Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
R9 480 let's the smoke out of motherboards
#1
So, this is a thread intended to track motherboards and systems which blew due to the R9 480. It's only been a few days but we already have several accounts. It's pretty simple, if you see or read about someone who had their system damaged due to running a 480, post it here.
Reply
#2
There are several here: https://community.amd.com/thread/202410

If you read how it plays out, it seems unlikely to be fake, the op had no clue at first and still thought that he had some other issue. The discovery of damage plays out thru the post ae a result of other members helping . Seems genuine enough.

More interesting, there is a reddit member claiming to be banned for exposing the issue of burning up boards, his thread gets pushed down while a bogus fake thread with totally bogus claims went straight to the top. This thread claims that motherboard pcie slots cab handle 300 watts with no problem what so ever....yeah, I am serious. I guess we all have been fools and there has never been a need for extra psu plugs...the slot has been capable of 300 watts all this time. WoW
Reply
#3


Virtual reality and festive neighborhood bonfires for the masses.
Reply
#4
A stick of apoppin's DDR4 RAM died after he tested the RX480. It could be just a coincidence but I have a feeling that the RX480 is to blame. He had a whale of a time reviving his motherboard afterward.
Reply
#5
https://www.techpowerup.com/223893/amd-4...d-into-8gb
Who wants to bet that AMD will get a free pass from the tech world when they patch this out?
Reply
#6
Tom's has an update: http://www.tomshardware.com/reviews/amd-...,4622.html
Reply
#7
RX 480 8 GB BIOS flashing works with retail RX 480s, AMD caught lying about this flashing: https://www.techpowerup.com/223913/amd-r...enchmarked
Reply
#8
(07-05-2016, 09:50 PM)SteelCrysis Wrote: https://www.techpowerup.com/223893/amd-4...d-into-8gb
Who wants to bet that AMD will get a free pass from the tech world when they patch this out?

Was that you intended link? 4gb unlocks to 8gb with bios flash? Amd does this cap on purpose. In the past they have released a batch of chips that unlock cores or shaders and then they have websites and forum posters advertising this to try to get sales. Now, I guess it's memory. Pretty lame because this info takes time to spread and by then, the unlock chips are gone out of the channels
Reply
#9
(07-06-2016, 01:33 AM)SteelCrysis Wrote: Tom's has an update: http://www.tomshardware.com/reviews/amd-...,4622.html

Wow. I don't understand the conclusion, it's like they went out of there way to say that this isn't an issue, that running over the MAX in the spec by 25% is okay? TOMs is saying they decided to do this follow up all on their own....

I also, interestingly enough, saw TOMs going on and on about how AMD hasn't contacted them at all. Not anyone at TOMs since they released their article, they were trying to use this as some proof that AMD wasn't behind the follow up article....but...I disagree. I believe AMD was 1000% the reason, not that they paid them out contacted them but it's the fact that they haven't, that they were afraid of being blacklisted. In all actuality, they have to be carefully in their wording so it's not surprising they would treat statements like the potential of killing motherboards with care, they can't say this outright....but...this second article went well and beyond with claims that their is no risk and it's safe....even shockingly offering a new scapegoat which I expect AMD troopers to pick up on, "as long as your slot is cleaned and card plugged in firm correctly, there is NO RISK"....

Wow...
Their have been reports popping up already, damaged boards and failure related to this card, everyone of those claims are met with AMD troopers accusing them is overclocking, now they can have an even better cop out,"did you even clean out the slot, that's common sense". Now they have the backed up fuel to question the victim and put it off on them...they do that already but you can bet this article will be linked to with the claim that it just has to be the victims fault cause the TOMs proved it, any failure is operator error. TOMs says it's fine, it's okay???

They say standards exist for a reason, even in the follow up. Yet then turn around and say that drawing, consistently and on average, 25% over the MAX allowed per the ATX slot specification is perfectly okay? That it won't hurt boards? How can they say this? That it will not burn up your board. They are actually crossing the reverse line in legal liability and doing a huge disservice to the entire community, the real people, the people likely to buy cards, those that read their site and make purchasing decisions. This is why they have the added disclaimer, as long as the slot was clean and card seated properly.

