Read more.Hardware maker to adjust the OPP values after many reports of failed or DOA PSUs.
Read more.Hardware maker to adjust the OPP values after many reports of failed or DOA PSUs.
Presumably, these are the same PSUs that Gamer's Nexus released something of an expose on late last week, where they had actually footage of a number of them 'exploding', after testing quite a few over several months. That description might be a tad sensationalist, but certainly, "bursting isnt flame" isn't.
This is not a good look for Gigabyte and together with lots of reports of user problems with some of their Aorus mobo's (it starts to look like either you get a good one or you don't) is enough to convince me to just avoid them at the moment, despite solid experience in past years. It's just convincing enough evidence of a chance I don't feel like taking.
A lesson learned from PeterB about dignity in adversity, so Peter, In Memorium, "Onwards and Upwards".
I've avoided Giagbyte since several Mono issues both hardware (DOA and short term use failures) and software since at least 2008.
These are the same ones that Steve from GN made a serious point about. That and the fact they're being bundled with GPUs that will certainly push them way over their bad engineering design/QA testing is also dire. Linus and Luke discuss this and from Linus' perspective they're trying to shift dead/bad stock while knowing they're bad hardware which is either gross incompetence or malice.
I'm also concerned the replaced units aren't actually "fixed", it technically means they'll have lower capabilities than before without generally fixing the issues.
Weird question, why does it say " Tom's Hardware reports that Gamers Nexus.." and not "Gamers Nexus reports" ? Is it something to do with print vs video ?
Yeah, I first saw it on the WAN show, which is why I then went to GN.
I guess whether it's "fixed" or not depeds on the fault??
I mean, if they cut the OPP from 150% to 120% and it still delivers the rated power in a correct manner, then it could be argued that the OPP was set too high, in erromr, and it's been corrected.
If I remember the results correctly, the GPUs they're bundled with shouldn't put them over. But what did, pretty regularly, was the maximum loading by power testing gear. Surely, if they still deliver rated rated, but the OPP is scaled back so that it delivers up to that point, they're still providing rated power, plus some, and the OPP stops them delivering so much that they cook themselves.
But either way, it's still not good for Gigabyte. It shouldn't have been able to happen in the first place.
A lesson learned from PeterB about dignity in adversity, so Peter, In Memorium, "Onwards and Upwards".
The issue is with how it is failing. An overloaded PSU shouldn't start fire lol. It should trip or some other mechanic should kick in, it flaming out or things catastrophically failing points to more than a simple bad setting. Hopefully this makes the product dead in the water, as normally with this sort of catastrophic failure, it becomes so widely known that people will no longer touch the product and the manufacturer bails on it and just quietly removes it. But eh I've also seen manufacturers double down, and considering the offered "fix" here, I fear Gigabyte is more that type of company.
Never been a fan of em, however I never really had a reason, it's weird. I just..didn't like em. It's nice for my irrational hatred to have some reasoning!
I'm pretty sure it was stated by GN they also failed or had issues even in normal use or at one occasion, load testing at 60%.
As Tunnah has quite rightly pointed out, any load testing should not cause a catastrophic failure regardless of artificial or actual intended use. Additionally, Gigabyte and their testers would have used the same or similar load testing equipment on their PSUs to validate them as operational so why wasn't this picked up then?
The issue with tweaking the OPP is meaning that the topmost threshold for dealing with a short spike in power requirements is lower which is more likely to outright trip and shut down a system. And IIRC they were being bundled with 3070s which weren't as bad as the 3080/3090 for transient spikes but still limits the maximum threshold of dealing with those.
As it looks like the FETs seem to be a primary issue, I wonder if they either got a bad batch with a poor-er tolerance than the actual specced ones they would have used or were they forsed to purchase inferior ones due to component shortages...?
I dont know if this is relevant but i purchased a Gigabyte CV27Q gaming monitor late last year has it had good reviews. A few months later it just died on me for no reason. Sent it back and bought an Asus instead. Due to things like this i wont be buying any Gigabyte hardware from now on. I am currently looking for a new 3080 / 3080 ti GPU, Gigabyte is now off my list!
That was exactly my point.
And presumably, it wouldn't, if the OPP point was set appropriately. Again, my point exactly.
Steve himself, in his 'review' concluded that at least one protection circuit wasn't operating correctly, though they had insufficient samples to be certain which one(s) but that OPP cutting in at 140% was, in their opinion, simply too high.
