Tom's Hardware re-examines the RX 480's power consumption

[Sintezza]

Like i said before, Toms has simply fucked their shit up pretty badly, as usual.

[KeltonDSMer]

11 hours ago, Supermangik said:

What are you saying, this is okay? First off, it isn't. AMD need to get their head out their _))) thinking they can fool the public by pulling something like this. Having the card draw more power out of the PCI-e slot is not okay. It may not cause harm for more updated mobo's, but people who buy this card get it for the very reason of budget cuts because that's exactly what it's suppose to be, a budget mid-gaming card. If you want to push that RX 480 and have a constant gaming experience for 2+ more years, good luck.

 

I can smell dead PCI-e slots if you choose to seriously game with that crap card after a certain amount of months to years. I'm glad Toms Hardware brought up this subject to the public because Linus didn't do it, and not many other Youtuber's would even do it. Because Tom's made it public, they forced AMD to correct their mistake, and put them in their place if they ever tried to pull another stunt like this.

 

The motherboard manufacturer PCPER spoke with claimed the point of failure for overdrawing from the PCIe slot is the connector itself and not the power plane in the MOBO.

 

A person claiming to be an EE on reddit did some digging and posted his/her findings. It is shown that each 12v connection at 65C (beyond worst case scenario) can sustain 2.8amps (there are 5)...that is almost 170W continuous power without fault. This was for the cheapest slot manufacturer he could find.

 

Well written post with sources: https://www.reddit.com/r/Amd/comments/4rbw8p/facts_about_pcie_connectors/

 

The expense of a motherboard has almost no bearing on whether or not it is "safe" to use a reference rx480 with it. If the contacts are clean and the card is seated properly, any board with a x16 slot should be fine with rx 480.

 

Edit: It seems more than one lower end mobo has failed with a reference RX 480 installed.

 

[Joe_Easy]

Tom's hardware and all the other media wannabees have lost a lot of respect. In their desire to be the first with a scoop and a negative story, they ignore the facts and rushed to publish nonsense without checking with AMD and experts. I can't say I'm surprised but with more and more options on reviewers, I make my choice based on people who show some integrity and value quality. 

 

The reality is that a lot of AMD folks will have invested a lot of time and energy to bring something strong to the market, and I believe they have, that competes with Nvidia and makes gaming/pc building more accessible to everyone and TH and others shit all over that. If they had taken the time to check their facts and validate their results, then fine - massive kudos but this smells like a hack job.

 

And no, I'm not an AMD fan boy..I have an EVGA 980ti in my rig at the minute.

[LAwLz]

Oh boy, here comes all the AMD apologists which still don't understand how to read properly...

 

• No, Tom's Hardware did not "fuck their shit up". They did not "ignore the facts". All they are doing is telling people "look, you are not reading our graphs correctly, and you are putting words in our mouths".
   
• When they re-examined their RX 480 it still draws way too much power though the slot (we are not talking about a tiny amount too much, we are talking like 20% more than allowed).
   
• If you buy a brand new, mid-range or high end motherboard then it will most likely not destroy other hardware. At worst you will probably just get some audio interference (especially on lower end motherboards). If you use old hardware then the risk of failure is higher.
   
• Tom's Hardware added a low-pass filter which is why their graphs will now look different. The actual results has NOT changed. Only the way they represent the data. This makes it easier for people who don't understand what is actually going on in the graphs read them (looking judgingly at everyone who said the 960 and 750 Ti had the same issues). They will try and do this for all future tests.
   
• This is still a problem that needs fixing.

[Joe_Easy]

13 minutes ago, LAwLz said:

Oh boy, here comes all the AMD apologists which still don't understand how to read properly...

 

• No, Tom's Hardware did not "fuck their shit up". They did not "ignore the facts". All they are doing is telling people "look, you are not reading our graphs correctly, and you are putting words in our mouths".
   
• When they re-examined their RX 480 it still draws way too much power though the slot (we are not talking about a tiny amount too much, we are talking like 20% more than allowed).
   
