AMD in Trouble? RX 480 Powergate

variant said:
It's a logical possibility that these reviewers got theirs from the same pre-production batch and thus all have the same flaw, whether it be bios or manufacturing.
Click to expand...

What about the 2 retail parts?

Also, PCPer shows the card pulls 155W at stock in Witcher 3, still over spec.
 
variant said:
What's suspect about it? Unless you think PCP is in some kind of conspiracy with AMD to lie about power draw?
Click to expand...
I never said they are lying just saying that based on the numbers they are actually the anomaly since the majority of cards tested for power draw did see the issue.
 
This is really fun to talk about but in a week we will most likely forgotten this rather small issue.
 
variant said:
What 2 retail parts? Random posters on Reddit and forums isn't remotely verification of anything.

155W isn't anything shocking.
Click to expand...

It's not at all, however when you're aiming for OEM systems that 5W out of spec will cause all sorts of issues that OEMs will prefer to just opt out in many cases.
 
Presbytier said:
If this is able to get repeated over and over on retail cards this will not go away soon.
Click to expand...

yes but right now (on a major release) it isn't that rampant or known yet, I have my suspicion that this exactly what AMD claims it is.
 
Trimlock said:
yes but right now (on a major release) it isn't that rampant or known yet, I have my suspicion that this exactly what AMD claims it is.
Click to expand...
Except in nearly every review where they tested power draw. If AMD does not address this quick enough it will not take long for word to travel.
 
Last edited:
Presbytier said:
Accept in nearly every reveiw where they tested power draw. If AMD does not address this quick enough it will not take long for word to travel.
Click to expand...

This issue in and of itself may not be a big deal but if the 1060 is really on the way and it's power efficiency is like that of its bigger siblings it could get ugly for the 480.
 
heatlesssun said:
This issue in and of itself may not be a big deal but if the 1060 is really on the way and it's power efficiency is like that of its bigger siblings it could get ugly for the 480.
Click to expand...

The RX 480 is about on par with the last gen GTX 970 when it comes to efficiency. the 970 has a 145W TDP (as well as similar performance). It would be difficult for the 1060 to be worse than that.

I wouldn't be surprised if AMD's fix for this is to discontinue the reference card and have AIB makers put more PCI-E connectors on their versions of the card.
 
Price may be a big factor here. I am thinking more and more OEMs probably know about the little over spec and that might be the whole reason they decided to push for the 6 pin. Because lets be real the OEMs try to save every penny and they probably don't want to have a 8 pin connector and have a 500 watt power supply or even less.
 
this is serious. the place where most OEM cut costs are on the motherboard and input devices. the card is a liability. no way OEMs selling low cost systems with the 4GB card are going to risk it. Come on AMD, you are a multibillion dollar company act like it.
 
variant said:
If it's an issue with the RX 480 PCB design, all cards would exhibit this issue. They aren't as PCPer demonstrates. It would have also never passed SIG certification, which it did. If it's a limited number of cards, it's a flaw with a certain batch of cards, either bios or manufacturing.
Click to expand...

lol, haha what? PCPer just demostrated the same behavior the card going out of specs above 150W. But not only that, the PCPer recorded what? only ~70 seconds of gameplay on those charts, but what about extended periods of time according their own test the card was still throttling to an average of 1214mhz and as low as ~1170mhz so guess what, to enjoy the fully performance of the card with full boost you need to crank up the power limit increasing the power consumption to around ~200W which it's worse as even overclocked by own PCPer findings guess what, the card still throttle and with greater and worse variances:

After running through our overclocking process, the results change quite a bit. The temperature is definitely hitting that 90C level but the clock speed is MUCH more variable than it was before, jumping between 1150 MHz and 1300 MHz.
Click to expand...

so at the end of the day what is a RX 480? a card that using full power and out of spec power doesn't perform at the advertised boost clocks without increase power target, temperature target and fan speed. so you end with a card at over 80C that still doesn't perform at full boost clock and overclock as much as 3%-5% making everything even worse in the power department but what's worse is that by your own linked review it show no performance improvement with overclock but not only that, also a worse gaming experience. damn, that's bad, for an overclocked card that still doesn't perform at advertised boost clock by still using out specification power?.

In Hitman at 2560x1440, average performance between stock and overclocked testing shows no change. However, it’s clear that the frame time data when overclocked is more variable and we see slightly less smooth gameplay because of it.
Click to expand...

