AMD confirms Power Issue for RX 480, blames memory bus, will patch

[thesmokingman]

Oh yeah they are hiding the fact they had a overly hot product and it missed its target by miles, just like Kyle stated.

Yeah look at the people that blasted him for saying that, and now everything is real.

Omg the deception? You fail at running your agenda to its conclusion. Come clean about what? What is to gain from hiding an "overly hot product?" How did they hide this? Does it actually RUN HOT?? Omg?

AMD doesn't need to that is true, but its better if they come clean with it.

 

[razor1]

Omg the deception? You fail at running your agenda to its conclusion. Come clean about what? What is to gain from hiding an "overly hot product?" How did they hide this? Does it actually RUN HOT?? Omg?

Come clean with it as fix it and tell us how they are fixing it.

Cause the problem is already obvious.

OMG, its a troll!

 

[thesmokingman]

Come clean with it as fix it and tell us how they are fixing it.

Cause the problem is already obvious.

Omfg, now I see who I am dealing with.

 

[razor1]

Omfg, now I see who I am dealing with.

I'm dealing with a troll that fine, I know too.

How long have you been here for and you don't know how motherboards distribute power over its components?

You don't know ohm's law and how it works?

Your posts in the past tell me you don't, I guess I know where to put you in, hmm that short yellow bus comes to mind lol.

 

[ChosenUno]

At the very least, this issue is going to cause a pretty big delay for AIB cards.

Usually AIBs will churn out custom boards that's very similar in design to the reference board for cheap in the first wave, but with better components. With this issue, though, they'll need to, potentially, scrap that and design from scratch. Might be a week or a few week's worth of delay.

 

[razor1]

At the very least, this issue is going to cause a pretty big delay for AIB cards.

Usually AIBs will churn out custom boards that's very similar in design to the reference board for cheap in the first wave, but with better components. With this issue, though, they'll need to, potentially, scrap that and design from scratch. Might be a week or a few week's worth of delay.

Yeah and that was what I was getting at, the first few AIB boards really have to be tested just like the ref board was.

 

[Wascrash]

We will know on Tuesday if they release a fix.

The problem with AMD right now they are screwed, literally, they took a product that could have made them money and now are looking at a potential loss!

This is what investors, stock market , the companies or people bank rolling AMD are seeing.

Do you think they feel comfortable with this right now, they will want answers![/QUOT

you are over thinking all this. Its really no big deal. Buy the card if you want. I have no reason to but I would and not worry about it if I did. This is not the first card to draw more power.

 

[razor1]

you are over thinking all this. Its really no big deal. Buy the card if you want. I have no reason to but I would and not worry about it if I did. This is not the first card to draw more power.

Gotta fix your quote

Well I don't think I'm over thinking it, AMD answers to who? Their board of directors right? Who do they answer to? stock holders and the people that they own money too, they have 2 years to start making money before they start paying back the primary on their bonds, right now they have just enough capital to stave off the first payment, while their working capital goes to nothing.

No bank, no independent investor will give them more money at this point. They are so leveraged, AMD as company is worth less than what their assets are truly worth.

This launch was supposed to be the turning point to get AMD back in the black for end of this year and beginning of next year, just before Zen comes out which was to really shift them to a proper functional company again.

 

[ChosenUno]

If anyone has a spare MB+CPU+RAM combo lying around and wants to test the claims of the twitch guy, it should be really easy. Just plug a RX 480 into the PCI-E slot, and don't plug in the 6 pin connector. Oh and please record the scene when you power on.

If he's correct, there's gonna be a lot of smoke and flames as the poor GPU tries to draw 160W from the PCI-E to power the Vcore.

FOR SCIENCE!

 

[NKD]

If this is indeed a flaw and can't be fixed with software or bios update Whoever was responsible for the PCB design should really be shown the door especially those who let is slip through Q&A. This is just unacceptable. May be Kaduri needs to do what he is good at go back to designing shit. May be if he is hard ass these guys are being pressured and that is usually when people cut corners.

 

[razor1]

They should be able to do it through bios or drivers, i can't see how they can do it while not touching their frequency though, but again, if they are able to do that Kudos to them, that would be an awesome fumble recovery.

I don't think they can blame anyone from engineering, they are told what to do once they get to the point of we need this card to perform this way.

And they will do what ever it takes to get to that point if its possible.

I don't know if you have worked in a large company before, but personal experience when its technology based and marketing has too much control, things get f''ed up, its just bad business. I would not be surprised if this is what happened, marketing stated somethings about this card to the public before anyone knew the actual outcome, and to reach that goal, the engineers did everything they could to get it there.

