Power Consumption and Overclocking
Direct GPU Power Consumption Testing
With a TDP limit of just 175 watts, I was curious how closely AMD would actually get to that level with the new Radeon R9 Nano. As it turns out: damned close. First, let’s look at our standard “from the wall” power testing that looks at the whole system, not just the graphics card.
The R9 Nano is definitely using less juice than its big brothers – 74 watts less than the Fury X and 68 watts less than the Sapphire Radeon Fury card. Idle power is also noticeably lower, though how much of a difference the 3-7 watts means to you there is likely pretty low.
Compared to the ASUS GTX 970 DC Mini card however, the R9 Nano is still a bit more power hungry. How that translates into real-world gaming performance has yet to be seen though.
Let’s take a more interesting look at power on the GPU though, by measuring on the GPU directly. Thanks to a setup we integrated late last year we can now measure the power draw of the graphics card through the PCI Express interface as well as through the ATX power connections on the power supply, giving us a card-specific power draw number that is much more compelling, if also much more complex to analyze.
Feel free to click this for a larger version, but the results are pretty clear with only three data points on the plot. The grey line on the bottom is the GTX 970 DC Mini card from ASUS, showing a directly measured power draw of around 125 watts. That’s actually a bit lower than I expected but since the reference card was rated at 145 watts, this seems in line with a lower power, small form factor version of the card. The blue line represents the R9 Nano in its stock state. You can see that the peaks of the blue lines (that are instantaneous measurements of total power draw) just top out at the 175 watt level, keeping the card in line with expected TDP levels.
The orange line represents the power consumption of the Radeon R9 Nano when “overclocked” – setting the power settings in the Catalyst Control Center to its maximum of 50%, without manually changing the clock speed at all. Clearly we are changing SOMETHING though as the peak power draw jumps all the way up to 240 watts!
Which leads us to…
Overclocking the Radeon R9 Nano – Doing the dumb things
Overclocking the R9 Nano is a bit different than previous AMD GPUs. Because there is a set, and somewhat low, power limit to the card of 175 watts, and we know the GPU can get quite a bit higher than that, changing the power limit in the CCC is likely to make a pretty big difference.
By swinging the power limit up to its maximum value, and not modifying the clock speed settings at all, I was able to make a significant change the behavior of the Radeon R9 Nano.
With the power limit jacked up, the R9 Nano had no issues hitting and maintaining the 1000 MHz clock rate nearly the entire time looping through our Metro: Last Light benchmark. That is a 112 MHz increase in clock speed, a 12% uptick over the stock settings.
This results in a noticeable performance increase of about 11% if we are looking at average framerate. Also, interestingly, it LOWERS the variability of the frame times as well – look at the Frame Times graph and compare the thickness of those plots. Both are well within the range of acceptable gaming experiences but it is interesting to note the potential is there for frame time variability with drastic, quick swings in clock speed.
Overclocking ABOVE the 1000 MHz level is more or less as fruitless as overclocking has been on the Fury and Fury X in our previous reviews – it’s just not worth the effort.
Obviously if you are looking to overclock your R9 Nano, you’re probably buying the wrong card. The Nano, as I have stated before, is for a very specific use case and user scenario. If you are looking to overclock it and push the power limit up, you should probably be looking at a Fury or Fury X option.