AMD once again violating power specifications? (AMD RX-480)

[stconquest]

15 minutes ago, Bleedingyamato said:

I have no idea what you're talking about sorry.  I was just asking if this is an actual problem or not.  I know you can ruin a CPU or GPU if you pump too much power through them so it stands to reason that a motherboard can have the same problem if a GPU draws too much doesn't it?

Absolutely.  You should grab a tea and read the whole thread

[Megazero]

Jay bring alot of valid point and things that we should be more concern about.

[Briggsy]

5 minutes ago, awesomeness10120 said:

GTX 960 had same problem, but was fixed by more power connectors.

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-960,4038-8.html

yeah that's been discussed to death already. it was the asus strix 960 that was peaking past 75 watts because it was factory overclocked and custom, not the 960 in general. And, even then the strix 960 is not as bad as the 480 reference.

[Zah]

41 minutes ago, ScootsMcgoots said:

I don't know which is worse.....sorta false advertising ( 970 fiasco) or potentially damaging your mobo. Also to the people saying " well if you have a motherboard with good power delivery it will be fine"; I'd imagine the people buying a budget friendly gpu won't have a high end motherboard with great power delivery ( unless you are saving money buy putting two of these in sli). 

Well, AMD did advertise it using SLI to beat a 1070, so...

[stconquest]

5 minutes ago, Zah said:

Well, AMD did advertise it using SLI to beat a 1070, so...

That was a GTX 1080, and it holds true from more than one source.  Dual RX 480s were like 9% slower than a single GTX 1080 in relative performance(average) in a slew of (like 20) different games.

[Camisto]

8 minutes ago, stconquest said:

That was a GTX 1080, and it holds true from more than one source.  Dual RX 480s were like 9% slower than a single GTX 1080 in relative performance(average) in a slew of (like 20) different games.

thats not the point.  the point is this power problem is compounded in crossfire configuration.

[stconquest]

2 minutes ago, Camisto said:

thats not the point.  the point is this power problem is compounded in crossfire configuration.

That's maybe your point.  It was not mine.

[Camisto]

1 minute ago, stconquest said:

That maybe your point.  It was not mine.

the comment you quoted was not discussing the performance of crossfire rx 480's, I was only helping you understand what you were quoting.

[stconquest]

1 minute ago, Camisto said:

the comment you quoted was not discussing the performance of crossfire rx 480's, I was only helping you understand what you were quoting.

Wow, okay.  Pay attention.  The guy I quoted said AMD showcased that dual RX 480s can beat a GTX 1070.  I was only correcting that remark. 

 

...and I added a bit at the end.

[Bleedingyamato]

48 minutes ago, stconquest said:

Absolutely.  You should grab a tea and read the whole thread

What should I do if I'm all out of tea?  ?  lol

 

Anyway, in the mean time this is a legitimate problem correct?  

 

The video that Megazero shared seems to say it is something to be concerned about.

[stconquest]

2 minutes ago, Bleedingyamato said:

What should I do if I'm all out of tea?  ?  lol

 

Anyway, in the mean time this is a legitimate problem correct?  

 

The video that Megazero shared seems to say it is something to be concerned about.

It very well could be.  It's day one.  This just started.  Grab a large case of tea from the supermarket tomorrow. 

[vanished]

1 hour ago, patrickjp93 said:

You forgot the 3.3V rail.

Actually that raises an interesting point.  We're all flipping out over how much power is being drawn through the PCIe slot, but if we're going to be "proper" about this, shouldn't we independently test the 12 v and 3.3 v rails?  Even if a card is only drawing, say, 74 W though the slot, if 100% of that is coming off the 12 v rail, it's still breaking spec, no?

[maulemall]

when the only thing you have to sell is crisis then everything needs to be a crisis for you to be relevant.

is there a problem? don't know really. the only thing that's actually known as fact is that there is a report of a problem. And It is being investigated.

NOBODY on this board or in the thread has done any actual testing and the screaming and accusations from a couple people only proves that when someone knows diddly about what's going on they say it really loud.

 

I do know that when you call a company liars and bad mouth them with no proof you open yourself up to litigious consequences and in some states it is a criminal offence . In my state you get charged with a felony.

[Megazero]

1 minute ago, maulemall said:

when the only thing you have to sell is crisis then everything needs to be a crisis for you to be relevant.

is there a problem? don't know really. the only thing that's actually known as fact is that there is a report of a problem. And It is being investigated.

NOBODY on this board or in the thread has done any actual testing and the screaming and accusations from a couple people only proves that when someone knows diddly about what's going on they say it really loud.

 

I do know that when you call a company liars and bad mouth them with no proof you open yourself up to litigious consequences and in some states it is a criminal offence . In my state you get charged with a felony.

... I read your post like 3 times and I still dont understand why are you quoting me. I dont call AMD lying, I dont bad mouth them, and if you get charged because you say bad thing about a company then I feel sorry for you, you live in a shit place. 