 

[Chimpee]

you are over thinking all this. Its really no big deal. Buy the card if you want. I have no reason to but I would and not worry about it if I did. This is not the first card to draw more power.

It is not really about whether you or myself will buy that card but how the market will perceive them from there on, it gives consumer reasons to not purchase the card since there is a flaw that needs to be fix or wait until 1060 comes out. Just reading some of the post here, while there are people who already bought the 480, there are other people who are gonna wait and see mode, this wait and see mode hurts AMD as it gives nVidia time to come out with 1060.

While there are cards that draw more power than it is spec due to spikes, AMD is the first to draw more power at a sustain level over the pcie slot specs when gaming at stock speed.

Edit. Fixing Quote

 

[razor1]

Yeah and Marketing 101 ya don't let a consumer think about things, if they think about things it rarely gives into your favor.

 

[Chimpee]

If this is indeed a flaw and can't be fixed with software or bios update Whoever was responsible for the PCB design should really be shown the door especially those who let is slip through Q&A. This is just unacceptable. May be Kaduri needs to do what he is good at go back to designing shit. May be if he is hard ass these guys are being pressured and that is usually when people cut corners.

I am kinda worry about the first couple of AIB's that will come out, while I know they will make changes from the reference board, do they have time to make tracer changes?

 

[Palladium@SG]

I am kinda worry about the first couple of AIB's that will come out, while I know they will make changes from the reference board, do they have time to make tracer changes?

I won't be surprised if early AIB boards are a copypasta of the reference design with same PCIE overloading problem despite a 8-pin present. They are probably just as put off guard as we did when this blew up this hard.

 

[primetime]

You know how some of these builds being put together these days use a riser cable (powered?) and then the gpu mounts in a specialized bracket, i guess so the card is able to be mounted in a more visual way. I guess the pins would still be overloaded but at least it would not be draining as much power from the motherboard?

 

[Chimpee]

You know how some of these builds being put together these days use a riser cable (powered?) and then the gpu mounts in a specialized bracket, i guess so the card is able to be mounted in a more visual way. I guess the pins would still be overloaded but at least it would not be draining as much power from the motherboard?

I am curious too, I do see some riser cables with molex connectors, I wonder if it is gonna draw some power thru the molex away from the motherboard. It will be an interesting solution though not really feasible.

 

[NKD]

I won't be surprised if early AIB boards are a copypasta of the reference design with same PCIE overloading problem despite a 8-pin present. They are probably just as put off guard as we did when this blew up this hard.

I doubt it. Because Sapphire is listing 175tdp on thier Nitro card. I highly highly doubt they didn't catch this. If they are listing the TDP at 175 I am pretty confident they know they won't be drawing a whole lot of it from PCI-e slot. Otherwise they are knowingly selling a flawed design and it will burn even more boards. Which I highly doubt.

 

[primetime]

I think the exact words were they are like prostitutes they will show you the good parts and not show you the diseased parts lol

Sounds like your speaking from experience tell us more lol

 

[Meeho]

They always do top down design easier to scale down then to scale up or reuse portions of design.

Any case ohms law

Keep ohm's law in your mind when you are talking about this, the card as whole is asking for certain amount of wattage but the card VRM are taking the memory and GPU current from two different sources, in this case the GPU is getting more of its current from the slot and because it needs certain amount of voltage to sustain a certain frequency based on wattage, they need a certain amount of current. So lets say they do drop 30 watts or so, by dropping voltage, you won't be dropping amps necessarily, because voltage is less, you might still need to up your amps.

You still keeping repeating this false information to people. The board's input volts are constant (12 V and 3.3 V), so if it draws less watts it draws less amps. If they downvolt the GPU or the memory they are not changing the board's voltage at the PCIe slot or 6pin.

 

[lolfail9001]

You still keeping repeating this false information to people. The board's input volts are constant (12 V and 3.3 V), so if it draws less watts it draws less amps. If they downvolt the GPU or the memory they are not changing the board's voltage at the PCIe slot or 6pin.

FYI, Vdroop is a thing. Also, undervolting is not a solution since it's silicon lottery based and as such AMD even runs a risk of freaking having unstable cards on stock settings. The only viable solution would really be dropping power limit on stock.

 

[NKD]

FYI, Vdroop is a thing. Also, undervolting is not a solution since it's silicon lottery based and as such AMD even runs a risk of freaking having unstable cards on stock settings. The only viable solution would really be dropping power limit on stock.