• If you buy a brand new, mid-range or high end motherboard then it will most likely not destroy other hardware. At worst you will probably just get some audio interference (especially on lower end motherboards). If you use old hardware then the risk of failure is higher.
   
• Tom's Hardware added a low-pass filter which is why their graphs will now look different. The actual results has NOT changed. Only the way they represent the data. This makes it easier for people who don't understand what is actually going on in the graphs read them (looking judgingly at everyone who said the 960 and 750 Ti had the same issues). They will try and do this for all future tests.
   
• This is still a problem that needs fixing.

 

No need to call anyone an AMD apologist and Tom's hardware don't need anyone apologising for them either.

 

Show me an example of where this actually impact a MB with proven fact. If you can't do that, then show me the 10 or 20 customers who have fried their MB since buying a RX480?

 

Rather than use terms like "too much" power and "old hardware risks failure", how about you provide some data? Old hardware increases the risk of fault for any number of reasons. Too much power - let all the people who certify PCIE and engineer MB and video cards for a living be wrong, let the people who sell PC components for a living also be wrong...based on the interpretations of some one who couldn't even be arsed to check their facts with AMD or board engineers.

 

If the best argument that people can share is look - we cleaned up the graphs so it's easier for you simple folks to see how right we were, then it's not really an argument at all.

 

Sure are there some improvements to this card that AMD can make with drivers, absolutely...but so does every card and technology device at launch.

 

Smells like a hack job to me and it's a disgrace to be honest. Can't believe how quick people are to get the pitchforks and rush to judge. 

 

[Deli]

7 hours ago, ivan134 said:

Obligatory, no shit. The people who claimed their motherboards were damaged were clearly trolls and yet people still quote them as evidence that the 480 is dangerous for your mobo.

I bet this myth will never die, and repeated again and again by AMD haters.

[79wjd]

@Joe_Easy No one is claiming that a 480 will trash your board instantly. The concern is the long term effects it may have. Standards exist for a reason and going 20% over spec doesn't sound too good if you've ever seen some of the trash quality boards that OEMs use. 

[Yamoto42]

It may well be within tolerance, but that doesn't mean it is within specification.  Those are two totally different things.

 

Ok, so current motherboards are fine.  So they draw 20% too much power.  As long as nothing is damaged it's fine, right?

 

So what about when the next card draws 25%?  Say 1% of motherboards paired with one are truly are damaged by it, is that OK?  1% is within a margin of error.


Tolerance is what actually happens.  Specification is purely the principle of the matter.

We can throw principle out the window, but unfortunately all engineering fields are based around designing tolerances around specifications.  As soon as you allow those to no longer matter, your tolerances are nothing more than pure guesswork.

[Joe_Easy]

15 minutes ago, Yamoto42 said:

It may well be within tolerance, but that doesn't mean it is within specification.  Those are two totally different things.

 

Ok, so current motherboards are fine.  So they draw 20% too much power.  As long as nothing is damaged it's fine, right?

 

So what about when the next card draws 25%?  Say 1% of motherboards paired with one are truly are damaged by it, is that OK?  1% is within a margin of error.


Tolerance is what actually happens.  Specification is purely the principle of the matter.

We can throw principle out the window, but unfortunately all engineering fields are based around designing tolerances around specifications.  As soon as you allow those to no longer matter, your tolerances are nothing more than pure guesswork.

 

Good point and well made. It's important to remember though that these boards have redundant fail safes and most boards will turn off before it gets so bad there is permanent damage. I'm not saying AMD shouldn't put out a fix and investigate what happened but the rush to crucify AMD is over done, mostly by people reacting to hysteria.

 

I mean if you look at how much people have been overvolting and overclocking CPU's, GPU's, RAM...there is rarely an issue. Doesn't mean it never fails...I recently had some DDR3 memory fail...took me 48 hours of swapping and trouble shooting the issue...drove me bonkers and couldn't figure it out...sent it back to the place I bought it from and got a new set a few days later.