Also for the most part, what we are doing here is report and discuss about some actual testing that have been done by actual people. AND AMD themselves already acknowledge this problem and said that a simple driver issue wont fix it, so this is fact.

[maulemall]

Just now, Megazero said:

... I read your post like 3 times and I still dont understand why are you quoting me. I dont call AMD lying, I dont bad mouth them, and if you get charged because you say bad thing about a company then I feel sorry for you, you live in a shit place. 

Also for the most part, what we are doing here is report and discuss about some actual testing that have been done by actual people. AND AMD themselves already acknowledge this problem and said that a simple driver issue wont fix it, so this is fact.

Dude I am sorry I didn't know it quoted you.

No offense meant in any way.

[gilang01]

http://www.pcper.com/reviews/Graphics-Cards/Power-Consumption-Concerns-Radeon-RX-480/Overclocking-Current-Testing

[zMeul]

43 minutes ago, Ryan_Vickers said:

Actually that raises an interesting point.  We're all flipping out over how much power is being drawn through the PCIe slot, but if we're going to be "proper" about this, shouldn't we independently test the 12 v and 3.3 v rails?  Even if a card is only drawing, say, 74 W though the slot, if 100% of that is coming off the 12 v rail, it's still breaking spec, no?

yes, it does - the 12V PEG rail has an Amperage rating

[vanished]

2 hours ago, zMeul said:

yes, it does - the 12V PEG rail has an Amperage rating

Exactly.  So, not only should we go beyond checking the total power consumption of the card against the sum of all suppliers by breaking it down by PCIe cable power and PCIe slot power, but we should break down the slot power into the different rails.  Has anyone tested it that closely?

[Bleedingyamato]

2 hours ago, stconquest said:

It very well could be.  It's day one.  This just started.  Grab a large case of tea from the supermarket tomorrow. 

True.  Hopefully AMD will resolve this quickly.  

 

Not a bad idea.  But what kind I'd have to think about.  So many kinds of tea it's hard to decide.  ?

[LAwLz]

4 hours ago, Ryan_Vickers said:

So, the 75 W slot is rated at 12 volts, 5.5 amps, +/- 8%... so 12 * 5.5 * 1.08 =  71.28 W... what?

 

So in one game the 480 falls short of the 970 and in another it's basically at 980 ti performance?  Hm...

You forgot the +3.3 volt rail. With that the maximum becomes above 75 watts, but you are not allowed to stress both rails to their maximum at the same time. Their combined power is not allowed to go above 75 W. 

 

 

3 hours ago, awesomeness10120 said:

GTX 960 had same problem, but was fixed by more power connectors.

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-960,4038-8.html

The GTX 960 did not have the same problem. One particular version of the 960 (the Asus Strix),which was factory over clocked and non-reference, was PEAKING above 75 W but the overall consumption was below it. Other versions did not show the same level of peaks either by the way. 

The RX 480 at stock, with the reference design, is above 75 W on average, and peaks even higher. 

 

Motherboards are built to handle peaks, but not consistent abuse. That is the problem. 

 

 

12 minutes ago, Ryan_Vickers said:

Exactly.  So, not only should we go beyond checking the total power consumption of the card against the sum of all suppliers by breaking it down by PCIe cable power and PCIe slot power, but we should break down the slot power into the different rails.  Has anyone tested it that closely?

Yes

So far, 6 different websites has tested it that closely, and it goes above the spec according to all of them. The PCPer article is the most informative and it was linked just a few posts above yours. They show a detailed breakdown of what each rail uses. 

Please bear in mind that the limit is 75 watts combined. PCPer mentions it in the article too. 

[vanished]

13 minutes ago, LAwLz said:

You forgot the +3.3 volt rail. With that the maximum becomes above 75 watts, but you are not allowed to stress both rails to their maximum at the same time. Their combined power is not allowed to go above 75 W. 

 

 

The GTX 960 did not have the same problem. One particular version of the 960 (the Asus Strix),which was factory over clocked and non-reference, was PEAKING above 75 W but the overall consumption was below it. Other versions did not show the same level of peaks either by the way. 

The RX 480 at stock, with the reference design, is above 75 W on average, and peaks even higher. 

 

Motherboards are built to handle peaks, but not consistent abuse. That is the problem. 

 

 

Yes

So far, 6 different websites has tested it that closely, and it goes above the spec according to all of them. The PCPer article is the most informative and it was linked just a few posts above yours. They show a detailed breakdown of what each rail uses. 

Please bear in mind that the limit is 75 watts combined. PCPer mentions it in the article too. 

 

So the conclusion I was looking for earlier is something like this:

• Yes, the 480 breaks specs in a pretty big way.  Its constant loads are above spec, not just the spikes
• Yes, this is a problem
• No, it's not just a problem because AMD does it
• Other cards (yes Nvidia cards) have had issues too but not like this.  The ASUS 960 was the worst example we can come up with and that only spiked above for short durations.  Yes, this is a difference that matters.

