The 480: power consumption, PCI-E powerdraw

[agfkfhahddhdn]

.

 

[Abwx]

We don't know what the component ratings are.

Peak ratings are often quite high for components.

It may be that 200W peaks are okay, but the 70W average isn't, for a given component.

Overall, I would think the average number is the important one.

You are looking things backward, if 200W peak is supported then 70W average should have no problem being supported as well, but the reverse is not true.

Anyway a given average mandate a given peak, for instance 75W average mandate 150W peak being supported.

 

[Shivansps]

There was a reason of why 750TI with 6 pins existed.

 

[Abwx]

It may do so in CF...nothing gona happen with a single card at all.

These are not strange events

Its more common that you may think

This wont happen with such low power card and you know it, these kind of failures are likely due to highly overclocked CPUs but certainly not something that is connected to the PCIe connectors.

There s several wires for the 12V and 0V rails, so there must be several hundreds watts going on here, something like 300W at least....

 

[Shivansps]

IIRC, that second image is from a rig running SLI video cards, and doesn't have anything to do with Crossfire...

Edit: Yep. Here's the original thread from 2010:
http://forums.evga.com/tm.aspx?m=500697&high=raw2dogmeat

So? it does not have to be about CF, its just to ilustrate a point. BTW, SLI is about using more than 1 card as well.

 

[LTC8K6]

You are looking things backward, if 200W peak is supported then 70W average should have no problem being supported as well, but the reverse is not true.

Anyway a given average mandate a given peak, for instance 75W average mandate 150W peak being supported.

There are too many variables to make any blanket statements.

Even the definition of "peak" and "average" will vary quite a bit.

I'm sure we've all seen speakers rated at a very high watt number only to read the fine print and see the very low RMS rating.

 

[Shivansps]

This wont happen with such low power card and you know it, these kind of failures are likely due to highly overclocked CPUs but certainly not something that is connected to the PCIe connectors.

There s several wires for the 12V and 0V rails, so there must be several hundreds watts going on here, something like 300W at least....

CPU does not take power from ATX as far i remember, ATX is used for expansion slots and other mb components.

Again, this is not gona happen on a single RX480, it may happen on CF if the pcie power draw is higher than normal, there is no way around it.

 

[Abwx]

So? it does not have to be about CF, its just to ilustrate a point. BTW, SLI is about using more than 1 card as well.

In CF the GPUs wont be maxed out, besides the 2 x 75W drained by the two PCIe connectors will increase slightly the voltage drop in the MB supply route and in the wires of the ATX connector, hence the 6 pins cables will present a significantly better conductance and thus drain most of the current, likely more than with a single card configuration as the ones reviewed.

 

[Erenhardt]

It was not current that did the damage. It was human error. Make sure your connectors are plugged firmly.

 

[LTC8K6]

There was a reason of why 750TI with 6 pins existed.

Well, the question is not whether you have a power connector, but how much power is pulled from the PCI-E slot, even if you do have a power connector.

As long as the PCI-E slot power stays within spec, and the overage is pulled from the power connector, it's fine.

For example, a reddit post:

Update 12

Purchased a Sapphire 8GB RX 480 today. After reading up about this issue, I decided to test for myself. I rigged up a riser to be able to measure 12V current with an AMP clamp from both the PCI-e slot, and 6 pin connector.

This isn't anywhere near being scientific, but I think it's accurate enough to confirm the problem. Running stock clocks with stock voltage while running ethereum mining = 83w from the 6 pin connector, and 88w from the PCI-e slot.

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

 

[Abwx]

CPU does not take power from ATX as far i remember, ATX is used for expansion slots and other mb components.

Again, this is not gona happen on a single RX480, it may happen on CF if the pcie power draw is higher than normal, there is no way around it.

The case you posted has nothing to do with CPU or whatever other component, it s a PSU failure apparently, so for the time you re just using some technical straw that is 100% irrelevant to the case discussed here..

GTX750 peaks at 140W and is most of the time well above the 75W line, so i dont see where would be the problem here, unless what is considered normal for one brand is not for another one...

 

[MrTeal]

This wont happen with such low power card and you know it, these kind of failures are likely due to highly overclocked CPUs but certainly not something that is connected to the PCIe connectors.

There s several wires for the 12V and 0V rails, so there must be several hundreds watts going on here, something like 300W at least....

The CPU gets its power from the EPS plug, not the ATX 24-pin connector. There's a lot of grounds on the 24 pin connector, but only two 12V ones. Up until they moved from 20pin to 24pin, there was only one 12V pin.

