Discussion PSU recommendations and power supply discussion thread - Tom's hardware

[Darkbreeze]

You're assuming that all these companies ACTUALLY adhere to those standards.

Mmm, hmm, yeah, just like there are none of them with overblown specifications or fake 80plus certifications. Right?

Or protections that are properly configured, per specification? Not set way too high, or way too low, or don't exist even though they claim to be there?

Pfffft.

Besides which, per your own post and quote, that is a "guideline", not a specification. Big difference.

Here is the ATX 2.2 specifications, and there is NO such recommendation, guideline or rule about 6-8 amps per pin that I can see.

https://cdn.instructables.com/ORIG/FS8/5ILB/GU59Z1AT/FS85ILBGU59Z1AT.pdf

 

[Mezoxin]

@Darkbreeze
that is a mistake by my side , i had the snapshot on my pc and forgot from which reference i took it , it was from the intel 002 design guideline at item 2.2.1
https://www.intel.com/content/dam/w...nts/guides/power-supply-design-guide-june.pdf

of course other PSU's that are of low quality are totally disregarded from the discussion , i was talking about psu's from reputable manufacturers who have credibility and actually implement the guidelines as i see from the reviews and cable runs that i attached

 

[Mezoxin]

Since they all go back (especially true for daisy chained 6/6+2) to the same source, even on most multiple rail psus, whats available to each pcie will be what's available to all pcie in total. So technically you could try to pull all 40A+ through a single 6pin, which is where SCP, OCP, OPP should be kicking in to prevent such from happening.

that was very informative thank you , I have always had a question about that 40A 12v OCP that i see set at multirail PSU's , because pci-e cables only at best has 3 hot 12v lines this means that they are rated for a maximum of 24A /288w so OCP in multirail when set at 40A won't protect that cable from damage , is this correct ?
neither will it protect the EPS cable 2x4 pin which is rated at a max of 32A/384w
so what difference does it make to have multirail 12v OCP set at 40A ?

 

[Mezoxin]

Fiirst 12VO psu ever released in accordance with intel design guide
https://www.intel.com/content/dam/w...ower-supply-platform-atx12vo-design-guide.pdf

hopefully we see compatible motherboards released soon

View: https://youtu.be/iNCfnKEoRU0

 

[Karadjgne]

That's going to be rough since drives and argb are run on 5v, the logic circuitry runs on 3.3v. A 12VOnly psu is going to have limited application.

 

[Mezoxin]

That's going to be rough since drives and argb are run on 5v, the logic circuitry runs on 3.3v. A 12VOnly psu is going to have limited application.

motherboards are going to do the DC-DC conversion and output 5v and 3.3v , some pre built systems have that already ,but i think its not very practical or sellable to the custom pc end user though

 

[Karadjgne]

I just don't see that happening. With power requirements, all the added bling like multiple m.2 slots, rgb/argb, networking needs, pcie, pch, traces etc there's little to no space on a mobo left. And you want to add at least 2 full rails? And apart from that, who is going to build them? Out of what? With what protections? Going to sacrifice a pcie slot in the lower left of the mobo to add a giant chunk of plastic to cover the lethal amperages those rails can produce?

Just what I always envisioned. A psu attached to my psu, sitting smack dab under my gpu.

 

[Darkbreeze]

It's VERY unlikely that motherboard manufacturers are going to want to have to shoulder the additional requirements themselves, or the cost. This will likely end up being just another niche Intel product, like Optane. I don't see it ever happening. At least, not anytime soon. Especially with AMD about to cut the legs off Intel using millions of small razor cuts. They'll still be there, but hard to do much when you can't walk anymore.

AMD is sure as heck not going to be willing to change THEIR standards and designs around to incorporate Intel requirements and neither are their board partners. No go IMO.

 

[Eximo]

Literally what Dell/Foxconn has been doing for years. 5V and 3.3V supplied from the motherboard. You actually plug your SATA drives into it for power. Not a lot of juice though.

And these wouldn't be lethal rails? If incoming is 12V, already 'non-lethal' You simply have a somewhat efficient DC-DC on board. Not taking the mains voltage and doing the conversion on the board. Besides, already happening on the motherboard. 12V -> ~2v -> ~1.2 volts for your CPU, GPU, etc, huge amperages. Lots of 12V only automotive PCs already as well, most NUCs. My little Atom stick computer runs on 5V.

One could argue, why 12V? Rewind the fans and stick with 5V LED strips, what is left that actually uses 12V?

Honestly not the worst idea ever. I think you would have pushback from the PSU manufacturers. Not much competition in making a single voltage PSU, whoever makes the most efficient or cost effective components would the winner.