I bought the Sapphire 480X Nitro as I tend to buy aftermarket cards when ever I can so looking forward to seeing what a 480X can do with a great power delivery, AIB design and 8 pin plug when I get it.

[xAcid9]

10 hours ago, zMeul said:

so get this:

-snip-

AMD specifically engineered the board to draw more power from PEG - they did it to mask the fact the card draws way more than 150W

They updated the picture. 

 

Similar to what buildzoid found in his 2nd video.

 

 

 

 

9 hours ago, 7850OC said:

i trust more on "PCper" now than on "toms hardware"

 

(i don't care if they apologize now with us for the mislead) 

-snip-

 

9 hours ago, Sintezza said:

Like i said before, Toms has simply fucked their shit up pretty badly, as usual.

Whats wrong with their first graphs?

[zMeul]

15 minutes ago, xAcid9 said:

They updated the picture. 

 

Similar to what buildzoid found in his 2nd video.

still, they have more power phases from PEG than the 6pin - AMD totally knew the fuck they're doing and hoped no one would notice

Edited July 6, 2016 by zMeul

[xAcid9]

1 minute ago, zMeul said:

still, they have more power phases from PEG than the 6pin - AMD totally knew the fuck they're doing and hoped no one would notice

No reason for them to do that. I think it just poor quality control.

[zMeul]

9 minutes ago, xAcid9 said:

No reason for them to do that. I think it just poor quality control.

poor quality control!?!? mate ....

someone(s) had to design that shit and had to be approved, then it went into manufacturing - why is it smelling like #dieselgate all over again!??!

[That Norwegian Guy]

Remind me why Tom's saw it as necessary to "update" on this issue mere hours before AMD's new driver?

 

What an excercise in pointlessness.

 

Though not in clicklessness. Gotta double-dip!

[Misanthrope]

9 hours ago, KeltonDSMer said:

 

The motherboard manufacturer PCPER spoke with claimed the point of failure for overdrawing from the PCIe slot is the connector itself and not the power plane in the MOBO.

 

A person claiming to be an EE on reddit did some digging and posted his/her findings. It is shown that each 12v connection at 65C (beyond worst case scenario) can sustain 2.8amps (there are 5)...that is almost 170W continuous power without fault. This was for the cheapest slot manufacturer he could find.

 

Well written post with sources: https://www.reddit.com/r/Amd/comments/4rbw8p/facts_about_pcie_connectors/

 

The expense of a motherboard has almost no bearing on whether or not it is "safe" to use a reference rx480 with it. If the contacts are clean and the card is seated properly, any board with a x16 slot should be fine with rx 480.

 

This guy managed to isolate a board turning off the system due to the 480. It's a fairly old board but the important part is that a 980ti was stable and the 480 drawing significantly less power overall shut down the system:

 

 

[NumLock21]

Old thg used to be awesome. Burning cpus on purpose, while playing techno music.

Now its fill with idiotic comments and psu recommendation at least a 80plus platinum.  And all aio are no good except for dual rads.

[KeltonDSMer]

47 minutes ago, Misanthrope said:

This guy managed to isolate a board turning off the system due to the 480. It's a fairly old board but the important part is that a 980ti was stable and the 480 drawing significantly less power overall shut down the system:

 

Yeah this board failed too, at the slot and it happened to be an old oem board... https://www.reddit.com/r/pcmasterrace/comments/4rfou0/brand_new_rx_480_just_destroyed_my_motherboard/

[NuclearPeace]

Regardless if it has the potential to damage motherboards or not its still exceptionally poor design from AMD. Stuff like this makes it hard for me to really get behind AMD as the underdog when they shoot themselves in the foot constantly. If this card was designed properly then there wouldn't have been any discussion about drawing too much current through the PCI-E slot.

