Beware of Exploding Gigabyte PSU's being Dumped by Newegg in Forced Bundles

Schro

Supreme [H]ardness
Joined
Mar 15, 2002
Messages
7,803
But there is a specification. Someone designed the PSU, and the OPP is supposed to do something, even if it isn't an industry standard.

The specification is just not public, and may differ from brands, but it is there. In any case, the GN test seems fairly straight-forward and not tainted.
This thread has come full circle a few times... My observations are as follows:
1. Everyone agrees the GM series PSUs are not ones that we want in our systems (whether you trust NewEgg reviews, or have read actual reviews of the unit's performance under load)
2. People seem to equate criticism of the OPP test to calling the power supply good when that is not the case.
3. The main issue with the OPP test appears to focus on borking power supplies by creating their own arbitrary test case. (to "figure out" why there have been so many DOA reviews on NewEgg)
4. The PR response from implicated parties leaves much to be desired.

I like they are extending testing to include Short Circuit Protection (which I think is very straightforward and very critical for a PSU to protect against), Over Power Protection and so on.
I'm still not convinced they've thought that all the way through. They are not experts in electrical safety, but they're going to make an assertion to their readers about the safety of the power supplies that they review without having industry defined specs to test against, without testing a statistically valid sample of units and probably without disclosing what they're doing to their worker's comp carrier.

So, given those points - if they do short circuit protection on one unit, declare it works fine, someone relies on that who then has their house burn down due to a unit they bought because of the review, the reviewer would likely lose a negligence case for not putting enough work into the assertion (even though such work is not cost effective for the reviewer).

Safety is a high liability thing to opine to - while it IS important to know that a power supply is safe, I'm not sure it's something that's appropriate to be in scope for reviews like these (which have a goal of determining whether a unit meets a standard and/or what the label on the box says it can do). Of course, most people in this thread will likely disagree with that scope, but does it really make you feel any safer to have a sample of one unit not go boom when tested?
 
  • Like
Reactions: noko
like this

Dvater

n00b
Joined
Jan 18, 2021
Messages
49
Man, I haven't even used Newegg in like 10 years. Once they started doing that marketplace type shit mixed with their other issues over the years I just simply gave up on them.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
6,644
That Spectre guy just wants to continue to get free shit from Gigabyte.
I very much doubt it. The cost of the PSU is pennies on the dollar over the total cost. Now I would think having a company submitting something for independent testing allows a certain level of repertoire to get answers and clarifications and iron out issues. If I was a company, I would use professionals to do independent evaluations of my products to make sure it is ready for prime time. It would be a very effective Quality Control process independent from bias and could help me alleviate potential disasters like what Gigabyte is and will further go through. Frankly the company has much more to gain then the reviewer, UL testing is not cheap but it is also not all inclusive as in it just checks for the safety of the product, not how well it will work in a PC for operations.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
6,644
To be fair, yes I think GN mentioned they tested about 10 units of the GB PSU and not all of them exploded, so I'll give Spectre credit there. I think 3 or 4 of them did explode, not 100%, but also not a great ratio either.
What is a good ratio? Any failures? One unsafe power supply is ok? My point is GN really needs to put a disclaimer on any safety related checks saying something like: They are not a professional safety check laboratory as in UL and others, that if the power supply passes their tests does not mean that the power supply will be safe for your use. UL and others are the certifications that one needs to look at for those concerns. GN only is reporting what they find with the tests, which are very limited and not full scope safety checks and for user information only. That when they fail a power supply may not indicate one way or another the safety of the tested power supply for users use or configuration.
 

GoodBoy

2[H]4U
Joined
Nov 29, 2004
Messages
2,187
People keep debating about testing methodology when Amazon reviews is full of DOA complaints
User reviews aren't worth shit.

I don't even know why Amazon allows users to review a PSU, as they are not experts with the required minimum 15 years experience in testing said power supplies...
I didn't watch the Gigabyte testing vid, but I did watch part of the EVGA testing. Now I'm seeing [Spectre]'s point on methodology. "We're going to from 20% to 60% to 100% load at 1 minute increments, then over power by 10% at a time every 10 seconds." Why? What's the point or rationale behind giving 1 minute at in spec voltages, and cranking up to the next step every 10 seconds at out of spec voltages? Documentation may say what OPP is supposed to do, but there is no documentation spelling out HOW to determine what it does.