Best solution is to cap the power limit at 150 no matter what. As soon as it hits 150w it should throttle. I bet it won't make much difference in most games. Because most at stock I saw was 163-165 watt. I am sure they can cap it at 150 hard limit and you won't see much difference and also reducing memory voltage will be helpful as well, it should still run at 8000 but it doesn't over clock much after that. Right now it's hitting 9000 so it will probably do 8000 with little less juice.

 

[lolfail9001]

Best solution is to cap the power limit at 150 no matter what. As soon as it hits 150w it should throttle. I bet it won't make much difference in most games. Because most at stock I saw was 163-165 watt. I am sure they can cap it at 150 hard limit and you won't see much difference and also reducing memory voltage will be helpful as well, it should still run at 8000 but it doesn't over clock much after that. Right now it's hitting 9000 so it will probably do 8000 with little less juice.

Once again, voltage manipulations are too risky to commit to. So yes, capping power limit would be the solution, considering that apparently half of VRM is wired to PCI slot for some ridiculous reason.

 

[NKD]

Once again, voltage manipulations are too risky to commit to. So yes, capping power limit would be the solution, considering that apparently half of VRM is wired to PCI slot for some ridiculous reason.

I think they can manipulate it on the ram, because almost every card runs at 9000 overclocked and there should be some level of wiggle room there. Now it might not overclock by a 1000 anymore may be 2-200 but still if they can save any power there thats a must.

Remember at stock I haven't seen any review above 163-165 watt worse case. I think they can easily save the 10-15w by capping it under heavy game loads. If people want to then unlock the power limit, let em take their chances.

 

[Meeho]

FYI, Vdroop is a thing. Also, undervolting is not a solution since it's silicon lottery based and as such AMD even runs a risk of freaking having unstable cards on stock settings. The only viable solution would really be dropping power limit on stock.

Vdroop is always a thing but not relevant to this particular point. I'm not saying it is a valid solution for 480's problem (it isn't), but downvolting certainly won't raise the amp draw from the MBO.

The only viable solution is to limit the max draw on the PCIe slot, either via software by reprogramming the board's power controller or by a redesign if the former is not possible.

 

[CSI_PC]

People are undervolting the 480 as low as 1.025V at stock speeds with a board power draw of 120W.

I am one who would also recommend adjusting volt and power target if buying now (ideally one would wait but not an option for those with it now and waiting for a solution).
But you would still either want to test it with the worst case scenario identified to date Metro Last Light 4K, or measure the power characteristics with a lot of games and a mix of fps and strategy and DX12.
The reason is that the 120W figure (and this needs to be validated by using accurate measurement process) will vary depending upon game, and more importantly I would want to see that measured properly from the riser and auxiliary connector.

My point is that this will ease pressure until AMD has a fix, but one should not take as gospel it is 120W at 1.025V just with some reports saying this is what it draws, without proper analysis and stability testing.
Cheers

 

[Tyns]

Dude seriously, keep these conspiracy theory out of here. AMD on the first day said the reviewers did bring it to their attention and they have already started looking in to it. So you want they to not look and speak before them checking? Go read AMDs first response on reddit. They did say it has already been brought to their attention and they are already working to identify the issue.

Both reddit comments were listed as four days ago and both after the reviews were released. gfxchiptweeter stated they had received feedback but didn't say when they received the feedback and the comment was made after the reviews were released. If AMD confirmed they were made aware of the problem and released the card anyway without a fix, is that better or worse? AMD_Robert made it sound like the issue was with a select number of cards and in specific situations (shown to be patently false on both counts, still have yet to see a test where it passes when testing for PCI-E slot power draw), gfxchiptweeter referred to the problem as being shown "in some cases", not all cases when testing for it.

They weren't proactive about warning their customers about the problem. They should have said "hey, there's this slight problem, most of you won't be affected, but don't try to overclock or use the card in an old or lower end motherboard until we have a fix ". Why is that, do you think?

They released cards without a fix or warning to their customers for a problem they know could damage customers' hardware, and you're bitching at me for being a conspiracy theorist? Nice priorities.

If you don't like the truth, go cry into your pillow or click the little Report link at the bottom of my post.

 

[Spartacus]

I'm guessing the dilemma the engineers are in is if they can drop power enough while impacting performance minimally.
They would rather do that than have to rework the boards. The rework would be very expensive for labor, shipping, etc.

Years ago I was a bench tech. I used to get circuit boards in for repair that had quarter size holes burned clean through
the PCB from overheated driver transistors. I had to do point to point wiring to fix the board (often replacing several chips
that got killed too).