[FatalityDiablo]

just wait a few weeks for 480s from board partners. Reference cards should be recalled thou, since not everyone has a modern mobo in their system.

[mariushm]

31 minutes ago, Ryan_Vickers said:

 

So the conclusion I was looking for earlier is something like this:

• Yes, the 480 breaks specs in a pretty big way.  Its constant loads are above spec, not just the spikes
• Yes, this is a problem
• No, it's not just a problem because AMD does it
• Other cards (yes Nvidia cards) have had issues too but not like this.  The ASUS 960 was the worst example we can come up with and that only spiked above for short durations.  Yes, this is a difference that matters.

Motherboards have to provide those 5.5A at 12v to EACH pci-e x16 slot on them, so they are designed with large copper fills and traces going to the pci-e slots area, they are capable of providing those watts to the slots. The motherboards won't really care that the RX 480 pulls an average of 80 watts instead of the maximum recommended in the standard of 12v x 5.5a = 66 watts.

If you don't populate every pci-e slot with cards, giving a bit of power that would otherwise go to the second slot, to the first slot, is not an issue.

 

The issue is that this 12v power goes into the video card through the pci-e slot through 5 tiny contacts. The engineers that designed the pci-e slots and these contacts inside the slots also had to figure out the contact surface, the contact pressure, how much resistance the contacts add, how much heat this resistance produces, how that heat in turn will heat the plastic over the contacts, how that heat will gradually weaken the metal in the contacts and loosen the pressure on the video card pads and so on and so on....

 

They rated everything and figured those pins can do in total a bit more than 5.5A at 12v, but recommended this 5.5A figure as a safe value.

 

With the same reasoning, motherboard manufacturers don't design their motherboards with the exact 5.5A per slot figure in mind, the traces going from the internal 12v layers to the pci-e slots are short and thick enough to handle much more than 5.5A, so they can handle those peaks of up to 150w for a ms or so without any issues.

That's why you see in various articles that motherboard manufactures say peaks are no problems and I remember reading in an article a quote from a motherboard manufacturer saying up to 95w constant consumption is relatively safe (at least for their boards).  Because it is. Nobody designs a product that would die if something uses more than a few percents of power over the maximum recommended, because they can never known when someone makes a bad card... would you prefer to have boards returned for RMA due to a bad pci-e card or would you rather overspec your design in case user buys such a bad card?

 

So yes, if we go by the standards and the values listed in the standards, the RX 480 doesn't respect the standard because on average it draws 82 watts instead of maximum 66 watts from the slot. The peaks of up to 150 watts from the slot are not an issue, all motherboards are designed to handle those. All video cards on the market will have peaks over the 66w rating. 

 

The extra constant power consumption also shouldn't be too much of an issue because it's not that much extra, any good motherboard will handle it just fine. It would have been a big deal if the constant power consumption would have been let's say 100 watts or more. 

 

Yes, the RX480 uses more from the slot than recommended, it's not nice but won't kill motherboards.

The fix could probably be something as simple dropping the maximum boost frequency from 1266 Mhz to 1200-1233 Mhz, which should drop the maximum power consumption by about 10-15 watts. it's logical the most power consumption would be at high frequencies, so limiting the maximum power would in turn limit the power drawn from the slot, so you'll have less power consumption in average.

[Megazero]

4 minutes ago, mariushm said:

 

 

The extra constant power consumption also shouldn't be too much of an issue because it's not that much extra, any good motherboard will handle it just fine. It would have been a big deal if the constant power consumption would have been let's say 100 watts or more. 

 

Yes, the RX480 uses more from the slot than recommended, it's not nice but won't kill motherboards.

The fix could probably be something as simple dropping the maximum boost frequency from 1266 Mhz to 1200 Mhz. it's logical the most power consumption would be at high frequencies, so limiting the maximum power would in turn limit the power drawn from the slot, so you'll have less power consumption in average.

Isnt the test from Tom say the average draw from PCIE of the 480 is 100W?

Also really? the fix is to underclock the card? Why should anyone buy a card and HAVE to underclock it to make it work? 

Another problem with your statement is "good motherboard", how do you define "good"? Remember who this card is targeting at: budget gamer. Those that spend only $200-$240 on a GPU is likely to have a cheap Mobo too. They ofc wont pair this with say... at $150+ mobo. So this make this problem even worse.

[laminutederire]

8 hours ago, patrickjp93 said:

Nope, 6 more hours. So far, so good. Also, no, I mean mathematics that prove, without testing needed, various things. Why does calculus work? It can be proved that it works. Why does physics work? Because the calculus says it will.

Mathematics doesn't prove much in real life. In that case, they allow us to use a model which happen to be close enough to reality, as experimentation prove it to be acceptable to a certain degree.

I know that mathematics are so much more intense and beautiful when they aren't done for physics though.