Performance isn't the only important metric when it comes to graphics cards. We also test power consumption using in-line monitoring tools and Powenetics software. We log power, clock speeds, temperatures, and fan speeds. We loop the Metro Exodus benchmark five times at 1440p ultra settings, and then run FurMark stress test at 1600x900 for over 10 minutes. We also measure noise levels using an SPL meter. We collected the data using the default profile, as well as the OC mode.
ASRock clearly threw caution to the wind on the RX 6900 XT Formula. Even in the default mode, it used more power in our gaming test than any other GPU in our charts. Turn on the OC mode and power use jumps an additional 30W in Metro Exodus and 40W in FurMark. The triple 8-pin power connectors are definitely required to deliver this level of power, and the PCIe power draw maxed out at under 40W even in OC mode. Compared to the reference 6900 XT, the ASRock Formula uses 35% more power in gaming and nearly 40% more power in FurMark — all for about 6% more performance on average.
We've known for some time that AMD's Big Navi / RDNA 2 architecture was really tuned to hit higher frequencies. The official boost clock on the reference card is 2250 MHz, but even that card averaged 2327 MHz in our gaming test. The ASRock matches the clocks we've seen on the RX 6600 XT and RX 6700 XT, breaking 2.5 GHz in average clocks, and OC mode adds another 60 MHz. Clock speeds are also quite high in the FurMark stress test, with the ASRock in OC mode basically doubling the clock speed of the reference RTX 3090 card. Again there was a 60 MHz bump in clocks with the OC mode, which honestly isn't much for the extra 30-40W of power that's required. It's a classic case of the diminishing returns we see when moving up the voltage and frequency curve.
Last, we have fan speeds and temperatures, which are very much interlinked. Some cards shoot for lower temperatures and ramp up fan speeds more quickly while others are okay with temperatures in the 75C range and take a more relaxed approach to fan RPMs. Given the power draw, it's not too surprising that ASRock needs both high speeds and relatively high temperatures.
The RTX 3080 Ti and RTX 3070 Ti are the only GPUs that ran hotter, and the reference cards definitely came near the limits of their cooling potential. ASRock only needed relatively modest fan speeds in the default mode, but OC mode needed a big bump in RPMs to keep thermals in check. Compare that to the Asus RTX 3080 Ti card that employs liquid cooling, and you can see that the combination yields lower temperatures and fan speeds.
We also measured peak noise levels using an SPL meter at a distance of ~10cm from the GPU fans, pointed at the center fan. Keep that distance in mind, because it's intended to show more of a difference between the graphics cards we test. At a normal distance of around one meter (where your head is relative to the case), noise drops about 10 dB(A).
At stock settings, the ASRock card generated 48.1 dB(A) of noise running Metro (just sitting in the game rather than looping, so there's no periodic dip when the benchmark loops). That corresponded to a fan speed of just 52%, so there's plenty of room for more noise and better cooling. Using the OC mode kicked the fans up to 62% and 52.3 dB(A), while setting a static fan speed of 75% resulted in noise levels of 60.0 dB(A).
MORE: Best Graphics Cards
MORE: All Graphics Content