The 480: power consumption, PCI-E powerdraw

[3DVagabond]

Just to interject something that seems to be being missed, average is not continuous. Continuous is a very much more demanding situation. I'm not going to pretend to be an engineer. I'm not a super computer geek either. I used to be a super audio geek though. So just to understand circuits and continuous power a bit.

If something is honestly rated for 75W continuous that is a lot of power. For example an amplifier that is rated at 200W continuous at 8 ohms would never be able to put out a continuous sine wave signal of 75W for more than a few minutes typically before shutting down, or worse. There are exceptions but that is because those are rated extremely conservatively and are actually capable of putting out far more than their rated 8 ohm power.

I know a particular studio monitor that is also sold as a consumer product. It's a real professional product but it was adopted by the audiophile community to the point that it actually became mainstream for consumer use. As a pro piece it's rated at 75W RMS and a continuous rating of 35W sine wave. When it's listed for typical consumer use it's rated @ 250W RMS. That is because the typical consumer rating is maximum amplifier power and actually has nothing to do with the maximum power the circuit, speaker in this example, can handle. That 250W assumes you are playing music and the amplifier is capable of delivering that power cleanly.

Anyway, that's a long way of saying that average power usage and continuous power usage are not the same thing. So, just because something is using more than 75W average does not mean it will damage a circuit designed to deliver 75W continuously.

 

[Abwx]

???

Don't know why.

I just quoted your post but it was not directed to you, of course, but to the
member to whom i was answering, sorry for the way i posted the things...

 

[Stuka87]

Your other GPUs probably don't pull more then 60watts thru the PCIe slots but don't let facts get in the way of a good rant.

I think his point was that he had three of them, so even if they were 60W each, thats still 180W through the motherboard.

 

[el etro]

I just quoted your post but it was not directed to you, of course, but to the
member to whom i was answering, sorry for the way i posted the things...

No problem!

 

[IEC]

Tore down my Sapphire 4GB cards.

They are using IR 3567B dual output 6+2 phase PWM controllers:
http://www.infineon.com/cms/en/prod...?productType=5546d462533600a401533d2546765ec7

Specific chip marking is:
IR
3567B
C547P
1DRY

Interestingly, my Sapphire 4GB cards are both "4GB" in name only - they physically have 8x Samsung (616) K4G80325FB HC25 GDDR5 modules.
http://www.samsung.com/semiconductor/products/dram/graphic-dram/gddr5-component/K4G80325FB?ia=759

 

[bononos]

........
The funny part here is that RX 480 4GB's reference PCB/VRM and MoSFETs are higher quality than they are on the $700 GTX 1080 FE card.
.......

Did the vrm components comparing the 480 vs Pascal get reviewed somewhere or did you check it out yourself?

Edit- I just saw the teardown thread by IEC.

 

[MrTeal]

Tore down my Sapphire 4GB cards.

They are using IR 3567B dual output 6+2 phase PWM controllers:
http://www.infineon.com/cms/en/prod...?productType=5546d462533600a401533d2546765ec7

Specific chip marking is:
IR
3567B
C547P
1DRY

Interestingly, my Sapphire 4GB cards are both "4GB" in name only - they physically have 8x Samsung (616) K4G80325FB HC25 GDDR5 modules.
http://www.samsung.com/semiconductor/products/dram/graphic-dram/gddr5-component/K4G80325FB?ia=759

That's... interesting. Is there a dual bios switch on that card? Not my card obviously, but I'd be really, really tempted to flash it with a 8GB BIOS and see what happens. For science.

 

[Doom2pro]

What will everyone talk about on July 6th when AMD fixes this? When there are no more sketchy reports of board failures?

I guess everyone will have to focus on trying to claim AMD gimped their cards to fix the problem when so far under volting the cards INCREASES performance per several popular benchmarking websites.

So that settles it... The next debate is over wether or not AMDs fix gimps the cards or increases performance.

Have fun

 

[Azix]

https://www.youtube.com/watch?v=yNMJ5U5ZcXk

this guy tests it. translation anyone? on reddit they are saying even 5.2 A did nothing to the pins

 

[SPBHM]

https://www.youtube.com/watch?v=yNMJ5U5ZcXk

this guy tests it. translation anyone? on reddit they are saying even 5.2 A did nothing to the pins

isn't 5.5 A the official limit for 12v, while PCper got a stock 480 using constant 6.5 with spikes to over 7 A?

unless you are talking 5.2 each and not total!?

Your other GPUs probably don't pull more then 60watts thru the PCIe slots but don't let facts get in the way of a good rant.

yes I posted a link, 295x2 uses 39.6W peak on the gaming test (their bitcoin test is much lower) with average under 30W
http://www.tomshardware.com/reviews/radeon-r9-295x2-review-benchmark-performance,3799-14.html

while the same site shows 85W average for the 480
http://www.tomshardware.com/reviews/amd-radeon-rx-480-polaris-10,4616-9.html

basically like PCper observed this is way higher than any other card, and it's kind of a bizarre behavior, like it looks like the slot and connector power usage are tied, they are always similar, while in other cards that's normally not the case at all.

basically if AMD had the 480 limited to something like 50W on the slot, and let the 6pin use 100W it wouldn't really be a problem, but the way they did it's 75/75 (and the official limit for the slot is 66W for the 12v portion)

 

[Abwx]

https://www.youtube.com/watch?v=yNMJ5U5ZcXk

this guy tests it. translation anyone? on reddit they are saying even 5.2 A did nothing to the pins

He s testing a single contact on the connector with up to 5.2A current and he measure the voltage loss so he can compute the power dissipation of this contact, there s 4 contacts, so we can deduce that this PCIe could take 20.8A, that s 250W...

