GIGABYTE Releases Statement on Allegedly Explosive Power Supplies, Will Offer Exchanges

[Paul_Johnson]

Yes, UL certification does exist for ATX12v/EPS, along with just about anything else you can think of. UL certification is specifically for safety.

Corsair, for instance, does UL certify almost all of their PSUs. Gigabyte, at least in this instance, appears to have not.

https://www.corsair.com/br/ko/blog/...iple-rail-psus-and-safety-it-serious-business

UL certification has nothing to do with OPP.

 

[Brian_B]

Just because ATX12 doesn't specify something like not exploding doesn't mean some other body or specification can't require that. You can (and probably should) make something fall under more than one specification and standards at a time.

Now, I will let you figure out exactly how many power supplies are not spec compliant for OCP that you guys love and have no problem with.

As long as it doesn't blow up I guess it isn't a problem.

UL certification has nothing to do with OPP.

I would beg to differ, at least tangentially, but I think this is one thing you and I just aren't going to see eye to eye on. My take is a PSU should fail gracefully, even when it sees something bad. You asked specifically for a cert for safety, I provided two with UL and IEC - but they are by no means the only ones available to test for safety and compliance.

If you want to get hung up on one particular feature and specification and defend Gigabyte here, feel free, but the bottom line is that I don't think you will change my opinion that a PSU or any other piece of equipment shouldn't destroy itself when presented with a plausible abnormal condition. You can't make them bulletproof - there will always be extraordinary conditions, but something like a short circuit or "faster than a short circuit" slew rate that seems to exist - those are definitely plausible and have been considered basic features in PSUs (not just ATX12v/EPC, but pretty much every power supply of various types I've ever used) for a long long time.

 

[Paul_Johnson]

Just because ATX12 doesn't specify something like not exploding doesn't mean some other body or specification can't require that. You can (and probably should) make something fall under more than one specification and standards at a time.

Except, there are not other specifications that are required to apply.

As long as it doesn't blow up I guess it isn't a problem.

So being "out of" a non-existent specification is worse than being out of an actual specification. Uh.......the logic ball there is mindbogglingly dense.

I would beg to differ, at least tangentially, but I think this is one thing you and I just aren't going to see eye to eye on. My take is a PSU should fail gracefully, even when it sees something bad. You asked specifically for a cert for safety, I provided two with UL and IEC - but they are by no means the only ones available to test for safety and compliance.

Neither of which are relevant.

If you want to get hung up on one particular feature and specification and defend Gigabyte here, feel free,

I am not. I already reviewed one them, they were not good and I said so. What I AM DOING is explaining to those you who don't know what is going on what is and it is very simple. Do not run a product way outside of spec and when it blows up make the surprised Pikachu and try to blame it on a non-existent requirement. With no design guide or requirement there is no way to design such a protection to catch all covered scenarios because there are no limits on how you approach it. It is a VERY simple concept. If OPP had a required design limit and slew rate and test parameters and someone tested it under that then there would be something to this other than we can blow up a unit by pushing it as far out of spec as possible. However, without that I might as well test the units function when feeding it 480V three phase and make the surprised pikachu phase when it dies.

but the bottom line is that I don't think you will change my opinion that a PSU or any other piece of equipment shouldn't destroy itself when presented with a plausible abnormal condition. You can't make them bulletproof - there will always be extraordinary conditions, but something like a short circuit or "faster than a short circuit" slew rate that seems to exist - those are definitely plausible and have been considered basic features in PSUs (not just ATX12v/EPC, but pretty much every power supply of various types I've ever used) for a long long time.

And they have defined protections and parameters in the design guide. We aren't talking about that here though. You can keep trying to make it seem like it is to make yourself feel righteous in your bashing but what you are bashing isn't actually a thing.

