The 480: power consumption, PCI-E powerdraw

[Goatsecks]

I am surprised that no one else has started a thread on this, but what is happening with the 480's power figures? Currently it uses 12% more wattage than the gtx1070 and 3 watts less than the gtx1080 (during typical gaming).

It has also been reported that:

AMD’s Radeon RX 480 draws an average of 164W, which exceeds the company's target TDP. And it gets worse. The load distribution works out in a way that has the card draw 86W through the motherboard’s PCIe slot. Not only does this exceed the 75W ceiling we typically associate with a 16-lane slot, but that 75W limit covers several rails combined and not just this one interface.

(http://www.tomshardware.com/reviews/amd-radeon-rx-480-polaris-10,4616-9.html).

It seems that there is something potentially wrong, no? Has the card shipped with too high clock speed?

 

[Silverforce11]

http://www.pcinvasion.com/amd-looking-rx480-pcie-compliance-failure-reports

AMD’s new RX480 card has launched today, which means benchmarks and power efficiency tests galore popping up across the internet. Among the many articles, reviewers at Tom’s Hardware reported something troubling regarding power draw on the PCI Express slot.

In their testing, they found that the RX480 they had received for review drew 86W through the PCIE slot. That’s 11W above the maximum 75W specification required to meet compliance.

AMD’s Senior VP and Chief Architect Raja Koduri was fortuitously present for an ‘Ask Me Anything’ event earlier today, and addressed the issue when asked (‘gfxchiptweeter’ is Koduri).

“We have extensive testing internally on our PCIE compliance and RX480 passed our testing,” he writes. “However we have received feedback from some of the reviewers on high current observed on PCIE in some cases. We are looking into these scenarios as we speak and reproduce these scenarios internally. Our engineering team is fully engaged.”

If widespread, the non-compliant power draw would be a serious issue for AMD. However, there are other possibilities. The card received for review at Tom’s could be faulty in some way, or something went array with the testing. Reviewers were working with cards that could flash the BIOS to use either 4GB or 8GB of VRAM (to save sending a pair of cards to everyone, presumably), so that’s another difference between review and market models.

For now, it’s something for AMD’s engineers to take a good look at.

Update: A further comment from AMD states that the RX480 met ‘Peripheral Component Interconnect Special Interest Group’ (PCI-SIG) compliance testing. So it passed external industry testing as well as internal tests.

“Obviously there are a few GPUs exhibiting anomalous behavior, and we’ve been in touch with these reviewers for a few days to better understand their test configurations to see how this could be possible,” that post says. “We will have more on this topic soon as we investigate, but it’s worth reminding people that only a very small number of hundreds of RX 480 reviews worldwide encountered this issue.”

A user on reddit also posted the GPUz shot of his RX 480 under gaming load and it was running at 1.29vcore, well above the 1.15v max allowed voltages under Wattman.

This looks like a Crimson problem failing to regulate the power and running some cards over-volted.

 

[khon]

AMD could have avoided this problem very easily by using an 8-pin connector instead of a 6-pin connector, which wouldn't cost them anything, but would look slightly worse in press releases.

Instead they did this, which risks damaging people's motherboards ...

 

[Dribble]

Tbh it looks like me like the board designers thought they were designing a 110W card, hence single 6 pin and limited cooling. Anyway due to either chip or performance problems the card is now pulling 150+W but this happened to late to redesign the power/cooling. Hence I'd avoid the reference cards and wait for the custom cards with custom pcb's, 2*6pin or 1*8pin and better cooling to fix this.

 

[Magic Carpet]

This could only be an issue for miners. Gamers are safe

 

[Abwx]

AMD could have avoided this problem very easily by using an 8-pin connector instead of a 6-pin connector, which wouldn't cost them anything, but would look slightly worse in press releases.

Instead they did this, which risks damaging people's motherboards ...

Dont think that there could be damage at this level, a 2 PCIe MB is rated for 150W total power for the connectors wich are simply connected in parrallel in the supply traces.

 

[amenx]

Dont think that there could be damage at this level, a 2 PCIe MB is rated for 150W total power for the connectors wich are simply connected in parrallel in the supply traces.

Where does that leave CF?

 

[Qwertilot]

If you look at what the Stilt posted here: http://forums.anandtech.com/showthread.php?t=2465645&page=83

It does seem possible to make a proposition that it is quite possibly a 12/3/40 or something W card underneath - enough to at least stay in specification I guess - but they're getting very variable quality chips back from GloFlo and seemingly selling them all anyway.

Worst case seems to be this review the Stilt references:
http://muropaketti.com/artikkelit/testissa-amd-radeon-rx-480-polaris/8/

Only barely operational that one.

Quite why they're doing this I'm not sure. Hopefully some kind of incompetent quality control rather than malice afterthought, not that that inspires either!

