How important are the power recommendations for graphics cards

I recently upgraded to a new graphics card and due to some limitations, I had to install it in a questionable configuration. My old card had one 6-pin power connection, the new card has two 6-pin power connections. There is only one 6-pin power cable coming from my power supply, but it has two connection ports in series on that line. After installing I read the manual to make sure I did it right and I saw a warning saying that I should use two separate cables to power my graphics card. However, everything seems to run fine (I'm getting good performance, temperature isn't being crazy, etc).

What are the potential risks of using an improper power configuration for my graphics card?

System info:

  • Windows 7 64-bit
  • First generation i7 860
  • ATI R9 270X graphics card

Thanks in advance

Here's a picture of the label on my power supply:

Power supply label


Solution 1:

Judging based on your specific situation, I think you should be fine. If your PSU has two 6-pin PCIE connectors in serial coming out of the same connector on your PSU (I'm guessing the PSU is modular, or else you wouldn't be able to see the number of pins on the PSU side), that just means the PSU is specifically designed to handle this load. It very likely has a single +12V "rail" with an extremely high amperage rating, or two very high amperage rails, but not 3+. PSUs that have more rails have less amperage per-rail. They wouldn't ship with such a connector, unless the wattage on the PSU was very high and each rail could carry a large current load.


What are the potential risks of using an improper power configuration for my graphics card?

It depends on the design of your PSU and your GPU. In the ideal case, your GPU or PSU (or both) would simply refuse to work, or power itself off, when it over-draws on the current, thus protecting the boards from damage. Attempting to over-draw can cause heat buildup, and in the worst case, fire, because amperage ratings are based on what the wires are rated to handle, based on their resistance (which is another way of saying "heat"). More current, more heat. So you need a lower resistance wire or a good heat dissipation strategy to carry a larger amount of current.

Of course, to be on the absolutely safe side, I am required to tell you that any improper hardware configuration could potentially lead to explosions, death, and loss of life, limb, or property, with no limitation (worst case, your computer catches on fire and causes your entire house to burn down, and then your house catches other houses on fire in the neighborhood, creating an enormous conflagration and billions of dollars of damage.... perhaps even the deaths of firefighters and residents.... but this kind of thing doesn't happen so much these days because of fire codes and electrical wiring standards, etc.)

That said, the only way to know if your PSU can bear your GPU's load is:

  1. Look up your GPU's Thermal Design Power (TDP) or the amount of power required on the +12V "rail". If you get TDP in watts, you'll have to convert this into amps @ 12V using the equation P = V*I (P = power in Watts, V = voltage in Volts, I = current in Amps). Using some basic algebra, if you have watts (assuming 12V connection, which is absolutely the GPU standard for many years) and you want amps, the equation is I = P/V, so divide the Watts by 12 to get the Amps required on the +12V rail.
  2. Look up your PSU's specifications and determine how many rails it has. Fewer rails will tend to have higher amps per rail; more rails will tend to have the total wattage of the PSU split up more, meaning that you'll have to be more careful not to overload one rail.

Now... here is an example

If you have a GPU with a maximum TDP of 120 Watts, 120W/12V = 10 Amps, using the formula.

Whenever you get your GPU's TDP, you need to subtract 75 Watts from it, because the PCI-Express slot itself (the slot you connect it to on the motherboard) will supply it that much power, per Wikipedia:

They can use up to 75 W (3.3 V/3 A + 12 V/5.5 A), ...

So 120 W - 75 W = 45 W. This is how much power this hypothetical GPU with a TDP of 120W would need to draw from the external power connectors (the 6- and/or 8-pin connectors you plug into the card after you have installed it into the slot). Now going back to amps, 45W / 12V = 3.75 Amps. That's not a lot.

If you have a PSU with two +12V rails, it's likely that the external PCI-E connectors are on a dedicated rail of their own, separate from the one that powers the motherboard's +12V.

If you have a PSU with one +12V rail, shared by the mobo and the PCIE connectors, all you have to do is make sure that the sum-total of the motherboard+CPU+disks+GPU+...(whatever else uses +12V) amperage is less than the amount rated by the rail.

Now, per TomsHardware, the R9-270X has a TDP of 180W.

Docking 75W by PCIE slot, that means the external connectors have to push at least 105W, or 8.75 Amps, in addition to whatever else might be bearing on that rail.

If you're lucky, if you have a modular PSU with more than one rail, you may be able to find a "rail map" that tells you exactly which pins on the PSU modular interface map to which rails. If you're unlucky, you won't be able to find such a diagram, and you'll just have to hope for the best :) Have your fire extinguisher ready! (just kidding, it's probably fine lol.)

Solution 2:

The power connections from the PSU to the Graphics board, are just multiple repeats of the same thing, connected to the same places. (I have dissasembled and checked 5 total so far in my own uses). The 8 Pin has an extra ground, which makes it weird. You have 3 sets, 3 +12 and the 3 grounds on each.

Different PSUs will have multiple "rails" (seperated curcuits) or single rails carrying. This is all covered in allquixotic answer.

There is the redundancy, but much more than that. The connection points used for the connections are rated to a certian ammount of amps, when the connection there is less than best, and the GPU card is working very hard, these connection point can heat-up the metals in the lesser connection places, and melt the plastic around the connection.
An often missed point of these connections is the "Crimp" where the connections are crimped onto the wires, The crimp also has a smaller area of connecting to the wires, wrapped around touching only the outside strands and only in limited contact area. It would be very rare to see that anyone had soldered these connectors onto the wires, or added soldering to the crimp to greatly improve that connection location. Sometimes a crimp can be very poorly done also.

Add in some oxidation to both of these areas, and over time the area of the conductor to conductor molecular connection locations is very slightly reduced. Oxides of metals are very poor conductors.

But that is not all. Each of the wires has some resistance , due to the length of the conductor from the psu to the gpu. The longer the wire used and the smaller the "guage" or thickness of the wiring , the more the voltage drops from the original source (PSU) to the card (under load).
The voltage drop will also get higher as the quantity of amps increases through the wiring. (the resistance stays the same, but the current increases).
The board having its own Voltage regulation, can fix a quantity of voltage drop. Each of the boards from different manufactures will have slightly different "qualities" of this power regulation, but most of them would have minor voltage drops covered.

Each wire , connection and crimp set (3sets) of this type could usually handle 10 Amps under most conditions, before things went badly. So with that card , with no overvoltaging/overclocking it should work fine.

PSU wiring can use either smaller guage or there is some PSU using a much larger guage of wiring, It does not have to be that much bigger wire (inside), to heft a lot more amps through it.

You have it working now, and it may very well work fine forever, it should work just fine. The GPU card can also get some of its power from the PCI slot, that is a bit more (lets call it) expencive way to get the power , if somehow it was overtaxed. The greater the voltage drop on this extra wiring, the more any power from the PCI-E will get applied.

The one reason why I would not leave it that way , Sometimes a computer has a minor problem, you will always remember that that item was left undone. was not set the way the manufacture originally intended. So it will always be "wrong" even if that is not part of any problem encountered, it would still make you wonder. Over time , it will not be self improving.

Side Notes: If you get modular wiring from another PSU, they are not all connected the same way at all, so doing extra checking before adding any wiring to a modular PSU from a different one.