I like how you matched the sticker color and added a recapped date too.

                Still not sure if the 12V rail will do 30 Amps continuous with two 20-Amp rectifiers in parallel... but you're probably not that far off. Reason I say this is because most OEM PSU's I've seen with this arrangement (at least on the 5V rail of older ones) usually state around 24-25 Amps max. But those were pretty conservatively rated overall. So I could be wrong on this.

                Comment


                  #28
                  Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                  Originally posted by momaka View Post
                  ^ Nice!
                  I like how you matched the sticker color
                  I'm surprised how well it came out, it wasn't anything fancy, just a Brother PT-2030 label maker and some 12mm white on black tape.

                  Originally posted by momaka View Post
                  Still not sure if the 12V rail will do 30 Amps continuous with two 20-Amp rectifiers in parallel... but you're probably not that far off. Reason I say this is because most OEM PSU's I've seen with this arrangement (at least on the 5V rail of older ones) usually state around 24-25 Amps max. But those were pretty conservatively rated overall. So I could be wrong on this.
                  Admittedly it was just a guess since right now the 257W (with 14A on the 12V) load I tested it with is the most I have (I need to pick up some bigger electronic loads some time), but I figured it is allot better than the original sticker.
                  Last edited by Per Hansson; 10-23-2018, 01:42 PM. Reason: fixed quote

                  Comment


                    #29
                    Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                    Originally posted by momaka View Post
                    +10

                    Definitely a decent power supply now that you can use in any average-powered rig (i.e. nothing with a big power-hungry GPU over 80 Watts.)

                    And yes, probably a good idea to re-label it anyways, in case you ever forget you recapped it and wonder what this PSU is doing in your stash.

                    In my case, I had a crappy J&C 235W PSU based on some Meico gutless crap. I gutted the original PSU a long time ago for components. But its case became a new home to a homeless HiPro 250W PCB that someone removed from its original case for whatever reason and left for recycling.
                    Oh well, I'm still using that PSU. It looks completely ghetto.
                    I was actually doing that on a daily basis last year, more or less.

                    450W Frontier - housed a 350W CWT ISO
                    400W Allied - houses a ~400W (realistically) Premier (DEER) I rebuilt in 2019 and looks pretty good inside
                    500W A-Line - houses an unknown wattage Wintech (though I'd estimate around 300-350W, judging by the WIN-35P transformer)
                    450W Torrent Computers (also Deer) - houses a 2012 KissQuiet PSU (original casing had rust, it was a Delux)
                    420W Spire (12cm fan) - houses a KME board w/ ERL39 trafo, beefy as heck

                    And the list could go on. I've also did one for my dad - he needed a replacement PSU for a friend's DVR (as the one supplied with the DVR just blew, everything connected to that PSU survived) so I just whipped up some quick unit, based on a 400W Delux (that originally housed a Sun Pro that died with BIG ass fireworks in my kitchen ) and some KME (though a larger PCB at that, probably from the PL series) board I recapped, added a wire in place of the fuse (I needed the fuse for some other PSU, the KME DID work beforehand), and gave it to him. For a few HD cameras, it was enough, and it probably still runs to this day.
                    Main rig:
                    Gigabyte B75M-D3H
                    Core i5-3470 3.60GHz
                    Gigabyte Geforce GTX650 1GB GDDR5
                    16GB DDR3-1600
                    Samsung SH-224AB DVD-RW
                    FSP Bluestorm II 500W (recapped)
                    120GB ADATA + 2x Seagate Barracuda ES.2 ST31000340NS 1TB
                    Delux MG760 case

                    Comment


                      #30
                      Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                      Since I finally have a big enough electronic load to put 30A on the 12V rail lets see if this PSU can handle it, I put 6A on the 5V rail and 4A on the 3.3V to better represent a modern system for a 403W total load:

                      (note the oscilloscope is on the 100mV per division scale)

                      12V rail (12.003V, ~260 mV ripple):




                      5V rail (5.1524V, ~220 mV ripple) :




                      3.3V rail (3.3987V, ~440 mV ripple):




                      It draws 562W from the wall for a 71.7% efficiency:



                      So it can handle that load without blowing up and voltages are in spec, but with out of spec ripple. Also note when I bumped the 5V rail up to 10A for a 423W total load the unit shut down after about 2 minutes (I assume it tripped the over temp protection since the Weltrend wt751002 has no over current protection and the voltages were still in spec, and it powered back on after letting it sit for a few minutes to cool off).

                      Originally posted by momaka View Post
                      Still not sure if the 12V rail will do 30 Amps continuous with two 20-Amp rectifiers in parallel... but you're probably not that far off. Reason I say this is because most OEM PSU's I've seen with this arrangement (at least on the 5V rail of older ones) usually state around 24-25 Amps max. But those were pretty conservatively rated overall. So I could be wrong on this.
                      It looks like you were correct on that (at least while keeping ripple in spec).





                      Let's dial it back to 350W (25A on the 12V rail, with 6A each on the 5V and 3.3V) and see how it fairs (This time the oscilloscope is on the 20mV per division scale):

                      12V rail (12.078V, ~80 mV ripple):



                      5V rail (5.1366V, ~55mV ripple):




                      3.3V rail (3.3989V, ~25mV ripple):




                      It draws 444W from the wall for a 78.8% efficiency:



                      So this appears to be a fairly solid 350W power supply that is being pushed to (or a little beyond) its limits when loaded to 400W.
                      Attached Files
                      Last edited by Per Hansson; 07-25-2020, 12:46 PM. Reason: Updated ripple numbers from post #32

                      Comment


                        #31
                        Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                        I think there is a small misunderstanding about the ripple, you say the scope is set to 100mV per division in this picture.
                        And then you write: "12V rail (12.003V, ~130 mV ripple)"
                        This is not so, the picture shows double the ripple, i.e. 260mV.
                        Also if the ripple only shows up like that in a picture but not in person (due to shutter time) then it is still a valid measurement.
                        In the past before DSO's existed you could even get a camera baffle for the scope for this purpose.
                        "The one who says it cannot be done should never interrupt the one who is doing it."