This doesn't mean these Gigabyte power supplies aren't shit, and I don't see anybody here saying they aren't shit. I still say Gigabyte's and Newegg's handling of the whole situation is shit, from bundling these shitty power supplies with video cards to not recalling both models. I can see the argument, though, that GN's testing isn't the way to prove these power supplies are shit.
Anything that simulates a potential use-case, is something the power supply should be able to handle. Boot up your pc - 20% load for a few minutes... fire up some apps - 60% load. fire up Prime95, 93 to 100% load with the 3.3v line feeling it, fire up a demanding game, load could exceed 100%, with the 12v rails feeling it.
See how the power increases, depending on what the user is doing?
It doesn't really matter what the user does, the PSU should handle it or safely power off when it can't.
You can put Prime95 in the startup folder, so the pc boots to 100% cpu load. Who cares? "Well, the power supply can't do that. There is no industry standard on bootup load, and bootup load is mentioned NOWHERE in the ATX spec. Full Stop. So, invalid." pfft..
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
6,644
Zero explosions.
Exactly, now how many samples are needed to have a confident answer? Gigabyte 10+ samples -> Fail. EVGA and AD, 1 sample each -> ??? Has GN really showed anything confident wise dealing with OPP with EVGA and AD with a sample size of 1 each? I don't think so. GN could have had 5 samples from Gigabyte pass their OPP test, would the story then be different? If GN had 10 samples of EVGA all Pass, then that has a better level of confidence in EVGA QC and Power Supplies especially if the PSU were randomly obtain and are from different lots/batches etc. They really do need a disclaimer, will they get sued? Possibly. Except in this case Gigabyte PSU actually does suck and are probably dangerous.
 

bigdogchris

Fully [H]
Joined
Feb 19, 2008
Messages
18,441
That Spectre guy just wants to continue to get free shit from Gigabyte.
Well, he could have chosen to not say anything, if this was the case.

People can just ignore him and buy what GN recommends if they want. However, Steve can come off as pretentious and that can turn off people to respecting his work, which seems to be the case here.
 

ZodaEX

Supreme [H]ardness
Joined
Sep 17, 2004
Messages
4,122
Well, he could have chosen to not say anything, if this was the case.

People can just ignore him and buy what GN recommends if they want. However, Steve can come off as pretentious and that can turn off people to respecting his work, which seems to be the case here.

I only ignore the most douchey of douchebags here. Spectre isn't a douchebag, just a very opinionated shill.
 

cybereality

[H]F Junkie
Joined
Mar 22, 2008
Messages
8,490
Yeah, I respect Spectre's opinion, I just disagree on the main point of this thread. Being that a PSU should never "explode" regardless of any specification.
 

Nafensoriel

Limp Gawd
Joined
Nov 23, 2015
Messages
382
Zero explosions.
Bad metric.
It assumes you can prevent any and all manufacturing faults. This is beyond impossible. You can, and should, minimize them. You cannot change the "weird shit" that just happens with anything manufactured though.

The better metric is how a company handles failures. EVGAs handling of the 3090s? IIRC was absolutely top-notch. They replaced them immediately, they didn't even try to prevaricate and cover their arses, and in the end, for the absolute cherry on top, they told people what the failure was when they had absolutely no financial reason to do so.
 

DukenukemX

Supreme [H]ardness
Joined
Jan 30, 2005
Messages
5,875
Bad metric.
It assumes you can prevent any and all manufacturing faults. This is beyond impossible. You can, and should, minimize them. You cannot change the "weird shit" that just happens with anything manufactured though.
This is why you don't allow a load to reach what is considered unsafe for the PSU. Lets say the components on the PSU are rated for about 750W but you don't want to let users run near that limit. Instead the PSU components should be rated for maybe 850W or even 900W and then limit the usage to 750. This is why OPP for a lot of power supplies can often go beyond the rated load on the unit but not blow up, but it usually can't do that sustained due to manufacturing defects that might occur. You see this sort of thing done with some LED bulbs as they triple the amount of LEDS but run them as half or lower load to extend their life while also consuming less power. Tesla and other battery manufacturers are also doing this by not 100% charging the batteries to prevent early failure. This is how a Tesla that's nearly 10 years old will still work just fine while your 3 year old iPhone needs a new battery.