The simple truth is that there is no way whatsoever they can say that this is not a problem and it will not damaged boards. What the hell is wrong with them? They cannot give any assurance, they can't make such a call...it's preposterous. Completely shocking to see the emphasis that this will not burn up boards. There is absolutely no way they could have any idea, they're are far too many boards out there and every one of them have their own spec, tolerances and guidelines on top of the components which have their own tolerances. None of them are designing out testing designs that are documented safe and certified for up to 7amp drawn thru the PCIe slot. No one has such a target and you cannot say it is safe or okay.

Another thing, TOMs is using terminology like standards, "standards exist for a reason", but this is an electrical specification. It's not just the card that is running out of the pcie standard. It's the card drawing more current than the max set for an electrical circuit. If you look at any PSU, their are ratings for each rail. Considering that you can usually run over the PCIe standard for a single given plug. most PSUs can handle more current thru a single PCIe 6pin than the standard sets. But their are specifications from the manufacture for each rail. Their are specifications for wires, pins, plugs...every single electrical component has a MAX specification. This is independent and has nothing to do with the PCIe standards set. You should not under any circumstances run over the MAX specification on an electrically rated component. In all truth, you never should run at the max ratings even.
Motherboards are designed specifically and as far as I can tell, there is no expectation that there would ever be a larger than the max load for PCIe. This is not an expected condition and not an area that extra money would be spent to beef up. Most cards stay well well below the MAX, this is not an known for pushing the limits on. No one designed boards for this kind of load. How TOMs can assure anyone this is okay, it blows my mind.

See their are pins on the slot,traces, all sorts of things to consider that all have their own ratings. They adhere to the system they are designed for. The expectation is no card should be drawing more than the MAX rated for the slot. They all have their own max,

I seriously question anyone who suggest that it is okay, running 25% on average over the ATX spec with spikes orders of magnitude higher. This has already damaged systems right off the bat, the real scare is that prolonged over draw in current will most likely damage and could corrupt components over time, it may not burn up immediately or show physical signs but a motherboards entire 12v power circuit is also at risk due to the strain. We already see data showing huge vdroop as the 480 load pulls down with its massive current peaks and over the limit averages. So, there are potential issues besides the more obvious killing your board, melted plugs, burnt pins, fire and smoke. Instabilities and component failures which may cause tremendous headaches which users never connect back to the strain from the gpus excessive load
Reply
#10
Reply
#11
Here is another Easter Egg. Anyone who thinks AMD can fix this power distribution issue with a bios flash or driver is out of their minds. This is a hardware issue and there is no way I can see this happening. I say this without seeing or touching their card but I feel quiet sure that these suggestions are complete nonsense. Sure, I guess there could be some special technology that they invented especially for this card, but I seriously doubt this. Sounds way crazy to me.

The only fix I can imagine is to choke the card down, limiting performance, clocks, and the overall power target. To fix this issue on the cards out now, it's gonna hinder performance. The card will have to be slowed down. This is the only way I see it.

This is a hardware issue, a design issue. Their are people saying that it is somehow their marketing screw up, it's absurd....insane what these people come up with. This is an engineering issue and their is no two ways about it. I don't know if it was poor oversight, incompetence, or the lack of funds but they all go hand in hand. Management could have rushed rushed but the engineering team fucked this on up big time. Their is simple no way in the world that anyone could believe AMD engineers didn't know he card was drawing more than 150 watts. It's quite clear the they knew. These cards are set to 165 watt limits, they all seem to run up and exactly to 165 watts. This is a set limit by their engineers. It's also not anything to do with the actual 6 pin used on the card. It's not the 6 pin which is causing this load distribution, that is stupid. It's the card design. It's hard wired and will draw the same amount from the board no matter if they had stuck an 8 pin there or not. It's the board layout. The phases are routed and this is his the load will always be distributed..by design.

So how could this be? Was it a monumental Fuck up or seriously questionable engineering? It really seems that there was some internal confusion, that there were people in and who thought the chip power draw of 110 watts was all there is to it. I mean, I never heard anyone quote or care about the draw of the chip all by itself. We buy GPUs, the entire gpu. You cannot use a chip without it's memory....without the board and it's very important components. Did some one seriously design the board with thinking that it was only needing 110 watts? Is this the fuck up? Cause looking at the figures, most board designs are aimed at significantly lower power draw from the motherboards. Like, 40-45 watt or so is super common. And doing the math, 75watt from the 6pin plus 40 watts and you end up with a very normal looking scenario. Typical scenario.