                        Comment


                          #32
                          Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                          Originally posted by Per Hansson View Post
                          I think there is a small misunderstanding about the ripple, you say the scope is set to 100mV per division in this picture.
                          And then you write: "12V rail (12.003V, ~130 mV ripple)"
                          This is not so, the picture shows double the ripple, i.e. 260mV.
                          Also if the ripple only shows up like that in a picture but not in person (due to shutter time) then it is still a valid measurement.
                          Some of it is due to shutter lag, I tried turning down the intensity (it was all the way up on the first set of pics) on the scope a little on the second set of pics (with the 350W load) to try and help some (since it is a CRT scope the camera also picks up a "ghost" image from the phosphors that still have energy in them but aren't visible to the eye).


                          You are correct though, I haven't used the oscilloscope in a while (with the Solid Gear 500W PSU review being the first time in several months) and forgot to use peak to peak rather than peak to reference.


                          Making that adjustment puts us right at (or slightly over) the 50mV ATX ripple spec on the 5V rail on the 350W test so that definitely seem to be the "safe" limit on this PSU (though it would probably be best to use it in a system drawing no more than 300-330W).

                          I would edit the post but the edit time limit has passed.

                          Thanks for catching that.


                          Updated numbers:

                          403W: 12V rail (~260 mV ripple), 5V rail (~220 mV ripple), 3.3V rail (~440 mV ripple ).


                          350W: 12V rail (~80 mV ripple), 5V rail (~55mV ripple), 3.3V rail (~25 mV ripple).
                          Last edited by dmill89; 07-25-2020, 10:57 AM.

                          Comment


                            #33
                            Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                            Originally posted by dmill89 View Post
                            Since I finally have a big enough electronic load to put 30A on the 12V rail lets see if this PSU can handle it
                            ...
                            So this appears to be a fairly solid 350W power supply that is being pushed to (or a little beyond) its limits when loaded to 400W.
                            Even so, IMO those are still pleasantly-surprising results. Given the components inside and seeing this PSU manage 400W continuous DC load shows it's probably designed quite alright.

                            Of course, looking at the efficiency between the tests with the 30A load on the 12V rail and the 25 Amp, you see it really starts to tank with the 30 Amp load. This suggests that the continuous DC load limit is indeed around 350W.

                            Not only that, but the ripple on the outputs jumps up quite a bit (though I imagine installing bigger caps can remedy that to an extend... though you're probably already pushing the output toroid inductor too hard, so no point.) I suppose this further supports the notion that this is a 350 Watts (continuous) PSU that can do 400W peak, if needed. So no need to change the label then. Just put "Max Peak" after the 400W rating, and you're all set.

                            In any case, no matter which way you look at it, for a $15 PSU with just a few $ spend on new caps, this is still a really good deal.

                            Comment


                              #34
                              Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                              I updated your post with the new numbers.
                              Also just to reiterate what I said, the "ghost" image or phosphor lag is not really "invalid data".
                              You can think of it like the option "persistence" in a modern DSO:
                              It is actual recorded events, just might not be that many of them vs the "baseline" that sits at a different level.
                              "The one who says it cannot be done should never interrupt the one who is doing it."

                              Comment


                                #35
                                Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                                dmill89, how are you scoping this, what's the load, ground clip length, probe position?
                                if you already wrote about it, please link it....

                                i sometimes feel "dumb" (purely resistive) loads are not the reality of these psu lives', and better idea would be to measure noise when they're powering the stuff they normally do, but the noise on the ground.....oughm....i mean all those big and small smps on the mobo....

                                Comment


                                  #36
                                  Re: The Cheapest ATX Power Supply on Amazon Prime, How bad can it be?

                                  Originally posted by i4004 View Post
                                  dmill89, how are you scoping this, what's the load, ground clip length, probe position?
                                  if you already wrote about it, please link it....

                                  The load amounts (amps) are listed in each post.

                                  For the actual hardware used in the most recent test a Kunkin KL283 300W 30A (15A X2) electronic load was used on the 12V rail, and a pair of the little 60W (10A max) Chinese "loose PCB" electronic loads where used on the 5V and 3.3V rails.








                                  I'm using a flashtree breakout board on the 20/24 pin connector (and I have the 5V and 3.3V hooked into the electronic loads for that), for the 12V I'm using the 4 pin motherboard and 6 pin PCI-E connectors.




                                  The Oscilloscope is an old Goldstar (LG Precision) OS-9020A 20 MHz analogue CRT scope I got cheap on e-bay several years ago.





                                  The probes are a pair of cheap "no-name" probes I got for $11 on Amazon.




                                  For probing I generally clip the ground lead on the probe to one of the ground wires (usually on a molex connector) and then probe the power wire (also usually on a molex except of course 3.3V which I would get from either a SATA connector or the motherboard connector directly), example below:

                                  (May not be a perfect scientific method, but I'm just trying to get a general "ball park" measurement).
                                  Attached Files
                                  Last edited by dmill89; 07-26-2020, 11:30 AM.

                                  Comment

                                  Working...
                                  X