PowerColor R9 290X OC Reviewccokeman -
» Discuss this article (9)
PowerColor R9 280X OC Testing:
As with just about every launch we get the press guides that answer questions and illustrate the architecture of the product, the positioning, and even performance numbers generated by the manufacturers' in-house teams on a "spec" system that usually differs from the setup we use here at OCC. One thing caught my eye while browsing through the press deck materials on this go round. It was a chart (shown below) that shows a distinct performance difference when running the different fan modes: Quiet for low noise in all situations and Uber that eliminates some of the constraints you have with Zero Core power profiles with an increased fan speed that jumps the fan speed up by about 1000 RPM. Not a huge change in the grand scheme of things when you look at it, but as I was about to find out we ended up with a slight problem that for the most part will go unnoticed until your game performance starts to drop when you have the baddest card that AMD makes in your system.
That being said the two modes are designed for different purposes. "Quiet" mode is designed to run as quiet as possible, forsaking everything else, and keeping the fan speed at a maximum of 40% to, as you guessed it, keep the noise in check. Utilizing a 28nm core with over six billion transistors, the core is huge and needs to be cooled. As we saw with the temperature testing, the R9 290X reaches 94 °C, just under the 95 °C thermal limit of the card. At this point something has to give if there is no way to keep the card cool with a higher fan speed. What gives is the clock speeds. Once the heat builds up sufficiently that the airflow of the fan is not enough to keep the core from overheating, it dynamically reduces the core clock speeds down as low as an observed 669MHz to maintain the thermal thresh hold. As the thermals improve so do the core clock speeds.
Now when you look at Uber mode we see the expected performance levels on the R9 290X as the fan speed can be maintained at a high enough level (55%) to keep the core clock speeds at optimum levels, or at least the rated levels. This we can plainly see in the screen captures below that effectively show the clock speed drop when the R9 290X is at its thermal limit. So lets put that in perspective with the NVIDIA GTX 780 that does dynamically manage clock speeds based on a thermal and power limit to keep the card performing optimally in game and with the noise it generates. By contrast the GTX 780 does drop clock speed as it warms up and the thermal rise. The caveat is that the GTX 780 does not drop below its rated clock speed of 863MHz on the core. Again this is seen below.
"Quiet Mode" "Uber Mode"
Seeing how the clock speed drops dynamically in "Quiet" mode, we had to revisit our testing and verify Uber mode was enabled to show the expected performance characteristics in the main testing. But we decided to show here how the R9 290X performs when run in "Quiet" mode and how that option compares to Uber mode while comparing this directly with the GTX 780. To do so we heated up all of the cards while in game for five minutes before a benchmark scenario was run. Resolutions tested are 1920x1080 and 5760x1080. We are using a sampling of our normal benchmark suite, but have included Batman: Arkham Origins as a replacement for Batman: Arkham City. The settings used mirror those in our benchmark suite tests for a direct comparison.
Batman: Arkham Origins:
Far Cry 3:
Metro: Last Light:
When you get to the end of this and draw a conclusion, you see that PowerTune does its job quite well at keeping the thermals and noise in check when running in "Quiet" mode. However that reduction in noise comes with a significant performance penalty across the board in this series of five games run through two resolutions. FPS drops associated with the use of Quiet mode at 1920x1080 start at just under five FPS and reach almost 13 FPS depending on the game tested. At 576 x1080 the lower overall FPS lessens the actual FPS impact in most games, but if looked at on a percentage basis you can come up with the same conclusions that running in Quiet mode will cost you FPS. If the noise associated with a reference fan is of no consequence to you since when gaming a headset is worn, then I suggest you run Uber mode for the higest FPS levels.