 

[Zarathustra]

Now, isn't that the crux of the problem. Lots of folks like you Brian, Gamer Nexus, etc. commenting on what should happen, how things are supposed to work, and what appropriate behavior is yet do not know anything about what the design requirements are. For instance Brian_B is going on about OPP and how it handles shorts. It does not. OPP is not even a required protection under the ATX12v or EPS design guide so it may be implemented however a manufacturer wishes. What does control shorts is the SCP which is defined in the design guides. I have have a nifty little button on the load tester to test the protection that drops the resistance to below 0.1 Ohm. OPP? Well yes a test for that too, but it has a slew rate defined by the load tester NOT a specification.



Not when there are design parameters required because it is impossible to design for every instance when there are no rules to play by.



Literally impossible to design out idiots. There is no way to design a unit for every failure or overload mode. That is why there is a design guide that establishes what unit has to do in order to be compliant. "Overload safety" is not a defined thing in this case. Again, it goes to people are commenting on what they don't know about. OCP does exist....and the spec calls for 240VA per rail with a defined slew rate. Now, I will let you figure out exactly how many power supplies are not spec compliant for OCP that you guys love and have no problem with.

Well, hold on just a moment.

I may not be familiar with the specifics of the standards that apply to power supplies, but I have been in product development my entire career.

You start with a set of user requirements. Those generally include a bunch of safety precautions. There is a risk management process designed to uncover additional risks that had not previously been considered. These are also added to the requirements.

In any industry the established standards are a bare minimum, not what you design to as a target.

In my industry minimum electrical safety standards are set by IEC 60601-1. That doesn't mean anyone designs to the standard and then washes their hands of everything else. As part of the user needs process they look at their own returns and failures and determine what kind of problems they have had in the past, and they add in their own internal design requirements to mitigate them.

Standards are nothing but bare minimums, and you know what they say about people who do nothing but the bare minimum?

Gigabytre can complain about how they are testing all they want, but there is a reason these two 750 and 850 watt Gigabyte units are failing spectacularly in testing, and others aren't. Maybe they weren't conservative enough in setting their design targets. Maybe they made a mistake in a calculation that didn't get caught. Maybe they made a mistake in trusting a sub-contractor they shouldn't have. I don't know the details, but the truth is their power supplies are a problem, and others are not.

I don't want to point fingers enough. You spend long enough in any industry and even the best company has a screw-up. That is inevitable. The difference between the good and the bad - however - is how they behave when things go wrong. Thus far Gigabyte is not off to a good start.

Blaming a standard, or a lack thereof is a cop-out. As a designer of any product, you are supposed to know your customer, know their usage scenario and design to match. You design your product for it's actual use condition to the users need, and mitigate the risks the user is likely to see. The standards are just bare minimums. You never rely on standards to ensure you have a good product.

You develop your user needs, you design to those needs, and then you test to make sure you have met those needs. It is your responsiblity to determine what the needs are, and make sure you have designed a product that works, standards or none.

 

[Paul_Johnson]

Well, hold on just a moment.

I may not be familiar with the specifics of the standards that apply to power supplies, but I have been in product development my entire career.

You start with a set of user requirements. Those generally include a bunch of safety precautions. There is a risk management process designed to uncover additional risks that had not previously been considered. These are also added to the requirements.

In any industry the established standards are a bare minimum, not what you design to as a target.

In my industry minimum electrical safety standards are set by IEC 60601-1. That doesn't mean anyone designs to the standard and then washes their hands of everything else. As part of the user needs process they look at their own returns and failures and determine what kind of problems they have had in the past, and they add in their own internal design requirements to mitigate them.

Standards are nothing but bare minimums, and you know what they say about people who do nothing but the bare minimum?

Gigabytre can complain about how they are testing all they want, but there is a reason these two 750 and 850 watt Gigabyte units are failing spectacularly in testing, and others aren't. Maybe they weren't conservative enough in setting their design targets. Maybe they made a mistake in a calculation that didn't get caught. Maybe they made a mistake in trusting a sub-contractor they shouldn't have. I don't know the details, but the truth is their power supplies are a problem, and others are not.