What Gigabyte probably did was source components from lesser quality manufacturers, probably due to COVID shortage, while also maintaining their insane 140% OPP which would have been fine with higher quality components. This is probably why not all Gigabyte PSU's of the same model haven't failed. This also means that Gigabyte is aware that their consumers might put a higher load than the PSU is rated for but has no problems allowing the load to go beyond the rated spec of the PSU. You killing your PSU is just another reason to go buy a PSU. It's not like a high OPP was a mistake. The only difference here is that going beyond the rated spec of the PSU would take years to kill it, not a few days. This is just planned obsolescence gone too far.
 

cybereality

[H]F Junkie
Joined
Mar 22, 2008
Messages
8,490
Not so sure about that. If my PSU blew up, taking my expensive rig with it, and possibly endangering my life, you best believe I'm not buying that brand again.
 
  • Like
Reactions: noko
like this

Schro

Supreme [H]ardness
Joined
Mar 15, 2002
Messages
7,803
The only difference here is that going beyond the rated spec of the PSU would take years to kill it, not a few days. This is just planned obsolescence gone too far.

I had an interesting conversation about this the other day. One could argue that running a power supply over its rated load is akin to overclocking your power supply much like any other component within your system. When you're overclocking, the user is using the component in a unsupported manner that also void the warranty for that item. Overclocked items can fail prematurely, spectacularly and take out other components in your rig. Are you saying that it's the manufacturer's responsibility to prevent their products from being overclocked so the users don't use them as they shouldn't?
 

Meeho

Supreme [H]ardness
Joined
Aug 16, 2010
Messages
5,344
I don't have too. The feature described is not covered by a specification. So, there is no way to properly test it to a standard. You can not even test it to the manufacturers/OEMs design without them giving your the design paramteres which they did not.


No, at [H] we never tested anything with power supplies that did not have an industry defined specification. You need to go back and actually look at what we did.


No, clickbait is used because someone is complaining about something that has no specification or standard. By definition, that is click-bait.


No. I still test power supplies professionally.

Yes, I do know more about testing them than someone who makes clickbait journalism videos. They were not even aware that there was no specification for what they complained about. That pretty much seals it.

I know how they tested OPP and it was not to any standard that is accepted. That is not a valid test.
What kind of nonsense is this?! Features and perfomance without specified standards get tested all the time, in various fields for all kinds of products. What matters is that one conducts such tests equally/consistently/on the same playing field...for comparison and evaluation purposes.

BTW, [H]'s PSU tests, while quality for what they did test, were lacking enough that one had to rely on other sources more than for other sections.
 

Meeho

Supreme [H]ardness
Joined
Aug 16, 2010
Messages
5,344
I had an interesting conversation about this the other day. One could argue that running a power supply over its rated load is akin to overclocking your power supply much like any other component within your system. When you're overclocking, the user is using the component in a unsupported manner that also void the warranty for that item. Overclocked items can fail prematurely, spectacularly and take out other components in your rig. Are you saying that it's the manufacturer's responsibility to prevent their products from being overclocked so the users don't use them as they shouldn't?
That's a bad analogy. One has to purposefully choose to and force a CPU/GPU...to overclock, while a PSU could go OP during normal PC use without intention. Just like one doesn't short circuit something on purpose.

It would be more like asking should manufacturers prevent fans to be able to fail in order not to trigger their temperature protections. It's a silly question.
 

Schro

Supreme [H]ardness
Joined
Mar 15, 2002
Messages
7,803
That's a bad analogy. One has to purposefully choose to and force a CPU/GPU...to overclock, while a PSU could go OP during normal PC use without intention. Just like one doesn't short circuit something on purpose.

It would be more like asking should manufacturers prevent fans to be able to fail in order not to trigger their temperature protections. It's a silly question.
Users can read the label on the box and get information about the power draw of each component that they plan to use from manufacturers and other sources. They can select a set of components with a known draw that is greater than what the PSU is rated to supply.

