The 480: power consumption, PCI-E powerdraw

[Cookie Monster]

You make it very hard to believe you are electrical engineer.

First, good luck burning connector due to excessive current. First think that would go pop would be traces or power regulators.

Short spikes is what kills connectors/pins.

The reason for burned connectors is most of the time human error, or sometimes damaged connector. To burn the pins like shown in the photos above, you don't have to draw any excessive currents beyond specifications.

Bad contact between two pins and even light load will make it burn. Add current spikes to that, and you have a recipe for a disaster.

I've seen a 10m cable 1.5mm2 @220V go pop at the connector with 1<kW load on them. Bad contact, sparks, smoke, tears.

On the other hand I've used 40meters long 1.5mm2 @220V with welder plugged in 2kW+. The welder had problems because of Vdrop, cable got pretty warm and the 16Amp fuse went off from time to time. But in this example the cable had no connectors other than your regular socket/plug.

Trust me

Just check the internet or even books if your old school if you don't believe me. The excessive heat (often caused by excessive continous current -> worst case shorts) is the reason for most connector/cable failures. Its the reason why they get hot sometimes if the cable is not able to handle the current draw. Remember, P=I*R (current x resistance).

Those photos you saw is exactly what high current does to connectors/pins. Like you mentioned it yourself. 1~2KW. At what voltage? now work out the current. Work out the resistance of those connectors/pins. Easy to calculate the power dissipated on those pins. Now just look at the amount of copper that exists on a PCI-e connector. Its not really designed for very high current >10A.

 

[Erenhardt]

At least in civil engineering the connector is overspeced compared to the line and breakers.
I think in electronics it is exactly the same.

Anyway. I'm curious if they can change the VRM controller load balancing with software.

 

[zinfamous]

redzo thank you for the posted video.

For all of the posters claiming that the Strix 960 does the same, per the video it clearly does not exhibit the same problem.

Until an AMD official fix, I would undervolt a RX480 a bit.

Gosh, I have used AMD since the 386-40 days but to cry that this is made up by Nvidia fan boys is really putting your head in the sand.

The good news is that AMD has acknowledged the problem and is actively working on the problem.:thumbsup:

no one is saying that this is made up by fan boys of team N. The (probably valid) criticism is that these exact problems when occurring on team N in the past or whenever are largely dismissed. Not that fans of team A are innocent of dismissing similar issues plaguing their parts when criticizing team N.

 

[Erenhardt]

Will AMD/AIBs take some responsability here?

Nope.
Does it matter that he plugged 50% more cards than the board supports?

 

[Zstream]

https://bitcointalk.org/index.php?topic=1433925.msg15438155#msg15438155

it's ok, guys

just a minor setback. it's a 200$ card, after all. can't expect miracles for that price.

He was using unpowered ribbon risers. A dumbass move if you ask me.

 

[William Gaatjes]

Trust me

Just check the internet or even books if your old school if you don't believe me. The excessive heat (often caused by excessive continous current -> worst case shorts) is the reason for most connector/cable failures. Its the reason why they get hot sometimes if the cable is not able to handle the current draw. Remember, P=I*R (current x resistance).

Those photos you saw is exactly what high current does to connectors/pins. Like you mentioned it yourself. 1~2KW. At what voltage? now work out the current. Work out the resistance of those connectors/pins. Easy to calculate the power dissipated on those pins. Now just look at the amount of copper that exists on a PCI-e connector. Its not really designed for very high current >10A.

Know you Ohms law

P= I^2*R = I*I*R.
U = I * R.
P = U * I.
P = U^2/R = U*U/R.

 

[MrTeal]

At least in civil engineering the connector is overspeced compared to the line and breakers.
I think in electronics it is exactly the same.

Anyway. I'm curious if they can change the VRM controller load balancing with software.

I don't have a datasheet for the IR3567B, but it it's anything like the IR3566 I've used before you can't.

Overspeccing connectors becomes a little more challenging/expensive when you have multiple connectors feeding off a line that can supply a lot more current.
You spec the connector to handle sufficient current for the ranges of devices that will be attached to it, not necessarily the power capability of the system upstream of it.

 

[MrTeal]

He was using unpowered ribbon risers. A dumbass move if you ask me.

Definitely, and I mentioned this in the Ethereum thread. You need powered risers in this case. However this failure wouldn't be caused by the risers per se. You can start to see issues with this even if you had three cards plugged directly into the slots.

 

[William Gaatjes]

At least in civil engineering the connector is overspeced compared to the line and breakers.
I think in electronics it is exactly the same.

Anyway. I'm curious if they can change the VRM controller load balancing with software.

Until we have a schematic, we cannot know for sure.

But in this scenario :
If the +12V pins from the pcie card connector and the +12V pins from the pcie power connector are connected to the same internal 12V plane of the graphics card, there is nothing software can do. Then it is hardwired together. But only someone with a RX480 card could test this with a multimeter. 0 ohm connection is tied together. If that is the case, a special cable from the psu to the pcie power connector with high diameter wire(low resistance could balance out the current draw).

 

[MrTeal]

Until we have a schematic, we cannot know for sure.

But in this scenario :
If the +12V pins from the pcie card connector and the +12V pins from the pcie power connector are connected to the same internal 12V plane of the graphics card, there is nothing software can do. Then it is hardwired together. But only someone with a RX480 card could test this with a multimeter. 0 ohm connection is tied together. If that is the case, a special cable from the psu to the pcie power connector with high diameter wire(low resistance could balance out the current draw).

I'll let you know when I get mine, but from what Tom's said the memory and three core phases are tied to the slot, and the other three core phases are tied to the PCIe power plug.