FTR that s a S775 P4 card...

 

[SPBHM]

He s testing a single contact on the connector with up to 5.2A current and he measure the voltage loss so he can compute the power dissipation of this contact, there s 4 contacts, so we can deduce that this PCIe could take 20.8A, that s 250W...

FTR that s a S775 P4 card...

did he actually test 20 A 12v!?!? I mean all pins at the same time for an extended period? + the CPU load while gaming? I think that motherboard would hopefully shutdown with such load...

also, I think you would need to simulate the VGA PCB contacts to properly test this?

one thing from all this is clear, the 480 is special case compared to other cards, it's out of spec.
now what we need is a clear official statement from motherboard manufacturers telling us if this is OK or not, and obviously from AMD... but nothing new for now?

 

[StrangerGuy]

did he actually test 20 A 12v!?!? I mean all pins at the same time for an extended period? + the CPU load while gaming? I think that motherboard would hopefully shutdown with such load...

also, I think you would need to simulate the VGA PCB contacts to properly test this?

one thing from all this is clear, the 480 is special case compared to other cards, it's out of spec.
now what we need is a clear official statement from motherboard manufacturers telling us if this is OK or not, and obviously from AMD... but nothing new for now?

Why would the mobo manufacturers want to do that? It's not their responsibility to ensure the boards work with out-of-spec hardware.

AMD shouldn't get a free pass for shipping out of spec hardware, the proper thing is to do a recall like Intel who did it for entire shipments of far more numerous 6-series B1 mobos in 2011 for broken SATA2 ports. Nobody should care how hard or costly that will take, they made their own bed.

 

[Abwx]

did he actually test 20 A 12v!?!? I mean all pins at the same time for an extended period? + the CPU load while gaming? I think that motherboard would hopefully shutdown with such load...

also, I think you would need to simulate the VGA PCB contacts to properly test this?

one thing from all this is clear, the 480 is special case compared to other cards, it's out of spec.
now what we need is a clear official statement from motherboard manufacturers telling us if this is OK or not, and obviously from AMD... but nothing new for now?

No, he tested a single contact with 5.2A but as shown in his voltmeter he had a crappy contact on purpose, yet the contact work reliably with low power losses at 3X its current rating, beside he measured the power dissipated by the contact.

The MBs are just routing the voltage from the ATX connector to the PCIes, in this case the power is limited by the virtue of the two slots, the more the slots the more the current capability before the MB detect the over powering and shut down the PSU.

If anything this show that all the outcry we read here is just posturing from a given public but technically speaking there s absolutely no issue at all, 80W is 1.67A/contact...

 

[Abwx]

AMD shouldn't get a free pass for shipping out of spec hardware, the proper thing is to do a recall like Intel who did it for entire shipments of far more numerous 6-series B1 mobos in 2011 for broken SATA2 ports. Nobody should care how hard or costly that will take, they made their own bed.

AMD are still within safe operating area but if you are concerned by blatantly outspecced cards you should post here and duplicate there your proposition :

http://forums.anandtech.com/showthread.php?t=2478547&page=3


Infraction issued for thread crapping.

-Rvenger

 

[StrangerGuy]

Yeah, we all know Abwx and the everything is A-OK crowd are so willing to put their money on the line for the poor sods with damaged motherboards using the RX480.

Maybe you guys should also say eating slightly expired food is no problemo too? Looks good and tastes good right, so whats the problem?

 

[Piroko]

Mmm you are sure? I remember that 75w is for a video card and 25 w for all other devices, no matter what the speed of the slot is. Mmn ill need to re check that one.

You are sort-of correct. 75W is specified for the use of video cards only and it's limited to physical x16 slots. However, it's not limited to slots that have 16 PCIE lanes, a slot with only 4 PCIE lanes but the x16 connector may be able to house a video card and be rated for 75W as well.

That's probably the big reason why board manufacturers are hesitant to only put x16 connectors on their MBs, they would either have to beef up their power delivery or be very thorough with warnings that only e.g. slots 1 and 4 are suited for GPU use (potentially by posting a UEFI warning if a device tries to initialize high power mode on one of the other slots?).

 

[zinfamous]

Yeah, we all know Abwx and the everything is A-OK crowd are so willing to put their money on the line for the poor sods with damaged motherboards using the RX480.

are there actual cases of people frying their motherboards with normal use of a 480X? afaik, overclocking remains a "user beware" proposition does it not? Also, do you think the guy that tried to mine with 3x 480s without powered risers has valid complaint against AMD?

