A little warning first. I've just done a swing shift followed by a split shift. I had 2 hours sleep from 19 to 21 hours ago. I might not be very clear in this post 
I have one of these cards. I don't have working fan speed control. I am particularly interested in hearing from someone who has replaced the stock cooler on one of these cards and still has it to look at. I'm also interested in anyone who knows more about this aspect of electronics than I do and who spots a flaw in my reasoning.
I'm currently at the conclusion that the XFX 7800GT Extreme LED Edition that has UDER in the model number has a design flaw that makes fan speed control impossible to implement. It's a strange conclusion, so I'm putting my line of reasoning up here for people to look and and see if there's anything wrong with it.
Premise 1 The only connection from the fan on the stock cooler to anything else is the two wires leading to the 2-pin power header on the graphics card.
Premise 2 This 2-pin connection carries only power. Fan speed information would require a third wire.
Conclusion 1 The only possible methods to vary the speed of the fan are to change the constant voltage supplied to it (e.g. 5V instead of 12V) or to use pulse width modulation (PWM), which achieves the same effect by switching the voltage on and off frequently.
Conclusion 2 Whichever method is used, it has to be implemented before the power header on the card. Even if the two wires carrying power from the power header pass through a suitable component connected somewhere between the power header and the fan, or embedded in the fan, neither method would work because there is no way to instruct that hypothetical component what the fan speed should be - the only connection is those two wires, neither of which carries fan speed information. Although it would be possible for this hypothetical component to also be connected to the graphics card and get that information from a different source. I see no evidence of any such component, anyway.
Premise 3 The LEDs on the card are powered from the same power header as the fan and therefore via the same wires. This is where it would be handy if someone can take a look at a stock cooler removed from the card. I have looked as closely as I can without removing the cooler or the cover on it, because I don't want to void my guarantee. The wiring is mostly visible and it certainly looks like the LEDs are powered from the same power header as the fan.
Premise 4 The LEDs require a constant voltage. Otherwise they would vary in intensity.
Premise 5 PWM would have the same effect on the LEDs as a constant lower voltage - it would vary their intensity.
Conclusion 3 Anything that would vary the fan speed would also vary the intensity of the LEDs.
Premise 6 The LEDs are designed to require the maximum voltage supplied by the power header, the voltage that results in 100% fan speed, to shine at their intended brightness. If this was not the case, they would be significantly over-volted every time the PC was powered up, as the fan speed is intended to be 100% from power up to the graphics drivers loading.
Premise 7 The card is designed to have the LEDs at a constant intensity. Having them glowing at normal intensity from power up to graphics drivers loading, quite dimly when 3D software is running and very dimly when 2D software is running would more or less negate the whole point of having the LEDs.
Final conclusion Fan speed control cannot exist on this card.
I have one of these cards. I don't have working fan speed control. I am particularly interested in hearing from someone who has replaced the stock cooler on one of these cards and still has it to look at. I'm also interested in anyone who knows more about this aspect of electronics than I do and who spots a flaw in my reasoning.
I'm currently at the conclusion that the XFX 7800GT Extreme LED Edition that has UDER in the model number has a design flaw that makes fan speed control impossible to implement. It's a strange conclusion, so I'm putting my line of reasoning up here for people to look and and see if there's anything wrong with it.
Premise 1 The only connection from the fan on the stock cooler to anything else is the two wires leading to the 2-pin power header on the graphics card.
Premise 2 This 2-pin connection carries only power. Fan speed information would require a third wire.
Conclusion 1 The only possible methods to vary the speed of the fan are to change the constant voltage supplied to it (e.g. 5V instead of 12V) or to use pulse width modulation (PWM), which achieves the same effect by switching the voltage on and off frequently.
Conclusion 2 Whichever method is used, it has to be implemented before the power header on the card. Even if the two wires carrying power from the power header pass through a suitable component connected somewhere between the power header and the fan, or embedded in the fan, neither method would work because there is no way to instruct that hypothetical component what the fan speed should be - the only connection is those two wires, neither of which carries fan speed information. Although it would be possible for this hypothetical component to also be connected to the graphics card and get that information from a different source. I see no evidence of any such component, anyway.
Premise 3 The LEDs on the card are powered from the same power header as the fan and therefore via the same wires. This is where it would be handy if someone can take a look at a stock cooler removed from the card. I have looked as closely as I can without removing the cooler or the cover on it, because I don't want to void my guarantee. The wiring is mostly visible and it certainly looks like the LEDs are powered from the same power header as the fan.
Premise 4 The LEDs require a constant voltage. Otherwise they would vary in intensity.
Premise 5 PWM would have the same effect on the LEDs as a constant lower voltage - it would vary their intensity.
Conclusion 3 Anything that would vary the fan speed would also vary the intensity of the LEDs.
Premise 6 The LEDs are designed to require the maximum voltage supplied by the power header, the voltage that results in 100% fan speed, to shine at their intended brightness. If this was not the case, they would be significantly over-volted every time the PC was powered up, as the fan speed is intended to be 100% from power up to the graphics drivers loading.
Premise 7 The card is designed to have the LEDs at a constant intensity. Having them glowing at normal intensity from power up to graphics drivers loading, quite dimly when 3D software is running and very dimly when 2D software is running would more or less negate the whole point of having the LEDs.
Final conclusion Fan speed control cannot exist on this card.