I don't want to point fingers enough. You spend long enough in any industry and even the best company has a screw-up. That is inevitable. The difference between the good and the bad - however - is how they behave when things go wrong. Thus far Gigabyte is not off to a good start.

Blaming a standard, or a lack thereof is a cop-out. As a designer of any product, you are supposed to know your customer, know their usage scenario and design to match. You design your product for it's actual use condition to the users need, and mitigate the risks the user is likely to see. The standards are just bare minimums. You never rely on standards to ensure you have a good product.

You develop your user needs, you design to those needs, and then you test to make sure you have met those needs. It is your responsiblity to determine what the needs are, and make sure you have designed a product that works, standards or none.

And the one thing that negates your entire argument is this......if they took the established standards as a bare minimum and did just that then they would not have even implemented any form of OPP. Which is not required. So, that kind of sinks your argument in port.

 

[Brian_B]

And the one thing that negates your entire argument is this......if they took the established standards as a bare minimum and did just that then they would not have even implemented any form of OPP. Which is not required. So, that kind of sinks your argument in port.

It means they didn't do it right, and the fact they felt the need to have it in the first place speaks more to it than it not being in a specific standard (just one among the ~nine~ various standards they purport to follow, based on the product specification page), I think.

 

[Brian_B]

Geez, the more I read on their web page, you'd think this was the most protected PSU made in the world:

• OVP/OPP/SCP/UVP/OCP/OTP protection

The good thing is it appears they are honoring their 5 year warranty.

 

[Paul_Johnson]

It means they didn't do it right,

No, you can't say they did not do it right. Since there is not set specificaiton for it there is no way to test it appropriately.

and the fact they felt the need to have it in the first place speaks more to it than it not being in a specific standard (just one among the ~nine~ various standards they purport to follow, based on the product specification page), I think.

You are hilarious. I mean this is comical. We have gone from they "didn't do something that they should have" to now "they did something they felt they needed to by even trying to implement OPP which is worse". WOW! The cognitive dissonance here is really getting into the absurd.

You guys need to like go take a break and figure out what you want to raise your pitchforks about.

 

[Denpepe]

I'm not familliar enough with industry standards, but it is appropriate to have overdraw protection and to have PSU's fail gracefully when they are overdrawn, not to pop in a display of thunder and lightning.

I guess that's a result of the RGB craze

I for one used to like GN's testing, but I kind of lost interest, Steve always seems to want every piece of hardware made the way he wants it to be made or it's crap even if it works fine for what is intended.

And he is soo god **** long winded, he has that TV thing down where he tells you constantly what's comming later more then giving actuall info. And he loves beating dead horses to a pulp etc...

I can see him pour gasoline of something light it on fire and call it a design flaw when it goes up in flames as that would be a reasonable usecase for him.

Anyways, I'm wayy of topic here so more on topic I do agree that these things should fail gracefully if used as intended.

 

[Paul_Johnson]

Geez, the more I read on their web page, you'd think this was the most protected PSU made in the world:

That is standard advertising copy for almost every power supply made. However, I can guarantee you that the OCP part is not in specification on the majority of power supplies. They may have it and set it higher than allowed or they may omit it and rely on OPP. No one seems to mind that though.......

 

[Brian_B]

You guys need to like go take a break and figure out what you want to raise your pitchforks about.

Totally ready to pounce on something more entertaining.

You lost me at "slew rate for short circuit" and then got hung up with the literal definition of OPP, and don't think I'm coming back. But I enjoy the discussion. I just used the term "OPP" because it was in the main article, and it fits, so long as you don't get pedantic about it.

I mean, what you are really trying to argue is the testers over at HN did a poor job of testing it. But it's coming out like your defending Gigabyte. And it's a tough position, because... not only did the PSU blow up, it did so spectacularly on video, and it was repeatable in a statistically meaningful fashion.