And that was my point. Of course it's not acceptable for power supplies to fail catastrophically and pretty consistently, and I say that as someone that lost a hardware RAID board, four (4) hard drives and a £1200 graphics card when it happened to me, about 20 years ago. My office stank of it for weeks. I still don't know what caused mine, but I have suspicions about a major mains spike, not any inherent PSU fault. A one-off like that? Bad luck. A series of failures of the scale Steve had, and was seeing in user reviews, not luck. Faulty design, faulty manufacturing, faulty components? But faulty, whichever.
However, as again pointed out by Steve, no end user should be pushing a PSU to the 130% or 140% point where, mostly, the damage occurred. The point of the protection circuits is to shut the PSU down before it gets to the physical damage stage, and certainly before it gets to the bursting into flames stage. But again, if Steve is right and the OPP point was too high, and especially if higher than other competing brands would be, and should have been maybe 120%, then with an adjusted OPP, the damage may well not have occurred to whichever component failed .... and it wasn't always the same component.
If, for argumet's sake, the components can cope with over-power to 130% without damage, and OPP was set at 120% then even that design, with those components, wouldn't keep failing. But, push the over-power to above 130%, like in the 140% test, and .... the damage is done. It may not fail at that point, but as he found, did subsequently.
As for 60%, it was indeed under load testing but after the first 140% test. And yes, a consumer could well hit 60% and more during normal use, but (according to Steve) not the 140% of the load test, and indeed then the 133% load test, that preceded that fail at 60%.
That is to say, it wasn't a PSU that just failed at 60%. It was one that, :-
- first, cut out at 140%.
- wait 5 minutes to cool down, then
- second run, they set to 130% then edged up to find the exact cut out point, which was 133%.
- then, third, turn on again, when it cut failed at 60% load.
By his statements, normal usage by end-users should not be able to get it to the 133% or 140% point in the first place, so the precursor stresses, not once but twice, to the 60% load test would not have occurred, and could not, if the OPP kicked in that much earlier, as he said he and others believe it should.
So just to stress, I am not defending PSUs that blow like this. At the very least, they are faulty in that the OPP, and/or other protections, are not preventing the circumstances resulting in the stresses that lead to catastrophic failure.
The root cause may also be much deeper than protection circuits. As I said, could be faulty design, faulty manufacturing, faulty or inadequate component quality, and so on. But whatever those faults, the point of the protection circuits is to shut the damn thing down before it either bursts into flame, or causes sufficient component damage that it later bursts into flame. It might be that no such incidents would have occurred, or occur in modified units, once the OPP etc were lowered. The only way to establish that would be a much larger scale testing, after lowering those limits. At the very least, Gigabyte ought to be doing enough testing to determine whether that is the case, or whether the problem is more insidious.
That said, would I buy a 'modified' version of one of these? Hell, no. Do I trust Gigabyte enough to buy ANY Gigabyte PSU, after this? No. How do I know that either the boards from these, or other boards with the same inherent issues, if it is more than the OPP being too high, don't get repurposed into new models? I don't, which is why I won't touch a Gigabyte PSU in anything remotely resembling the near future. As it happens I wasn't going to anyway. I have my preferred brands for PSU, based on my own previous experience and see no reason to switch. But if I had been going to, I wouldn't be any longer.
A lesson learned from PeterB about dignity in adversity, so Peter, In Memorium, "Onwards and Upwards".
Iota (18-08-2021)
Didn't Steve also mention it may not be just the OPP being set too high towards the end, I watched it last week out of curiosity. Either way it's simply bad design that should have been picked up with basic batch testing prior to shipping. As an end user I'd have to question Gigabyte's QA processes, because it sounds like they either have none, or what little they do have is inadequate and not fit for purpose.
I wouldn't be accepting a "fix" like this, I'd be asking for a full refund.
Steve did a new video in response, its clear gigabyte are lying and doubling down by trying to make GN look like its intentionally being misleading.
https://www.youtube.com/watch?v=Xts3pvbcFos
Its a good watch (although long as all GN videos are, do something at the same time!)
I never bought gigabyte power supplies but I definitely wont be buying them in the future due to how they are handling this, using shoddy made components and then not making a proper recall (rather than a small press release). This is clear its not the overcurrent protection thats the problem as GN say they had units which failed simply at 60% load, no attempt to run higher but it blew up!
I went back and listened very closely to that bit.
As I understand it, they did run that much higher before it failed.
Their process was to load test at several fixed points, like 60, 80%, 100%, 120% and 140%. I don't remember the exact bands (I think there was a 130% in there, too). It passed all these util 140% at which point, OPP cut it. That told them that OPP didn't cut in at previous point, but did at 140%.
So they rested it and let it cool down.
Then they went back to previous point it passed at (130%, IIRC) and then nudge up a bit at a time to find the exact point OPP kicked in at. That meant something like 0.5% steps, and that particular point OPP'd at 133%.