 

Firstly, AMD should have stopped being so adamant about only having a 6 pin auxiliary plug. With a TDP of "150w", nobody is going to be using this on a 250-350w crappy no 80+ certified PSU as part of a prebuilt upgrade. 99% of RX 480s are going to be used by system builders who have quality PSUs equipped with 2x 6-8 pin power outputs or more. The 970, which uses the same amount of power, more or less, comes with a 8 pin AND a 6 pin on the reference board. That's 225w of in spec power just from the PSU which allows the card tons of headroom to operate safely. Leaving the RX 480 so little headroom with just a 6 pin isn't going to fool or impress anyone.

 

Secondly, why the hell would you wire half of the power phases up to the +12V PCI-E slot? In the event that the card needs to draw more power, the 6 pin is the safest way to do it due to the lower resistivity (and thus less  ohmic heating) of the 6 pin cables rather than the ultra thin PCI-E connectors. Not only that, but the PCI-E power sources its energy from the 24-pin motherboard power which also needs to power things such as the southbridge, I/O, and audio. Pulling too much power via drawing too much current through that 24-pin motherboard power slot exponentially increases resistance, which is also going to greatly increase the chance of failure via melting or even a fire.

 

I'm glad that add in board partners are fixing the lunacy of the reference design by adding things such as an 8 pin and a DVI connector.

[xAcid9]

1 hour ago, zMeul said:

poor quality control!?!? mate ....

someone(s) had to design that shit and had to be approved, then it went into manufacturing - why is it smelling like #dieselgate all over again!??!

Someone that created the BIOS for 480 probably half asleep/drunk since this can be fix via vBIOS.

 

http://www.overclock.net/t/1604979/a-temporary-fix-for-the-excess-pci-e-slot-power-draw-for-the-reference-rx-480-cards/0_20

[zMeul]

1 minute ago, xAcid9 said:

Someone that created the BIOS for 480 probably half asleep/drunk since this can be fix via vBIOS.

 

http://www.overclock.net/t/1604979/a-temporary-fix-for-the-excess-pci-e-slot-power-draw-for-the-reference-rx-480-cards/0_20

yeah sure ... BIOS will move the traces on the PCB too - don't kid yourself

the most thing they can do is to limit the power that goes into the 4 phases linking to the PEG and increase the power that goes trough the rest

[xAcid9]

6 minutes ago, zMeul said:

yeah sure ... BIOS will move the traces on the PCB too - don't kid yourself

the most thing they can do is to limit the power that goes into the 4 phases linking to the PEG and increase the power that goes trough the rest

3 phases.. and i don't see a problem with that since the VRMs is beefy.

[zMeul]

2 minutes ago, xAcid9 said:

3 phases.. and i don't see a problem with that since the VRMs is beefy.

the issue here is that AMD made changes to the power delivery system to make the card look like it's drawing 150W - by doing that they failed to meet PCI-SIG specs for PEG

by increasing the power draw from the PCIe 6pin they will also fail PCI-SIG specs

 

no matter how they turn it, the RX480 is not PCIe Express compliant as per PCI-SIG specs - and it's not the 1st time this happened with AMD

the bigger issue here is that AMD claimed they sent the card for certification - if PCI-SIG actually tested the power draw and still certified it .... then PCI-SIG is fucking worthless and also makes their certifications worthless

[airdeano]

question? i saw they were routing a 3.3 reference voltage through the PEG connector.

how are they passing 3.3v through 3 +12v and 3 ground circuits?

 

 

[zMeul]

2 minutes ago, airdeano said:

question? i saw they were routing a 3.3 reference voltage through the PEG connector.

how are they passing 3.3v through 3 +12v and 3 ground circuits?

they aren't routing anything

3.3V will always be from the PEG connector

[airdeano]

your first graphic showed the 3.3 passing through the PCIe slot connector, then a follow up shows that same 3.3 powered through the PEG (PSU lead) ("PEG" stands for PCI Express Graphics) you cannot power 3.3v signal with 3x 12volt leads.