I think most of the rest of us are saying - that isn't right. Paul seems to be the only one arguing about an ATX specification - no one else really cares, other than we know our PSUs shouldn't throw sparks, and at least from the video, it doesn't look like GN did anything too absurd like plug it into 480V 3p. Maybe they did test it wrong, but no one is coming out with specifics, and apart from the Press Release from Gigabyte that was in very generic terms, no other indication that it was abnormal testing that caused these failures.

I mean, Paul, if you want to come back with something other than "ATX Standard" I'm all ears, but you haven't really put out anything yet that supports your position that it was tested improperly and that's why they fail.

Attacking my logic as circular and absurd is just a weak straw man really. As true as it may be, a better argument is always made with facts.

Also, I agree 100% with Denpepe here as well - the term "Explode" is very hyperbolic and the entire thing is just clickbait for GN - but it did throw sparks, which is entertaining to watch at the least, and most PSUs don't fail in such a high profile manner. So as weak as this is, it's still the most entertaining thing we have in this scalper/miner GPU-devoid messed up logistics COVID world to talk about.

 

[Paul_Johnson]

Totally ready to pounce on something more entertaining.

You lost me at "slew rate for short circuit"

That would be because you don't do this work.

When I set the test here:




The SCP test drops the resistance at a set rate until it meets the spec limit of 0.1 Ohm. It is not "instant", it is a measurable time that you can see in your waveform output and in the programming file.

and then got hung up with the literal definition of OPP

Literal defintions of technical terms are the only kind of definitions. And there is no way to test OPP to a standard. If i hit this button:




It increases the draw in set steps that are defined under StepI for a period of Step T until it reaches OPP End I in the programming rinse and repeat until it reaches 1560W or the unit blows up. That, however, is not found anywhere in the specification and it is in steps that plateau and hold before moving to the next. Also, not a Short circuit protection scheme like you said OPP should be but that is how it is tested on load testers.

and don't think I'm coming back. But I enjoy the discussion. I just used the term "OPP" because it was in the main article, and it fits, so long as you don't get pedantic about it.

But you didn't know what it was or what is actualy being tested here or how.

I mean, what you are really trying to argue is the testers over at HN did a poor job of testing it.

I literally keep telling you there is no way to test OPP appropriately because there is not spec for it. I have said that 3 or 4 or 5 times now. How you keep missing it, I don't know.

But it's coming out like your defending Gigabyte. And it's a tough position, because... not only did the PSU blow up, it did so spectacularly on video, and it was repeatable in a statistically meaningful fashion.

I reviewed the same unit before they did and told you it wasn't a good unit. Flip side, Aris did approve it at Cybenetics so that is a doozy. LOL

I think most of the rest of us are saying - that isn't right. Paul seems to be the only one arguing about an ATX specification - no one else really cares

Ah, now we don't care about standards when earlier all you wanted were more standards? It is just boggling the mind to figure out how you keep twisting in the wind here.

other than we know our PSUs shouldn't throw sparks, and at least from the video, it doesn't look like GN did anything too absurd like plug it into 480V 3p.

And that is because you don't know what the correct testing procedures are or what it is they did exactly. You are literally commenting about something you 1) don't know anything about, 2) don't know what they did, and 3) won't listen to the documentation that tells you what is what.

 

[Brian_B]

I literally keep telling you there is no way to test OPP appropriately because there is not spec for it. I have said that 3 or 4 or 5 times now. How you keep missing it, I don't know.

Because there are a lot of specs - I've mentioned at least two of them. Just not in ATX12v, and not on your one special fancy tester. We test short circuit (among a lot of other various types of) protection all the time, just not on computer power supplies.

Ah, now we don't care about standards when earlier all you wanted were more standards? It is just boggling the mind to figure out how you keep twisting in the wind here.

No, I care very much about them. You only care about the one specific standard your tester happens to support though.

That would be because you don't do this work.

You are right, I don't test power supplies. I test large scale generators that hook up to the national power grid, and have to meet a lot of different standards and fight a lot of larger and more difficult problems than computer power supplies ever dream of. So I don't know a whole lot about Computer Power Supplies, but I wouldn't say I'm entirely ignorant when it comes to electrical protection in general, and electrons don't care a whole lot about where they came from.