This card it's an amazing piece of electronic failure not matter how much you and the other couple of AMD fanboys seek to make up this situation. isn't bad at the price? yep, enjoy that mid range performance crown for two weeks.
 
shinotenshi said:
this is serious. the place where most OEM cut costs are on the motherboard and input devices. the card is a liability. no way OEMs selling low cost systems with the 4GB card are going to risk it. Come on AMD, you are a multibillion dollar company act like it.
Click to expand...

Don't forget power supply.
 
480X are in for a major recall. Hopefully someone gets under AMD's skin so they put proper QA money in the future. Half baked and over hyped products just shouldn't make it to the market.
 
razor1 said:
Its out of spec, it has to be fixed otherwise they can't get the pci-e certification, pretty simple, then OEM's will have to make non PCI-e complaint systems, why would they want to do that? They won't. Either AMD fixes this problem or they loss OEM sales.
Click to expand...
Actually, OEMs will first test, whether thier systems are stable and working as intended. And since OEMs typically use so-so components to cut down price, I am sure it would be the Dells, HPs and Lenovos out there, who would raise first red flag on RX480 not meeting specification. At stock.

OC'ing is another matter, but then OC is tuning out-of-spec by definition. This may be one of the reasons of very poor overclockability we see now.

AMD should fix this, though, no debate.
 
If anyone can find info on another card that draws an AVERAGE above the PCI-E spec please send me link
 
lolfail9001 said:
Well, OCN thread recalled tom's review of 960 AIBs that had 960 Strix consume very inconsistent amount of energy from the slot with peaks over 200W and averages... actually i did not find them but the graph was ridiculous.

Power Consumption Details - Nvidia GeForce GTX 960: Maxwell In The Middle
Click to expand...

Yeah I saw it, already confirmed it's just the asus Strix though, and I didn't find averages above 75w.

On maxwell you can change distribution of power from the bios, why the hell isn't this possible here
 
well a poster on OCN posted this,
upload_2016-6-30_5-54-3.png


So if AMD can alter software wise to the 6 pin it should alleviate the PCI motherboard issue.
 
All this arguing in various threads by people who claim to know everything and no one actually has a retail card yet? Someone has to have one to verify issues with the retail cards compared to the review cards.
 
Diseaseboy said:
All this arguing in various threads by people who claim to know everything and no one actually has a retail card yet? Someone has to have one to verify issues with the retail cards compared to the review cards.
Click to expand...
It's already been verified outside of this forum. Plenty of people already have retail cards.
 
IIRC, in the PCI-E certification process, power draw is NOT one of the things they test for. So RX480 PCI-E certification is not in any danger. I looked it up after the first card (Nvidia) started pulling more than 75w from the slot and that was years ago. I'm sure many other cards have exceeded 75w draw through the slot since then. Also, as was posted above, you can supply a lot more amperage through the 6 and 8 connectors if you're using HCS or HCS+ pins without any danger of a meltdown.
 
Creig said:
Guys, if you actually read the PCI-E certification process, power draw is NOT one of the things they test for. So RX480 PCI-E certification is not in any danger. I looked it up after the first card (Nvidia) started pulling more than 75w from the slot and that was years ago. I'm sure many other cards have exceeded 75w draw through the slot since then. Also, as was posted above, you can supply a lot more amperage through the 6 and 8 connectors if you're using HCS or HCS+ pins without any danger of a meltdown.
Click to expand...
The only card i remember pulling more than 75w from the slot from nV was 950 no-connector edition. Which one else? And finally, yes, overcurrenting 6-pin would be more okay.
 
Last edited:
Creig said:
Guys, if you actually read the PCI-E certification process, power draw is NOT one of the things they test for. So RX480 PCI-E certification is not in any danger. I looked it up after the first card (Nvidia) started pulling more than 75w from the slot and that was years ago. I'm sure many other cards have exceeded 75w draw through the slot since then. Also, as was posted above, you can supply a lot more amperage through the 6 and 8 connectors if you're using HCS or HCS+ pins without any danger of a meltdown.
Click to expand...


Different components have different tolerances, as an example you can draw up to 150 watts on a 6 pin connector but because you have crappy power supplies they might not have the gauge of wire to do something like this sustained. The only major difference between an 8 pin and 6 pin is the 8 pin has a extra ground. But is it advisable to do that? NO its not that's what the specs are for, just to make sure even if there are shotty components they will still hit the mark.
 
razor1 said:
Different components have different tolerances, as an example you can draw up to 150 watts on a 6 pin connector but because you have crappy power supplies they might not have the gauge of wire to do something like this sustained. The only major difference between an 8 pin and 6 pin is the 8 pin has a extra ground. But is it advisable to do that? NO its not that's what the specs are for, just to make sure even if there are shotty components they will still hit the mark.
Click to expand...