Also, it's even more interesting when you look at what is being powered by the phases that are from the board slot...It is really bizarre. Looks like a big Fuck up to me. It wouldnt have cost anything to have the load distribution set to draw The remainder from the six pin. At a split of 60 watts to the board and 85 from the pin. And this would have been much safer.

It's like someone really thought they were designing for a 110 watt card. It's an engineering Fuck up all the way. So what it this was the goal, 800mhz cards that only used 110watts total then the top execs said, oh shit...The pascal is blistering fast, make the 480 as fast as possibly possible...It is still engineering who set the. 165watt cap with a load distribution that ends up risking motherboards.

I doubt that so much though. An 800mhz Polaris wouldn't need 8ghz ram. The time that the 1070 launched, it's not enough time to pull such a change. It's not enough time to have cards with slower ram then turn around to 8ghz and have the retails full of stock for their launch. I think the 8ghz goes back awhile, months prior to release. So far back that the boards could have adjusted to distribute the load split.

It's really hard to imagine any thing but a huge blunder
Reply
#12
http://www.tomshardware.com/reviews/amd-...,4668.html
Tom's seems to think the fix was a success.
Reply
#13
(07-08-2016, 03:55 AM)SteelCrysis Wrote: http://www.tomshardware.com/reviews/amd-...,4668.html
Tom's seems to think the fix was a success.

Gosh, let the domination of NVIDIA begin.

Even fixed the card is what I thought it would be- the Radeon 290 v.3.

Same stuff year after year from AMD, gets hard to care much.
Reply
#14
Guys I'm buying a GTX 1060 at launch.
Reply
#15
Ocre, you are awesome, just awesome.  For Poppin to not want you to join BTR at first..  and then recently experience some kind of pita failure with his 480..  well he did ibvite you later on anyway.  

I love your illumination to scenarios like this!
Reply
#16
Had no idea AMD released the driver yet. I do find it very interesting. Of course, as I predicted, the only way they got under the ATX specifications was to reduce the performance.

TOMs is being so out of the way nice on this. So very strange...
TOMs is claiming amd recalibrated the PWM load with a driver? Does this driver flash the bios? Also, there are a lot of people out there that buy cards and put the disc in, install the driver and never ever update.

But nonetheless I find the data interesting. TOMs says something important near the end, "AMD really delivered on a driver update that overrides the BIOS defaults to relieve the motherboard slot without a performance loss, we have to wonder why the Radeon RX 480 wasn’t configured this way originally.".

Sounds great...But there is something funny. The configuration should had already been set, there is only so much you can do, I really don't get why TOMs is so vague on the data now. I just don't think it's possible to dramatically rebalance the load all by itself....those circuits are attached, for the most part...There should be only very little tweaking. But think about it, if you reduced the speed or function, the load on the circuits that are mainly drawing from the motherboard....
in the numbers.
Take a look at the power consumption, the new driver not in compatibility mode pushes the card up to right at 170watts. But only that, it is also down to have the card running at higher clocks. This certainly would account for the increase in power from the 6 pin, we see an increase there. If you look at TOMs earlier article, you will see they have a good idea what is drawing from the six pin and what is drawing from the board. The higher clocks would draw more power from the 6pin. Take note

AMD increased these clocks to offset whatever they did to lower the train from the motherboards slott. In my estimation, there would be little redistribution possible so what did amd do? Slow down the parts that draw from the slot. I would like to see so test ran, because the increase in clock speed could offset a reduction in memory/bandwidth speed.

I wonder if this would show in the GPUz? Or, could AMD hide it and the software not report a difference? But yet, in this case the performance could tell the story, run software and check every memory system, compare the old drive and new on every level. There is something done to which the increase in clocks is gonna offset.

Here is another worry of mine. Is this software a universal adjustment or is it just specifically doing something in the apps and cases that reviewers have found and reported on. See, even the supposedly fixed driver, the power draw is still going over on the amps. It's also running up to the limit, right on the limit. It's close, really close.