So they rested it again.
And then, having established the OPP at 133%, it was running at about 60% (i.e. normal load) when it catastrophically blew up. It was under that 60% when it blew, but after having previously OPP'd twice, onceat 140%, then at 133%, but were not doing so when it blew. Others blew at other points. Listen to the vid at about 20:30 to 22:00 ish.
A lesson learned from PeterB about dignity in adversity, so Peter, In Memorium, "Onwards and Upwards".
You seem to be referring to the original GN video or am I mistaken?
Listen here: https://youtu.be/Xts3pvbcFos?t=748
12:30 onwards, they specifically note having a test rig which NEVER tripped OPP and ran at 60% load and blew up.
They are just flat out lying by saying its not a problem yet user reviews say other wise and then they say third parties are testing wrong (which is incorrect) and then eventually add in ok we swap it for you.
How can you say everyone is wrong and then offer a replacement? Its obviously built incorrectly while trying to blame everyone else its pathetic from a corporation haha.
I listened to both, but yes, I was referring to the unit that blew after the OPP sequence in the first.
In the second "60%" test, the one yu referenced which we might call the "endurance" test, it wasn't the PSU that blew. It was the GPU. Steve did make the point that :-
1) They only did one endurance test, intended to be 30 days.
2) After 72 hours, the GPU blew, they found it on narriving on day 3.
3) They replaced the GPU with, IIRC, a 6800XT and it then completed and survived the endurnce test.
Steve pointed out they don't know what blew the GPU and, so far at least, don't have concrete evidence that it was, or wasn't, the PSU. We might infer, given the extent of PSU problems that that is a distinct suspect, but we don't know. It then did apparently drive the 6800XT for the duration of the endurance test. It could not have done that if the PSU blew, 72 hours in.
He also pointed out, whether the GPU was blown by a PSU that blew that board but survived itself, then didn't blow the 6800XT, it's bad for Gigabyte either way but he can'tbe sure the culprit was the PSU. Tis is all covered at about 24 mins into the original video.
Which brings me back, in the absence of a detailed list from Steve, to the conclusion that :-
- the PSU that did blow ay 60% was after the OPP trip 1 at 140% and trip 2 at 133%.
- the inference, perhaps, was running those loads may have caused the damage that resulted in the 60% failure.
- it was emphatically not, as Gigabyte asserted, "extended testing at load". He made clear that the original load testing was at several levels, briefly, looking for an OPP test.A few minutes, at most.
It is that last point he seems to be stressing in the time point you mentioned, that ONLY during the endurance test was there an "extended" (per Gigabyte's claims) period of load, but that that load was 60% which,first, should not cause any PSU ti even breathe hard, never mind die. And second, wasn't under artificial load testing (i.e. actual PC build, not PSU test equipment), as implied by them. And third, was an utterly normal real-world load any high end card is capable of sticking a PSU under.
What would be nice is a documented list from Steve of :-
- units tested
- test regime
- pass, or fail and if fail, what, why, where, when and how it failed.
However .... the situation between GN and Gigabyte appears to be not yet calmed down. Gigabyte have doubled down once already. It's not beyond possibility that Gigabyte double down again and refer this to their lawyers.
If I were Steve, I would not be willing to show all my cards yet. Indeed, I have been in that exact position, with a company threatening legal action over a review they didn't like. The power roles were rather reversed. though. It was a small startup company, threatening legal action against the extremely large publishing empire I was freelancing for. I had a chat with the editor and a lawyer about the extent and nature of my testing, results and record-keeping, and whether should it come to it I was prepared to back my conclusions in court. I was. The publishing company told the supplier to sue if they felt the need, see you in court. Sensibly, the company thought twice about that.
If I were Steve, I'd be saying enough to stand up my methodology and conclusions, which he is, but certainly not handing Gigabyte his whole hand of cards. He may yet need them .... though my personal view based on publicly known data and my experience if likely press reaction if Gigabyte were daft enough to do that is that they'd better batten down the hatches and prepare for rough seas in the review arena .... if they can get anyone to review their stuff after they did go to lawyers.
On your last points, about what Gigabyte have said, how they've handled it, that reeks to me of someone in a press office gettling rattled and trying to CYA. Their response seems to me to utterly stink, and it's why I just pivoted from buying anything Gigabyte in my current build, to almost anything except Gigabyte. It has become a trust issue, and I don't have any.
A lesson learned from PeterB about dignity in adversity, so Peter, In Memorium, "Onwards and Upwards".
There are currently 1 users browsing this thread. (0 members and 1 guests)