It would tickle me to see a giant like 8 gauge receptacle to the motherboard for +/-. Maybe a keyed Anderson connector so it couldn't be screwed up. All the PCIe cards just be plug in with no cables. Apple is over there running PCIe connectors through the motherboard, so already sort of happening.

 

[TJ Hooker]

Aris [...] he is an electrical engineer with 3 degrees and a PHD and decades of industry experience.

As an aside, where are people finding Aris' credentials/experience? I remember briefly searching for him (google, linkedin, cybenetics page etc.), but didn't really find anything. Just his Twitter and TH pages, which just link to his TH articles.

 

[TJ Hooker]

yes the tech has to be present in order to discover its shortcomings , but it doesn't have to be released to discover ALL of it , as there is a lot of predictive analytical tools available for that
Back in the dark ages the only way of predicting the future was by words of oracles and superstition , but ever since Pascal and Fermat founded the basics for which would be later known as probability theory , there has been a scientific method of predicting the future , and although they were made to solve a puzzle and win games of chance, later on their first business application was with life insurance policies in the uk 18th century and has been the founding stone of what is known as risk management , whether in finance or reliability engineering or safety and also for conducting valid researches and linking a cause to an effect with enough certainty
So not everything we see in safe design or protection features are reactive most of them are predictive as well

Using statistics to predict mortality rates, failure rates, etc. is totally different than predicting which direction technology is going to move in the future...

 

[Karadjgne]

Ah, Eximo, 😉 it's not the voltage. Anything more than 88mA and you loose the ability to 'let go' of DC voltage, and 500mA will produce fibrillation in 0.2 seconds. Both the 5v and especially the 3.3v rails deal with amperages well in excess of those, and if those rails were open air on a pcb, well that's a disaster in the making with all the caps used etc. The box the psu guts come in serves a higher purpose than just a place to park a sticker.

 

[Mezoxin]

Using statistics to predict mortality rates, failure rates, etc. is totally different than predicting which direction technology is going to move in the future...

what i meant is that it doesn't need to be released , not invented , when everything is still on paper you can use probability theory and statistical inference to predict the future performance of that technology , and what are the possible failures and their ultimate consequences , and decide which are credible scenarios and which aren't , and accordingly alter the design in a risk based manner , before procurement or production happens and this is how most of the safe design and protective features are determined

you can find aris credentials here
https://landing.coolermaster.com/pages/cybenetics/

Honestly not the worst idea ever. I think you would have pushback from the PSU manufacturers. Not much competition in making a single voltage PSU, whoever makes the most efficient or cost effective components would the winner.

for the PSU manufacturers its easy and cheaper , motherboard manufacturers and end users are the problem here , from the motherboard side in addition to increased costs and point of failures , it's currently very crowded with no space left at all around the edges to accomodate for additional power connectors so they can be routed nicely behind the tray and allow for cable management , and the end users are all going toward rgb and tg panels which means any additional cables showing are not that welcomed ( unless the make them RGB power cables 😛)

It's VERY unlikely that motherboard manufacturers are going to want to have to shoulder the additional requirements themselves, or the cost.

I agree and even if motherboard manufacturers did they will not find a buyer for it , but i just see it as trend that has been going for a long time in some major prebuilt system manufacturers like HP and Dell so intel decided to standardise that practice which is a good thing , but i was surprised when fsp released a consumer product .
And AMD is sweeping the rug from under the feet of intel as you mentioned may weaken their ability to force that standard

Ah, Eximo, 😉 it's not the voltage. Anything more than 88mA and you loose the ability to 'let go' of DC voltage, and 500mA will produce fibrillation in 0.2 seconds. Both the 5v and especially the 3.3v rails deal with amperages well in excess of those, and if those rails were open air on a pcb, well that's a disaster in the making with all the caps used etc. The box the psu guts come in serves a higher purpose than just a place to park a sticker.

But its Voltage that determines How many AMPS will be pushed into your body after dealing with both your Skin's Capacitance (in case of DC) and Resistance and cause the adverse effects you mentioned
DC at 12V wont be able to push that many amps through your skin capacitance and resistance

 

[Darkbreeze]

So, per Jon Gerow.

whats available to each pcie will be what's available to all pcie in total.

And

OCP in multirail when set at 40A won't protect that cable from damage

so what difference does it make to have multirail 12v OCP set at 40A

I've never seen a unit that had any kind of OCP set for a single or even multiple connectors fail or burn up connectors. I have, however, seen connectors burnt when they didn't have OCP.