Agree that the AIB cards should probably fix a lot of this, presuming that their attrition rate when getting good chips is low enough they can keep costs sane that is.

 

[khon]

Dont think that there could be damage at this level, a 2 PCIe MB is rated for 150W total power for the connectors wich are simply connected in parrallel in the supply traces.

That's true, but a lot of people buying the RX480 will have single PCIe MB's, and the power limit of 75W exists for a reason.

 

[Sweepr]

Has this been posted?

"With Tom's Hardware reporting that the RX 480 draws (substantailly) more than the 75W allowed from the motherboard (for example, the PCI Express high-power card spec allows a mazimum of 66W to be drawn from the 12V pins of the PCI Express slot, and the RX 480 averages79W from the 12V lines alone) AMD seems to be violating the PCI Express(R) spec.

According to the licensing contract for the spec, if they do not fix this within 3 months, AMD will NOT be able to call the card a PCI Express card. If they do, they face not only litigation, but if my understanding is correct an action before the U.S. International Trade Commission (ITC) to ban the importation of the card as counterfeit goods. You might think the PCI-SIG will give AMD a pass, but if they do, they risk loosing the trademark entirely. An unforced trademark gets invalidated. The SIG won't let that happen.

So what does this mean to the consumer? I think there are two possibilities, if we assume AMD will not choose to remove the PCI Express logos from these cards: Either they will alter the boards to have an 8-pin socket and to more power from there, or they will neuter the card to ensure it doesn't draw more power than the PCI Express specification allows. I don't see any other options.

Disclaimer: I am an attorney, but I practice patent law, not trademark law. This post does not constitute legal advice and does not create an attorney-client relationship".

https://hardforum.com/threads/amd-radeon-rx-480-video-card-review-h.1903637/page-3#post-1042386067

 

[Silverforce11]

To get certification, the cards would have to pass testing in the first place. This is what AMD mentioned, they got certified for the RX 480. If the sample you have exceeds the spec, then it's just a matter of RMA.

 

[itsmydamnation]

What determines where power is drawn from? is it just resistance ? (ie is it just physics) if so does that mean MB and PSU manufactures also need their PCI-E certifications .

 

[StrangerGuy]

AMD could have avoided this problem very easily by using an 8-pin connector instead of a 6-pin connector, which wouldn't cost them anything, but would look slightly worse in press releases.

Instead they did this, which risks damaging people's motherboards ...

It's classic AMD really: for the want of a bit of more pre-release hype or cutting corners for pennies, ends up shooting itself massively in the reputation foot several times over. These guys will never learn.

 

[William Gaatjes]

http://www.pcinvasion.com/amd-looking-rx480-pcie-compliance-failure-reports



A user on reddit also posted the GPUz shot of his RX 480 under gaming load and it was running at 1.29vcore, well above the 1.15v max allowed voltages under Wattman.

This looks like a Crimson problem failing to regulate the power and running some cards over-volted.

It could be the calibration went wrong. Or the chip has trouble calibrating itself. Or one of the sensors is faulty.

http://arstechnica.co.uk/gadgets/2016/06/amd-rx-480-polaris-review/

First up is a new power calibration routine that takes place at boot time, which removes the safety limitations associated with variations in consumer power supplies versus those used in manufacturing. By comparing the voltage analysis run during manufacturing with the boot time readings, the voltage regulators on the board can be precisely dialled in to match. This eliminates the waste power associated with accommodating individual motherboards and power supplies, allowing for more stable boost clocks.

I think that is the issue toms hardware has. And why the card at toms hardware draws more power.
Something wrong with the card is very likely.

edit:
Maybe, the remaining oc issues might also be coming from this.
It will be a great card. Teething problems only. :thumbsup:
I am sure of it.

 

[rancherlee]

My thinking is that the Chips varied a larger amount than they expected and they needed to either Bump up the voltage or lower the clocks for more of the chips to be used, obviously they bumped up the voltage so they could use as many chips as possible for the launch of the 480 instead of binning lower quality ones for 470 use. I'm personally waiting for an AIB version then its time to retire my 7950.

 

[William Gaatjes]

If you look at what the Stilt posted here: http://forums.anandtech.com/showthread.php?t=2465645&page=83

It does seem possible to make a proposition that it is quite possibly a 12/3/40 or something W card underneath - enough to at least stay in specification I guess - but they're getting very variable quality chips back from GloFlo and seemingly selling them all anyway.

Worst case seems to be this review the Stilt references:
http://muropaketti.com/artikkelit/testissa-amd-radeon-rx-480-polaris/8/

Only barely operational that one.

Quite why they're doing this I'm not sure. Hopefully some kind of incompetent quality control rather than malice afterthought, not that that inspires either!