Maybe you guys should also say eating slightly expired food is no problemo too? Looks good and tastes good right, so whats the problem?

Um, yeah? You realize those dates are broad recommendations, right? The date is almost completley irrevelvent, because it can not control how an individual person stores their food compared to the next person. Do you actually toss out your food simply because of some extremely arbitrary date stamped on the package?

 

[Minkoff]

Yeah, we all know Abwx and the everything is A-OK crowd are so willing to put their money on the line for the poor sods with damaged motherboards using the RX480.

Maybe you guys should also say eating slightly expired food is no problemo too? Looks good and tastes good right, so whats the problem?

Yeah, we all know others and the "everything with a certain company is doomed" crowd.

I don't see a problem with what Abwx said. Yes, the card maybe out of spec (don't have full access to PCIe specs) but there's no conclusive evidence (or any evidence) that it damages mobos and the video is actually very interesting.

BTW, if it looks good, smells good and tastes good, there is no problem.

 

[Erenhardt]

Yep. It definitely is not as bad as nvidia killing hardware with drivers. Talk about forced upgrades.

 

[StrangerGuy]

are there actual cases of people frying their motherboards with normal use of a 480X? afaik, overclocking remains a "user beware" proposition does it not? Also, do you think the guy that tried to mine with 3x 480s without powered risers has valid complaint against AMD?

Um, yeah? You realize those dates are broad recommendations, right? The date is almost completley irrevelvent, because it can not control how an individual person stores their food compared to the next person. Do you actually toss out your food simply because of some extremely arbitrary date stamped on the package?

OC? The 3x 480s guy? Both are already voluntary "at your own risk" out-of-spec cases. How is that supposed to prove anything about a stock 480 that exceeds 75W average on the PCIE slot spec out of the box?

And just because nothing happens to normal reference RX480 users yet doesn't absolves AMD of wrongdoing or guaranteed safe operation for the long term. Because of the very reason it was shipped out of spec. It's simple Occam's Razor at work.

Yeah sure man, bring your opinion to the food regulatory agencies since you think you know better than them. Another of your worthless "talk is cheap" opinion.

 

[IEC]

I changed the arrangement of my cards to have one in a PCI-e 3.0 slot, and one in a PCI-e 2.0 slot. I've also overclocked my cards. Still monitoring for any abnormalities.

Running without issue so far.

 

[zinfamous]

OC? The 3x 480s guy? Both are already voluntary "at your own risk" out-of-spec cases. How is that supposed to prove anything about a stock 480 that exceeds 75W average on the PCIE slot spec out of the box?

And just because nothing happens to normal reference RX480 users yet doesn't absolves AMD of wrongdoing or guaranteed safe operation for the long term. Because of the very reason it was shipped out of spec. It's simple Occam's Razor at work.

Right, but that is different. I agree that AMD shipped something out of spec. I'm not saying that isn't a problem--but you're making claims about lawsuits and legal issues regarding safety for individuals and their property that currently have no standing. Not yet, anyway. This, for now, is grasping at straws. Until something actually happens to point to real wrong doing from AMD, it isn't really a problem.

Yeah sure man, bring your opinion to the food regulatory agencies since you think you know better than them. Another of your worthless "talk is cheap" opinion.

I'm sorry you feel this way, but this isn't opinion. I take it you were raised on boxed and prepared food and haven't really had much experience in preparing food yourself? I advise you not to criticize easily-acquired facts when in this case, it appears that I am far more educated on this issue than you. Humans did quite well for generations developing universal preservation techniques that are still well in practice. Only a few simple steps can be taken to understand these points. Arbitrary doesn't exactly mean useless--if you understand why expiration dates are recommendations and not requirements, you might start to be a less wasteful consumer of packaged food substitutes.

The FDA has a useful role but is also extremely far reaching when it comes to food safety in certain cases. Simple question: do you think a tub of yogurt will last longer in a fridge or in the back of your car during the summer? After answering this question, what do you think the stamped exp date on your yogurt container assumes when it comes to how that stamp expects you to keep your yogurt. 4 degrees C? 10 degrees C? 28 degrees C? Lacking in information, don't you think?

Tell me--do you keep your eggs in the fridge? If so, why?

I know, off topic here

 

[Abwx]

Yeah, we all know Abwx and the everything is A-OK crowd are so willing to put their money on the line for the poor sods with damaged motherboards using the RX480.

Maybe you guys should also say eating slightly expired food is no problemo too? Looks good and tastes good right, so whats the problem?

Unlike you i m not talking out of fairy tales and other inventions, the german video show that PCIe connectors wont sweat with a current that is 3X the connector rating, so 1.1x the rating is not even worth discussing as far as reliability is a concern.

 

[coercitiv]

I know, off topic here

Off topic but not as far as one might think: both engineering specs and food expiration dates take into account certain conditions and provision for variance.

This is why:

1. AMD should and will issue a fix to make their product adhere closer to the standard.
2. people using the RX 480 right now can do so safely because of the provisions built into these specs.

Only by targeting the specs can the safety margins do their job properly.