Let's put it this way: A normal PC with OCP on the +12V set at40A would never have a problem. Spikes go up to 40A, which is unusual, and the PC keeps running.

Now look at something like a mining rig. 40A load on a single +12V rail and after time the connector melts. Why? Because the load is constant and that's beyond what the pins can handle.

 

[Mezoxin]

Yes 32A for 2x4 eps and 21 to 24 A for pcie cables is the max rating for sustained load , 40A was done to accomodate for spikes , most possible scenarios that would cause sustained high amperage would most likely be (as johnny guru said in one of his videos) resistive loads formed by pinched wires or broken connectors , I now understand that is because these type of failures will most likely continue to draw a high amount of amps much higher than 40A so 12vOCP would work well in mitigating those effects
When i asked johnny guru recently on his forums he said that the optimal method would be having a time dependant ocp setpoint but that would require extra cost of hardware or psu with a dsp

To some extent, you can build in a current / duration circuit protection, but dialing that in requires more (expensive) components. A PSU with DSP is really the only way to properly nail down allowing a maximum current for only a predefined period of time to prevent transients from tripping OCP.

 

[Darkbreeze]

They are not "his forums". Not for a long time now.

 

[NightHawkRMX]

So I bought a secondhand TT SMART 80+ 600w for $10 off of a friend today just for my LGA775 system in order for me to put my other HiPro PSU back into my NAS.

I was actually surprised. The PSU felt heavy and so far has been fairly quiet, although not under much load.

I have read the reviews, the spec sheets, and seen how medeocre this PSU actually is, but I could see how someone could get fooled into thinking this PSU is as good as a HX850i since they see a recognizable brand, 80+ sticker, and the PSU looks fancy.

 

[Karadjgne]

You can Gold electroplate it. Dip it in beer batter and deep fry it. Roll it in diamond dust, cover with Royal glaze and Godiva shavings. But at the end of the day, it's best friend is still a wad of toilet paper...

 

[Darkbreeze]

Hanging out with his friends the door and the floor is where that PSU belongs. IOW, as a door stop. But yes, it is definitely better than SOME other units out there, especially a lot of what is available in very limited infrastructure areas that have mostly local region fire starter quality brands that you'd never much (Or at all) hear of outside that specific region.

Better, by far, than a Linksys, or Bestec, or Ultra, or Apevia, or Supercase/Allied or Chieftec, but being better than those or the tons of others out there LIKE them, isn't saying a whole lot really. For a light powered system it's probably perfectly fine. I just wouldn't use it with a graphics card that requires more than slot power.

 

[Jeffsta]

Is anyone familiar with the EVGA SuperNova 850 G3?

In the introduction to this thread, the G3 is said to be a good EVGA model. I was reading the reviews and the EVGA SuperNova 850 G3 is 5 eggs with very few negative reviews. The few negative reviews suggest that the fan is loud, but EVGA responded that they must have gotten a faulty unit. The positive reviews say the unit is quiet.

1. Would it be a decent PSU for a new build? I looked at Seasonic, but they are out of stock on newegg.

2. Will it's "Compact 150mm Size" fit properly into my Case?

 

[Darkbreeze]

That PSU is fine. Yes, it will fit your case. Are you sure you NEED an 850w unit? What graphics card and CPU are you running? If you start a thread and post a link to it here I'd be happy to help you make certain you get not only a good unit, but the right unit.

 

[Jeffsta]

That PSU is fine. Yes, it will fit your case. Are you sure you NEED an 850w unit? What graphics card and CPU are you running? If you start a thread and post a link to it here I'd be happy to help you make certain you get not only a good unit, but the right unit.

I'm not sure actually. You and I were talking in another thread about adding a GPU to my PC. At the end you were mentioning I might want to try to save up for a couple months and upgrade my platform. I plan on repairing my Dads computer with my 5 year old PSU, then upgrading to whichever PSU you think I should put in for a future platform upgrade. I started a thread earlier here about which PSU I should get. I appreciate your help.

 

[Darkbreeze]

I popped in there and made a recommendation.

 

[Mezoxin]

Why you shouldn't use 4pin molex to (6+2)8 pin pcie connectors , specially if they are well made
This is a great example of a person that bought a quality adaptor that took the beating , but sadly the single 12v pin in the back of that psu couldn't , because its not made for a sustained load greater than 8A , Ironically he might have been better off buying a crappy adapter that would melt before the pin in the psu

 

[Darkbreeze]

Why would anybody convert PCI to Molex? Systems already come with plenty of molex connectors, and there are very few things that use them anymore so I really doubt they could have USED them all up to the point where they needed to do that.