The Minifit Jr pins ratings vary based on surface finishes and the number of connections, but generally they're in the 7-11A range. Most are 8A or so with 16 or 18 gauge wiring, which gives a max rated input power of 200W or so. That's for a properly constructed connector in good shape though. If you have a female pin connector that's a little stretched or damaged, with only two pins it's not that hard to overload the single remaining 12V wire. I'm actually surprised you don't here more reports of ATX connectors melting with how much draw is on the 12V rail in modern computers.

 

[MrTeal]

In CF the GPUs wont be maxed out, besides the 2 x 75W drained by the two PCIe connectors will increase slightly the voltage drop in the MB supply route and in the wires of the ATX connector, hence the 6 pins cables will present a significantly better conductance and thus drain most of the current, likely more than with a single card configuration as the ones reviewed.

In most GPUs I've investigated for this, there is no internal combined 12V plane. In most high power GPUs the 12V feed for the core power comes from the PCI-e power plugs, while some other smaller 12V loads (like the fan) are run off the slot. If Tom's is to be believed on the RX480 three of the core phases are fed by the PCI-e power plug, and three are fed by the slot. The card will pull pretty much equal power through both paths even if the droop in the socket is higher.

It should make for some interesting poking around when I get my card next week.

 

[sm625]

Y
AFAIK, on pretty much any motherboard I've seen the PCI-E 12V power is just directly connected to the MB 12V plane, so it's not like you have a dedicated supply feeding the card that could by overloaded. The bigger danger (IMO) with high draw cards like this is the possibility of overloading your MB 24 pin connector which only has two 12V wires coming into it if you're running multi-GPU.

A motherboard also has a pair of dedicated 4 pin 12V headers, half of which are 12V and half are ground. I'm like 98% sure that these 12V lines are tied to the same internal power plane as the two 12V pins on the 24 pin ATX header. So that gives a total of six +12V connections from the PSU to the motherboard.

 

[William Gaatjes]

It may do so in CF...nothing gona happen with a single card at all.

These are not strange events

Its more common that you may think

Current spikes of only milliseconds long would not damage those female terminals. These terminals are rated a 9A (20awg) continuous current each. That would be (12*2*9) 216 watts that can be drawn from the atx connector (Not taking into account the current the pci-e connector can handle or the copper plane in the MB pcb).
Note that the current the terminal can handle is limited by the used wire awg. At 16awg, the pin can handle 12,5A. It is not uncommon that sometimes a female terminal is misfolded during manufacturing and when connected in an MB atx connector, has a higher contact resistance with the male terminal than when the pin is correctly folded. Or that the wire has a bad crimped contact with the female terminal. Billions of these terminals have been produced. It is not uncommon that something like this happens. It is a ppm failure.


Three possibilities (or a combination of):
* That what ever was connected was constantly drawing a very high current, outside the spec of the terminals.
* The terminals were misfolded.
* Or the wires were not crimped correctly to the female terminals.

 

[MrTeal]

A motherboard also has a pair of dedicated 4 pin 12V headers, half of which are 12V and half are ground. I'm like 98% sure that these 12V lines are tied to the same internal power plane as the two 12V pins on the 24 pin ATX header. So that gives a total of six +12V connections from the PSU to the motherboard.

Are you talking about the 4 or 8 pin EPS plug? Grab a multimeter and test it some time, the 12V ATX pins are not in common with the EPS 12V plane. The EPS connector just feeds the CPU, that's why some sites use it to measure CPU power draw.

Now the motherboards that include a separate Molex connector near the PCIe slots to help power them are in parallel with the ATX connector, but that's to be expected.

 

[rgallant]

IIRC, that second image is from a rig running SLI video cards, and doesn't have anything to do with Crossfire...

Edit: Yep. Here's the original thread from 2010:
http://forums.evga.com/tm.aspx?m=500697&high=raw2dogmeat

just saying , most burn outs were mainly caused by loose/bad fitting plugs

with my sli gtx580's I used a power boost from evga for my 760 classy evga mb that did not come with a aux. 12v plug.
I plugged it in the booster into the pci-e slot then to a 12v molex to take the load off the main power plug to the mb.

fun fact with the booster
I won one of the free ones , when it came it was something like $3.xx shipping [to Toronto] with 15+ real stamps to make up the shipping on the little package lol

 

[William Gaatjes]

In CF the GPUs wont be maxed out, besides the 2 x 75W drained by the two PCIe connectors will increase slightly the voltage drop in the MB supply route and in the wires of the ATX connector, hence the 6 pins cables will present a significantly better conductance and thus drain most of the current, likely more than with a single card configuration as the ones reviewed.