How is that different than a stated clock speed and operational voltage that a user can choose to set above the spec?

This analogy is more aimed at the whole OPP thing, not necessarily relevant to short circuit protections.
 

Ebernanut

[H]ard|Gawd
Joined
Dec 15, 2010
Messages
1,557
Users can read the label on the box and get information about the power draw of each component that they plan to use from manufacturers and other sources. They can select a set of components with a known draw that is greater than what the PSU is rated to supply.

How is that different than a stated clock speed and operational voltage that a user can choose to set above the spec?

This analogy is more aimed at the whole OPP thing, not necessarily relevant to short circuit protections.
Because you're not setting anything, you're estimating how much wattage you need and trying to match it with a little headroom(but not too much). That also assumes that components are 100% predictable in their power consumption which they're not, some graphics cards are notorious for consuming way more power than they should under certain circumstances and I've seen motherboards that consume much more power for the cpu than they should due to a BIOS bug or simply by having an aggressive load line calibration setting enabled by default at stock for some stupid reason. There can also be variation between individual CPUs and GPUs where one boosts more or less than others or is more/less efficient.

My point being that even if you know how much a system should be using it can sometimes exceed that through normal use which is one of the main reasons we have OPP.

It's also worth noting that graphics cards usually don't actually specify power draw and instead simply list required wattage for the PSU that doesn't take into account the rest of the system. The only good way to get an estimate on power consumption is to use a psu calculator that takes into account all hardware, any overclocks, and capacitor aging(depending on how long you plan to use it) but it's still just an estimate that will almost certainly be off by at least a little.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
6,644
There is an industrial standard so to speak, equipment/power supplies are operated within design specifications. A power supply underated for the system loads has no guarantees on any protective mechanisms. All protective mechanisms for Power Supplies are based upon single fault analyst. That does not mean it cannot be designed and tested with multi-faults but unless it is stated and what conditions, one cannot assume it will. Any multi fault analyst can run into thousands of tests while single fault is very clear. GN testing methodology to me is questionable at best with the narrative that follows. Testing 1 EVGA calling it good and running through 10+ Gigabytes calling it bad is not equal in testing methodology. While yes, GN was exploring the reasons for an abnormal reports about Gigabyte 750w and 850w models, they also changed the methodology between Gigabyte and EVGA and AD for their conclusions. The poof out of thin air specs they want to see like in ripple was laughable unless they can back it up with something meaningful.

As a customer, I would like to know if something in my system shorts out, my power supply will shutdown preventing possibly my computer/place/life from harm due to fire. This to me is a easy test to formulate, just something reasonable, startup short as in new build and operating short at 60% power.

As a customer OPP/OCP for a worsening component, slow condition getting worse as time goes on or a more rapid over power scenario. From my experience in electronics the slow and steady increase is rare while the more rapid failure is more common, as a component fails it also sees a rapid heat increase of it which happens quick causing the component to fail quickly. Anyways a normal in spec power level to OPP setpoint is more reasonable and would be more a single point failure vice an extended out of spec condition followed by a fault. Yes I would like to know offhand if the PSU did not blow up, shutdown and better if it can restart normally and protect my system from whatever component that failed. In other words OPP testing by GN, even currently, has information value for me while I may not fully 100% or 70% or 40% like how they do it.

Overtemperature makes a lot of sense - stop the fan, plug it up and see if the PSU shutdowns or frys itself.

As for Brown outs and over voltage, I see no easy way or consistent way to do those.
 

cybereality

[H]F Junkie
Joined
Mar 22, 2008
Messages
8,490
Well, yeah, to be fair, you'll probably never go to 140% power under normal usage. Even slight spikes (when starting a game, etc.) should not be that high assuming you left some room in your PSU calculation.

But it is still interesting to see in a video because, as I've said, we should have no explosions. Even under the artificial test, the PSU should shut down or safely die, fine, but not explode.

Also, my spare machine has an RTX 3060 and a 500W PSU. Nvidia claims you need 650W, but I have a Killawatt and I seem to be under 400W so I think I'm safe. But yeah, kind of riding the edge there.
 
  • Like
Reactions: noko
like this
Top