Power Consumption

Power consumption has been AMD's biggest problem area in recent years, especially when compared to the huge advances NVIDIA made here. RX Vega comes with an updated approach to power management, thanks to a beefed-up power-management microcontroller in the GPU die, which monitors and regulates all aspects of the board's power delivery and thermals.

Single-monitor idle power consumption seems surprisingly high, especially when considering that cards like the RX 580 managed to go below the 10 W mark, making them competitive with NVIDIA's designs which run single digit numbers in this test. One reason for this seems to be the relatively high idle GPU clock of 852 MHz, while memory goes down all the way to 167 MHz.

Multi-monitor power consumption seems solved, finally. With 16 W, the power increase over single-monitor idle is just 2 W, a huge improvement over the 15 W or so that were previously added in that state (on Polaris); the same goes for media playback.

We saw very high power consumption in today's other review of the RX Vega 64. The RX Vega 56 does a lot better; the difference is almost magical. With 229 W, power draw is 30% better than with RX Vega 64, at the cost of 10% performance, resulting in greatly improved performance per watt. With those numbers, Vega efficiency competes with some Pascal cards - these are the numbers I was hoping for with Vega.

It seems that RX Vega 56 runs much closer to the optimal voltage/frequency curve than RX Vega 64, which seems to have been clocked very high just to match GTX 1080 performance. Each chip design has an optimum efficiency operating point that delivers a certain performance; if you want higher performance, you have to sacrifice some efficiency to drive the GPU hard enough to achieve your goals.