 

[Brian_B]

I'll even go a bit futher

Each one of these:

• OVP/OPP/SCP/UVP/OCP/OTP protection

They have a standard. Otherwise they wouldn't exist. If it's not from a regulatory body or an industrial standard, then at the very least it's an internal Gigabyte standard.

Just because it doesn't exist in ATX12V, and you don't have a button you can push on to test it, doesn't make it irrelevant. There's a spec for it someplace.

And, going out on a limb - as a reviewer, semi-professional at that - I would think it would be in ~your~ best interest to find out what that was supposed to be, and why it's failing. Not just saying "Eh, it's not required". You should ask Gigabyte. You should be talking to the folks at GN and try to get more info. You should be digging deeper.

I'm just an armchair quarterback with circular logic, and I'll throw the "not my job" card at it - I've invested about as much research into this as I'm willing to devote in my spare time (*with the caveat I have ~plenty~ of time to drink beer and continue to armchair quarterback). Which, I suppose, you would be entitled to as well, and I wouldn't blame you if you did, as I'm sure the pay isn't as great as it should be, unfortunately. I think you do a fantastic job in your reviews, I just hate to see you ostrich your head in the sand and pass up a good opportunity to dig a bit deeper, especially if they only reason is because you don't have a button for it on your tester.

 

[Paul_Johnson]

No, I care very much about them. You only care about the one specific standard your tester happens to support though.

No, I care about the one that pertains to my product segment. That is what we work with so that is what we do.

 

[Paul_Johnson]

I'll even go a bit futher

Each one of these:

• OPP

They have a standard. Otherwise they wouldn't exist. If it's not from a regulatory body or an industrial standard, then at the very least it's an internal Gigabyte standard.

No, it doesn't that is what I have been trying to tell you. It is a feature, but it does not have a standard. Thus, it is not testable to a standard. SO, I can make up whatever I want and say it passes or fails but that is not a valid test.

 

[Brian_B]

No, it doesn't that is what I have been trying to tell you. It is a feature, but it does not have a standard. Thus, it is not testable to a standard. SO, I can make up whatever I want and say it passes or fails but that is not a valid test.

No, there are many standards which could pertain. You just don't know which one it is, and your stuck on the one standard you know of.

In this case, I'll do some homework for you:

We were made aware by third parties of concerns regarding potential issues of the GP-P850GM and GP-P750GM tripping at high wattages when tested via DC Electronic Load equipment for extended lengths of time repeatedly close to the 120 percent to 150 percent OPP trigger point
...
Gigabyte is now adjusting the OPP trigger point on both the GP-P850GM and GP-P750GM, with the 850W model now operating at a maximum of 1050W instead of the 1300W limit before. The 750W model has also been reduced to an OPP trigger limit at a maximum of 925 watts, much less than the 1125W limit before.

In this case, it was an internal Gigabyte standard (but that doesn't preclude the fact that it may also be related to an external regulatory standard, that I haven't done enough research on, and I probably won't unless i start building my own PSUs). It has a number you could test to. I stumbled across this here:

Gigabyte to replace ‘exploding’ PSUs that Newegg bundled with RTX 30-series cards

Gigabyte is offering a return and exchange service.

www.theverge.com

 

[Grimlakin]

I came in here to read about the power supplies. Instead I got a bazillion posts about how Mary Jane is right because X and how spider man is right because of Y.

Simply put if Gigabyte didn't have a **** power supply on their hands rife with failures in the field, AND someone didn't blow this up in a highly viewed public manner they probably wouldn't have bothered replacing anyone's PSU's.

From a company stand point they are in the hook and were on the hook, figured out they were on the hook so are doing all they can to mitigate damage to their bottom line and to their brand.

I suspect they won't be releasing power supplies that fail when a video card power demand spikes hard. (Or whatever the hundreds of power supply failures are reported on Newegg alone before any youtube video was published with tons of views.)