What the specs are NOT meant for: being broken by reference cards
 
Unfortunately, everything on the PCI-SIG site requires you to be a member to access. Last time it took me quite a while to search and find the PCI-E certification tests and I was surprised that power draw was not one of the things they test for. Yet they have a spec for power draw. I remember scratching my head wondering why they would set a spec if they're not going to bother testing for it. Maybe someone else can dig up the certification white paper and post it. That would hopefully tell us for sure.
 
WorldExclusive said:
Because they've already been exposed as liars for page views. No need to tell the truth now.
Click to expand...
They don't report negative AMD rumors, it drives away their target demographic. Same could be said about a few other clickbait sources, who will also probably not discuss this.

Related: Linus tried to bury RAM-gate last year (Nvidia side).
 
Creig said:
Unfortunately, everything on the PCI-SIG site requires you to be a member to access. Last time it took me quite a while to search and find the PCI-E certification tests and I was surprised that power draw was not one of the things they test for. Yet they have a spec for power draw. I remember scratching my head wondering why they would set a spec if they're not going to bother testing for it. Maybe someone else can dig up the certification white paper and post it. That would hopefully tell us for sure.
Click to expand...
They wouldn't be a SIG if everyone was allowed in.
 
Creig said:
Unfortunately, everything on the PCI-SIG site requires you to be a member to access. Last time it took me quite a while to search and find the PCI-E certification tests and I was surprised that power draw was not one of the things they test for. Yet they have a spec for power draw. I remember scratching my head wondering why they would set a spec if they're not going to bother testing for it. Maybe someone else can dig up the certification white paper and post it. That would hopefully tell us for sure.
Click to expand...


I don't know about the check list but power draw is one of the catagories they stipulate on.

Many companies from OEM's to motherboard manufacturers are part of the PCI-SIG committee, They have all weighed in on what they think the specs should be, and then ratified those specs, and those specs give leeway to really crappy components too. Its just a way to self regulate component quality when put together and this has much more to do with warranties and OEM's than anything else.
 
Sheppard said:
Umm your logic is not sound, if all pcie gets its power from the same traces, wouldn't that mean you violate that spec every time you run crossfire then?
Click to expand...

Here's the thing: a 290X doesn't violate the spec - it's 36W average draw from the slot. That means this does not put undue stress on your mobo, and instead shifts the burden to the PCIe Power Connectors. Crossfire all you want, no issues.

The RX480 puts 86W average (stock) of load towards the slot. Alone and in a well-built board, you probably won't run into issues. Overclocked, you may run into some even on good boards (Tom's indicated it was going up to 100W on the slot - even they had to halt CF testing because of the possibility of damaging their test bench). CF is worse, and what more of you OC'd in CF?

Even if it doesn't blow, it could introduce issues with other components running through PCIe (like audio, etc). And on cheap, budget boards (which this GPU is likely to be placed on), you may not be able to expect the same kind of tolerance that more expensive ones do.
 
some of the debate unfortunately seems to focus on no change, but believe AMD is gonna fix this, they have no choice. Also reading elsewhere about the 6990 which did the same thing, but was it ever "fixed"? If not were there any failures reported?
 
JustReason said:
some of the debate unfortunately seems to focus on no change, but believe AMD is gonna fix this, they have no choice. Also reading elsewhere about the 6990 which did the same thing, but was it ever "fixed"? If not were there any failures reported?
Click to expand...


Yeah but that card wsn't a OEM seller lol. So at the end it didn't matter for it.

Actually I don't think it drew much from the PCI-e its power draw was all from its connectors.

Anyone spending that kind of money for such a card probably are getting top end components anyways.
 
razor1 said:
Yeah but that card wsn't a OEM seller lol. So at the end it didn't matter for it.

Actually I don't think it drew much from the PCI-e its power draw was all from its connectors.

Anyone spending that kind of money for such a card probably are getting top end components anyways.
Click to expand...
AFAIK there is no evidence that any of the oem cards are having any issues.....All we know about are these non OEM cards supposedly. Reason i say this is OEM version being clocked slower likely dont have any issue at all?
 
primetime said:
AFAIK there is no evidence that any of the oem cards are having any issues.....All we know about are these non OEM cards supposedly. Reason i say this is OEM version being clocked slower likely dont have any issue at all?
Click to expand...


Yeah and we also don't know if there are any OEM cards out there yet lol.

But for the r6990 the cost of such a system an OEM can get away with a few failures because its going to be only at a card or power supply level.

At the cost of system that will house a rx 480 and the volume sales of such a system, is a big risk for them.
 
Last edited:
You must log in or register to reply here.
Back
Top