 

[William Gaatjes]

I'll let you know when I get mine, but from what Tom's said the memory and three core phases are tied to the slot, and the other three core phases are tied to the PCIe power plug.

Thank you.

(reading further)
AI !

But i understand it. 150W design. 3 phases to one 75W connector. 3 phases to the other 75(66W ?) connector.
But there are 2 more smps on the board. Add that the polaris 10 should be 110W max, 40watt left for the other 2 smps, powering memory and other logic. And of course the smps switching losses. The design is sound.

And the six pin connector gives the idea to the oem and customer that it is low power. Because no low power board would use 8 pin connector ?
I can understand the reasoning.

 

[jpiniero]

Seems to me that AMD is pushing it too hard and 6-pin isn't enough. I really think if you were interested in the 480 you should wait until you can get your hands on an custom model with 8-pin.

Especially since it looks like cheaper motherboards won't be able to handle it.

 

[William Gaatjes]

Seems to me that AMD is pushing it too hard and 6-pin isn't enough. I really think if you were interested in the 480 you should wait until you can get your hands on an custom model with 8-pin.

Especially since it looks like cheaper motherboards won't be able to handle it.

That is the idea i had too . The 6 phases on the 8pin pcie power connector and the other smps from the pcie card slot. But i wonder if that is feasible. It could be that some serious damaging issues would arise for the polaris 10 chip when the pcie power connector is not present and the pc is powered up anyway with the card plugged in, in this hypothetical scenario.

 

[Erenhardt]

Thank you.

(reading further)
AI !

But i understand it. 150W design. 3 phases to one 75W connector. 3 phases to the other 75(66W ?) connector.
But there are 2 more smps on the board. Add that the polaris 10 should be 110W max, 40watt left for the other 2 smps, powering memory and other logic. And of course the smps switching losses. The design is sound.

And the six pin connector gives the idea to the oem and customer that it is low power. Because no low power board would use 8 pin connector ?
I can understand the reasoning.

Memory has its own 1 phase VRM.

 

[Erenhardt]

That is the idea i had too . The 6 phases on the 8pin pcie power connector and the other smps from the pcie card slot. But i wonder if that is feasible. It could be that some serious damaging issues would arise for the polaris 10 chip when the pcie power connector is not present and the pc is powered up anyway with the card plugged in, in this hypothetical scenario.

I remember booting up with some power connector unplugged. I don't remember if that was with 8800 gts, or some other card. You could boot without all of the PCI power connectors plugged Card artifacted under load, but it worked.

Nowadys if you don't have plugged power connectors, gpu will not start at all.

 

[William Gaatjes]

Memory has its own 1 phase VRM.
Could be that there are unused phases on the board?

No, the advantages of 6 phases is that the smps controller has more regulating control over the output voltage. Also, in all the pictures of the pcb there a 6 identical phases and one aux phase on one side. And on the other side there is another smps. I think it is because i see the same inductor, capacitor and mosfets combination.

 

[William Gaatjes]

I remember booting up with some power connector unplugged. I don't remember if that was with 8800 gts, or some other card. You could boot without all of the PCI power connectors plugged Card artifacted under load, but it worked.

Nowadys if you don't have plugged power connectors, gpu will not start at all.

I can remember reading once that the driver of some card would notify the user that the pcie power connector is not present. I do not know the details anymore.

 

[William Gaatjes]

Wccftech has an article about this dilemma :

http://wccftech.com/amd-rx-480-pcie-power-issue-detailed-overclocking-investigated/

 

[Cookie Monster]

Know you Ohms law

P= I^2*R = I*I*R.
U = I * R.
P = U * I.
P = U^2/R = U*U/R.

Yes but the voltage is what? 12V/5V/3.3V? unless your talking hundreds of volts, the PC in general deal with very low voltage things. Like video cards for one, have hundreds of Amps flowing @ ~1V. Trust me when I say current is the thing that is causing those burned marks.

Unless the application at hand is in the hundreds or thousands of volts it doesn't really apply to PCs because the highest voltage one would see is 12V.

 

[William Gaatjes]

Yes but the voltage is what? 12V/5V/3.3V? unless your talking hundreds of volts, the PC in general deal with very low voltage things. Like video cards for one, have hundreds of Amps flowing @ ~1V. Trust me when I say current is the thing that is causing those burned marks.

Unless the application at hand is in the hundreds or thousands of volts it doesn't really apply to PCs because the highest voltage one would see is 12V.

I was pointing out the error you made in your post.

 

[Cookie Monster]

I was pointing out the error you made in your post.

Oh yeah, forgot the I^2 bit. Haha thanks for the correction. I should retire now.

 

[Sonikku]

Here's an interesting video:

https://www.youtube.com/watch?v=rhjC_8ai7QA

Also, user issue over on Reddit. Who knows if true?

https://community.amd.com/thread/202410

We're entering an age where wars are fought between competitors with information and misinformation. Nvidia was proven so much more savvy with it than AMD I'm almost not sure of who to believe on the internet any more.

 

[Abwx]

I'll let you know when I get mine, but from what Tom's said the memory and three core phases are tied to the slot, and the other three core phases are tied to the PCIe power plug.

That s impossible, if that was the case the 6 pin would provide less than half the total power and could never go above half the GPU power, wich is obviously not the case according to THG s own numbers...

 

[torlen11cc]

They will lower the clocks or voltage and their is no other logically way.

 

[Yakk]

We're entering an age where wars are fought between competitors with information and misinformation. Nvidia was proven so much more savvy with it than AMD I'm almost not sure of who to believe on the internet any more.

We've been in the misinformation age for many, many years now. Misinformation has, and continues to be one of the biggest tools in the marketing warchest. Google, forum shills, and the "tech" site circuit of unconfirmed news just amplify it exponentially.