I do not know exactly how all these 12V inputs are connected, i assume directly with each other on the gpu pcb. So it is true, that through the path with least resistance the most current will flow.

As a side note :
smps step down buck converters do have one counter intuitive effect. An smps is a power converter. This means that the smps to reach the target load (output) voltage and current, will actually increase the input current intake when the input voltage lowers. Lower input voltage means more input current for the same output load. But, the 12v line will have to dip seriously before that becomes significant.

 

[William Gaatjes]

Are you talking about the 4 or 8 pin EPS plug? Grab a multimeter and test it some time, the 12V ATX pins are not in common with the EPS 12V plane. The EPS connector just feeds the CPU, that's why some sites use it to measure CPU power draw.[/B]

Now the motherboards that include a separate Molex connector near the PCIe slots to help power them are in parallel with the ATX connector, but that's to be expected.

This indeed can be the case with the better psu models. This is because the 12V rails are each from a different secondary winding from the smps transformer in the PSU. The secondary windings of the transformers each have their own rectifying diode and capacitor. It is a design choice to limit the stress on the components.

 

[MrTeal]

This indeed can be the case with the better psu models. This is because the 12V rails are each from a different secondary winding from the smps transformer in the PSU. The secondary windings of the transformers each have their own rectifying diode and capacitor. It is a design choice to limit the stress on the components.

I don't mean with the PSU attached. Most power supplies would measure a short if you measure it with the connectors still plugged in.

 

[William Gaatjes]

I don't mean with the PSU attached. Most power supplies would measure a short if you measure it with the connectors still plugged in.

I thought you meant measuring the 12V pins from the PSU separately of course.
On the MB, the local 12V rails should not be connected.

 

[tonyfreak215]

I don't mean with the PSU attached. Most power supplies would measure a short if you measure it with the connectors still plugged in.

You could back probe it to get voltage. You mist be in series to get amperage (or use an amp clamp)

 

[Abwx]

I do not know exactly how all these 12V inputs are connected, i assume directly with each other on the gpu pcb. So it is true, that through the path with least resistance the most current will flow.

The 12V rails are connected together at the card level and in the PSU as well, the line of lower resistance will force a higher current path, in principle the 6pin cable has the lower voltage drop since it supply only the GPU and that the ATX + MB routing has significantly higher resistance, the more other components drain power the more the GPU power will be dispatched to the 6 pin connector.

As a side note :
smps step down buck cvoltage droponverters do have one counter intuitive effect. An smps is a power converter. This means that the smps to reach the target load (output) voltage and current, will actually increase the input current intake when the input voltage lowers. Lower input voltage means more input current for the same output load. But, the 12v line will have to dip seriously before that becomes significant.

That s accurate, there s two dynamic systems in serial, higher peak power can be due to a more beafy card or PC PSU that take less time to compensate the reservoir capacitor s losses of charge..

Anyway i find those power matters blown out of proportion in the case that interest us, moreover given that previous competing products did display the same behaviour, sometimes even more blatantly, without anybody only noticing before Polaris was released...

 

[Stuka87]

It may do so in CF...nothing gona happen with a single card at all.

These are not strange events

Its more common that you may think

*SIGH*

Those four lines are rated for 192W sustained (16amps @12VDC).

The photos you show cannot be caused by a single card using 10W over the rating, or even two cards that are a combined 20W over the rating.

Just because you can find photos of burned up connectors (with zero background on to what caused them) doesn't give your "theory" merit. If there was a weak link in the whole power delivery system to a GPU, its on the pins in the PCI-E connector.

You can keep trying to blow this out of proportion, but its not going to make what you are saying suddenly become truth.

 

[Doom2pro]

Three possibilities (or a combination of):
* That what ever was connected was constantly drawing a very high current, outside the spec of the terminals.
* The terminals were misfolded.
* Or the wires were not crimped correctly to the female terminals.

I agree, I had an old GA-7VT600 1394 Socket A (Athlon XP) motherboard that fed off the 5V rail (No Intel style 12V VRM design) and the power supply and or the socket on the motherboard (or the combination of the two) caused bad connections between the two, this caused increased resistance and thus voltage drop, coupled with high currents (albeit normal for the socket) this caused browning and deformation and in some users boards, melted and fused together!