Agree that the AIB cards should probably fix a lot of this, presuming that their attrition rate when getting good chips is low enough they can keep costs sane that is.

That is a very informative post by the Stilt. Thank you.
Will there be a freeware program to read out and interpret all the fuses of polaris ?

For those interested, The site is from Finland.

 

[William Gaatjes]

My thinking is that the Chips varied a larger amount than they expected and they needed to either Bump up the voltage or lower the clocks for more of the chips to be used, obviously they bumped up the voltage so they could use as many chips as possible for the launch of the 480 instead of binning lower quality ones for 470 use. I'm personally waiting for an AIB version then its time to retire my 7950.

Yeah, it is strange because the 14nm finfet process GF is using is promoted as giving more along the same result with respect to all produced dies.

https://www.techpowerup.com/reviews/AMD/RX_480/2.html

I was already waiting anyway.
Every manufacturer in the computer business has sometimes early adopter issues (teething problems ). After some time these are resolved and there is a great product to buy.

 

[NomanA]

Has this been posted?


https://hardforum.com/threads/amd-radeon-rx-480-video-card-review-h.1903637/page-3#post-1042386067

Yes, and I think you were the one who posted this forum post from HardOCP before.
I'd consider it a typical over-reaction from a forum user about lawsuits and products getting banned

But, have you seen this ?

https://www.reddit.com/r/Amd/comments/4qfwd4/rx480_fails_pcie_specification/d4sy0c3

1) The RX 480 meets the bar for PCIe compliance testing with PCI-SIG. //edit: and interop with PCI Express. This is not just our internal testing. I think that should be made very clear. Obviously there are a few GPUs exhibiting anomalous behavior, and we've been in touch with these reviewers for a few days to better understand their test configurations to see how this could be possible.
.............................................................................
We will have more on this topic soon as we investigate, but it's worth reminding people that only a very small number of hundreds of RX 480 reviews worldwide encountered this issue. Clearly that makes it aberrant, rather than the rule, and we're working to get that number down to zero.
/edit for absolute factual clarity.

 

[sm625]

Who knows what PCIe compliance testing actually entails though. They could simply plug the card in and check its idle power usage. I imagine that under certain gaming loads, the power draw is lesser than others. So unless they use the most demanding games they would not find a compliance issue.

 

[Beer4Me]

Review samples are likely using special BIOS/firmware that are running it at higher boost clocks during gaming/benchmarks which is overvolting the card. These chips/cards ARE not the glorious overclockers people were hoping for...because they're already overclocked quite a bit out the gates.

I couldn't have said it any better than ^StrangerGuy^ above. AMD(ATi) has put a bunch of egg on their face trying to give the 480's a better showing on the tests/benches, but I guess no one @ AMD realized that the Internet would be judging them on noise/power harshly. After all, noise/power is just as important as performance these days, given how much performance we have. It's a delicate balance ppl.

 

[LTC8K6]

Well, it makes no sense to send any cards to reviewers that might have this problem.

One would think that AMD would check the cards before giving them out to make sure there are no such surprises.

So, I have to conclude that somehow AMD was unaware of the possibility of some cards/chips being out of spec.

 

[Creig]

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, you can supply a lot more amperage through the 6 and 8 pin connectors if you're using HCS or HCS+ pins without any danger of a meltdown.

 

[Zstream]

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, you can supply a lot more amperage through the 6 and 8 pin connectors if you're using HCS or HCS+ pins without any danger of a meltdown.

I've read the PCI spec several times and still haven't found the 75W number. I don't really care either way, but how did that number come about again?

 

[Despoiler]

Review samples are likely using special BIOS/firmware that are running it at higher boost clocks during gaming/benchmarks which is overvolting the card. These chips/cards ARE not the glorious overclockers people were hoping for...because they're already overclocked quite a bit out the gates.

I couldn't have said it any better than ^StrangerGuy^ above. AMD(ATi) has put a bunch of egg on their face trying to give the 480's a better showing on the tests/benches, but I guess no one @ AMD realized that the Internet would be judging them on noise/power harshly. After all, noise/power is just as important as performance these days, given how much performance we have. It's a delicate balance ppl.

Except that the power variation across several reviewers already makes your hypothesis impossible. If AMD loaded special BIOS they would all be more or less consistent. Link to the Reddit megathread from the dude pushing the fails PCI-E spec. Also confirms that some reviewers used retail samples for testing.

https://www.reddit.com/r/Amd/comments/4qfwd4/rx480_fails_pcie_specification/

 

[Stuka87]

Who knows what PCIe compliance testing actually entails though. They could simply plug the card in and check its idle power usage. I imagine that under certain gaming loads, the power draw is lesser than others. So unless they use the most demanding games they would not find a compliance issue.

No, it quite a lot more involved than that. There are very detailed test procedures that must be adhered to. Nothing about standards compliance is simple.