If any one finds some more test
On specific differences between the old and new driver, please post them. I am really busy this week but I have a good idea that they had choked down one side and ramped up another to offset the performance
Reply
#17
TPU is happy as well: https://www.techpowerup.com/223981/amd-r...rmed-works
Reply
#18
The load distribution is still an issue with the fixed layout. This hack AMD it's doing is interesting, but look at the results. Going from 0 to 13 has barely dropped the motherboards draw down, we are talking 5 to 6 watts. This is because there manipulating the PWM to try to compensate but te board is still the same. The design is the issue and it it's hard wired.

The interesting thing to note, this is still super high current to draw from the slot, most cards don't draw anywhere near 70watts sustained. Even the 290 x2 only draws 30 something watts.

I still would love to see a detailed look into each phase, like w had before the fix. Also, I would love to see the cards results on a full st of synthetics, see if they did any gimping on the memory side. This is a work around, not a fix.
I think there could be more to it, some trade offs ..

Right now the media is being really really over nice to AMD. This is them cutting amd a lot of slack. Very strange.

I believe there is more to this work around, keep an eye out. I also think the default mode is still to high, the amps are what matter...Not the watts. As far as amps goes, this new driver still has too much current. The amps are too high
Reply
#19
https://www.techpowerup.com/225081/high-...or-listing
It's of doubtful practical result, but it's still telling in its own way.
Reply
#20
(08-18-2016, 02:02 AM)SteelCrysis Wrote: https://www.techpowerup.com/225081/high-...or-listing
It's of doubtful practical result, but it's still telling in its own way.
Read some of the comments...wow
Those AMD troopers are messed up in the head.
They say, "its not out of spec, the new driver fixed it."
The sad part is, they can link to articles that say the power issue is fixed because those sites cowardly backed down for some reason. AMDs driver did not bring the 480 to PCIe spec compliance. It merely shifted the load somewhat. They brought down the motherboard slot power draw by sucking even more power from the already over the limit 6 pin connector. So, the 6 pin went from being over the limit to being way over the limit.
THe motherboard slot still draws over the PCIe specification with the new driver. But somehow these tech websites considered all was good...You know cause it went from not being compliant to not being compliant. I guess there is a reduced risked to motherboards..so completely resolved, LoL.

Anyway, those sites went super soft real fast. I didn't understand that. I almost wonder if they were influenced, such as with cash or by threats.
See, one could make the case that those articles could have cost and hurt AMD. Legally, a case could be made and TOMs would have to prove these gpus burn up motherboards...and that is where things could get slippery. See, the cards wouldnt burn up every board and even if they could produce one that was damaged, it wasnt the card but the electricity that damaged it. Unless there was real time evidence that the wall socket and psu were normal at the time of the incident, evidence that can hold up in court.
And there is the catch 22. The psu will surge the instant there is an electrical fault. Which then one can question all of it, was this not the psu that caused the damage? Without the psu, there would be no damage.

Anyhow, the sites backing down is one thing but they actually went further and were more than assuring that all was fine after the new driver. But still, both the motherboard slot and the 6 pin both draw over the PCIe specification.
The obscured "compatibility mode" that can be enabled by opening up the driver software. Its not a setting that everyday users would find for starters, but if they did they surely wouldn't feel like this is something they need to enable. There are a few "compatibility mode" boxes throughout windows and they are almost always left alone.

But even in the case, some random users just finds it and enables it. The 480 still is not PCIe compliant. See, it does finally bring the motherboard slot down to proper levels but does not for the 6pin. It would still go over
Reply
#21
[Image: 195wfi.jpg]via Imgflip Meme Generator
Reply
#22
Lol
Reply
#23
(08-20-2016, 07:17 AM)ocre Wrote: Lol

apoppin has let the smoke out of 4 Polaris cards now and none of them would overclock worth beans. He will be letting the smoke out of a 5th card within the coming days. No more damaged hardware, thankfully. I think it's pretty telling that AMD lost their PCI-E certification though. And I suspect that this issue is part of the reason why we have not seen the RX480 aftermarket cards yet. This is a major issue and I agree with ocre 100%, his comments are completely spot on.
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)