I haven't seen the board, but I'm guessing they could rework the board to fix the power issue with some point to point
wiring. It wouldn't be "elegant" but it would work. They could also replace the 6-pin with an 8-pin, fold the two extra
legs and run the power where they need it with wires. Probably several options for a fix so they don't have to scrap the
boards (that would truly be a disaster).

I don't understand how they painted themselves into a corner on this, but I wish them luck getting out of it.

.

 

[Palladium@SG]

Both reddit comments were listed as four days ago and both after the reviews were released. gfxchiptweeter stated they had received feedback but didn't say when they received the feedback and the comment was made after the reviews were released. If AMD confirmed they were made aware of the problem and released the card anyway without a fix, is that better or worse? AMD_Robert made it sound like the issue was with a select number of cards and in specific situations (shown to be patently false on both counts, still have yet to see a test where it passes when testing for PCI-E slot power draw), gfxchiptweeter referred to the problem as being shown "in some cases", not all cases when testing for it.

They weren't proactive about warning their customers about the problem. They should have said "hey, there's this slight problem, most of you won't be affected, but don't try to overclock or use the card in an old or lower end motherboard until we have a fix ". Why is that, do you think?

They released cards without a fix or warning to their customers for a problem they know could damage customers' hardware, and you're bitching at me for being a conspiracy theorist? Nice priorities.

If you don't like the truth, go cry into your pillow or click the little Report link at the bottom of my post.

I know right.

Then we still have those "Nah, RX480 not adhering to standards doesn't matter. It's not like I'm going to fix or reimburse your system when it goes poof anyway" irresponsible people still recommending the reference RX480 in the face of conclusive evidence.

I'm guessing the dilemma the engineers are in is if they can drop power enough while impacting performance minimally.
They would rather do that than have to rework the boards. The rework would be very expensive for labor, shipping, etc.

Years ago I was a bench tech. I used to get circuit boards in for repair that had quarter size holes burned clean through
the PCB from overheated driver transistors. I had to do point to point wiring to fix the board (often replacing several chips
that got killed too).

I haven't seen the board, but I'm guessing they could rework the board to fix the power issue with some point to point
wiring. It wouldn't be "elegant" but it would work. They could also replace the 6-pin with an 8-pin, fold the two extra
legs and run the power where they need it with wires. Probably several options for a fix so they don't have to scrap the
boards (that would truly be a disaster).

I don't understand how they painted themselves into a corner on this, but I wish them luck getting out of it.

Everyone could care less how hard AMD is going to deal with this mess. If they don't want to deal with the aftermath of broken hardware, then maybe don't release broken hardware to begin with?

 

[madgun]

Not sure why people are being overly defensive about the rx480. Is there a way to know who has vested interests and the ones that don't.

Let AMD bring the card to specifications and the matter can be closed. If there wasn't any hue and cry made on this issue, AMD would have most likely sat upon it with their pants up.

 

[Deleted member 93354]

In defense of AMD, sometimes engineers tweek things up to the last minute trying to make it as good as possible before it goes out the door. Sometimes what seems like an innocuous change can have dramatic consequences in certain scenarios as I am unfortunately aware. (Even if initial testing passes QA)

What likely happened is they likely went to PCIe-SIG for certification and at the time they MIGHT have been under the wire for power consumption. But a later tweak might have put them out of spec.

This kind of leads to a sticky situation. If the situation becomes bad enough, the PCIe-SIG could be forced to recertify every driver change or firmware update for units that are close to the spec limits. I know for a fact when we re-release a change to one of our rating engines, it has to be recertified by an independent professional group. And they are VERY strict about certification.

Again it was stupid decision to push it so close to the limits on the bus. The traces appear to be there for the 8 pin plug. They should have just slapped a 8 pin plug on it for 2 cents more. But it's a mia culpa at this point.

 

[razor1]

You still keeping repeating this false information to people. The board's input volts are constant (12 V and 3.3 V), so if it draws less watts it draws less amps. If they downvolt the GPU or the memory they are not changing the board's voltage at the PCIe slot or 6pin.

Vdroop is always a thing but not relevant to this particular point. I'm not saying it is a valid solution for 480's problem (it isn't), but downvolting certainly won't raise the amp draw from the MBO.

The only viable solution is to limit the max draw on the PCIe slot, either via software by reprogramming the board's power controller or by a redesign if the former is not possible.

You GET VDROOP, and reviewers have noticed it!

Its just purely ignoring the facts that are presented to us if you didn't see that!

This is why the amp's are going up so much on the rails. The motherboard it trying to compensate.

ANY Case Happy 4th everyone!

 

[razor1]

For everyone that wants clarification on the it, like I want on it, just because I like to know things just in case some one asks later on or another problem appears later on a totally unreleated product

AMD admits RX 480 power tuning is “not optimal”

While AMD's statement doesn't indicate precisely what's causing the issues with the RX 480, if it is exceeding its 150W target, either something is causing the card to run in the wrong power state, or it's exceeding its power limit and needs to be throttled back.

Hopefully we get clarification on Tuesday and not another run around the block.

What irk's me is every single time they have been cagey about anything, its always bad lol

They never told us straight out what the power/watt and power consumption was this happens,

They never told us the usual performance graphs (well they did with Mgpu lol) yeah that right there should have a huge red flag, well reason for that we can see now, nothing to see.

They start marketing with malarkey about TAM and VR, well man that is just beating around the bush, what is the performance?

 

[Palladium@SG]

You GET VDROOP, and reviewers have noticed it!

Its just purely ignoring the facts that are presented to us if you didn't see that!

This is why the amp's are going up so much on the rails. The motherboard it trying to compensate.

ANY Case Happy 4th everyone!

It's pretty clear no other GPU except the RX480 tested by PCPer had any vDroop on the PCIE supply rails, or drew average slot power remotely near 75W. It's really hard to fathom why are there still so much inane mental gymnastics to defend the out-of-spec RX480s.

I almost bought the reference RX480 myself, and will be demanding a refund if I had. Out-of-spec products right out of the box are simply unacceptable no matter how much the vendor says it isn't.

 

[razor1]

Razor, ur like a Ferengi some times, seeing consperies and plotting ...

I just want to know what happened without all the crap around it, its not that AMD wanted this, I know that, they didn't plan for this to happen as such, I think they had something that couldn't be saved.

Come on, look nV knew they had a crap product with the FX right? They didn't say anything about it till after the 6 series came out but they made fun of themselves after that.

Thats a good thing, when you can admit you had problems, and AMD failure after failure they haven't been able to look at the themselves and laugh it off, cause they keep f'ing up. But at least after its all said on done, just accept it ya know.

 

[razor1]

It's pretty clear no other GPU except the RX480 tested by PCPer had any vDroop on the PCIE supply rails, or drew average slot power remotely near 75W. It's really hard to fathom why are there still so much inane mental gymnastics to defend the out-of-spec RX480s.

I almost bought the reference RX480 myself, and will be demanding a refund if I had. Out-of-spec products right out of the box are simply unacceptable no matter how much the vendor says it isn't.

Well meeho isn't trying to defend the rx480, I know that, I think I should have been more clear with what I stated from the get go. And the last two days have been drinking pretty heavily lol and still am So yeah I am to blame for that.

 

[c3k]

I'm pulling for AMD, but not a fanboy. I'd like to update the gtx970 in my sig and put that in the place of the 670. The RX480 looked good....until the pcie thing.

Now, I am confident that AMD will fix the power draw issue, I just don't know what effect, if any, their fix will have on performance. If, as stated elsewhere, they have a powerful programmable power control module ic chip, then a software fix may make all this moot. Or, maybe they'll have to send everyone an 8 pin adapter.

The bigger fail was how AMD shipped it this way and how they are handling it...or not, since we have yet to hear their solution.

Ken "who's willing to spend ~$350 for a gpu, not $400+. I'm looking at you, Mr. 1070..."

 

[razor1]

Its a real problem, they probally fix it, but it will take time to get trust back, this shell not happen, the HW fix maybe be easy, but the reputation fix more difficult.

Actually after fixing the problem and then laughing about it, will help their image lol, Its something people that like their products can get behind. Its simple marketing but very effective.

 

[razor1]

Well if AMD blames RTG they are truly fools then, cause Raja and the how he has set up RTG, has nothing to do what happened with this launch. Raja leadership should not be in question because of this, he came too late to his position for him to become a factor of this outcome.

But Raja's leadership will be in question of it takes a longer time to fix this problem..... it has to be fixed before reviews of the 1060 hit cause if it isn't well you are now going to have a huge batch of reviews that are going to point it out and this time there will be no doubt they will get hurt. Too many people will see it and it will become even worse than it is now.

 

[jwcalla]

My theory is that this entire "goof-up" was intentional so that everyone would be distracted away from the real news story, which is that the Polaris architecture is a complete turd; a sort of "Bulldozer" of GPUs.