Overclocking the RX 9060 XT 16GB:
In our RTX 5060 Ti 16GB test, we had three models to review on launch day, so we weren’t able to test overclocking. That’s what we’ve done with this report, comparing the performance of our three models. So we’re going to evaluate the capabilities of our XFX Swift RX 9060 XT 16 GB when it comes to overclocking.
Power Limit, the RX 9000’s weakness?
Another key element to highlight is the ” Power Limit “, the box we’ve highlighted in GPU-Z’s ” Advanced” tab. But what is this Power Limit? Our XFX Swift RX 9060 XT 16 GB has a TDP of 180 watts, which means that when the card reaches this TDP, performance will be maximized. If the card were to exceed 180 watts when overclocking, it would automatically “throttle” itself in order to maintain its TDP, by lowering the frequency, for example, in order to stay within the limits. The margin offered by certain models allows this TDP to be exceeded. Our example of the day has +10%, i.e. the card can theoretically go up to 198 watts. If you’d like to visualize the Power Limit principle, we invite you to read our article on undervolting the ASUS TUF RX 9070 OC.
STEP 1: performance at original frequencies
We launch a first run under 3DMark Time Spy to see how the card behaves. The idea, of course, is to have a reference score to measure the gain throughout this overclocking session. We’ll start by keeping an eye on three values: the overall and GPU score obtained in Time Spy, the maximum frequency of our Boost frequency and the power consumption recorded in GPU-Z.
The overall score is 17269 points and the GPU score is 16864 points. The Boost frequency reaches a maximum of 3232 MHz and the GPU-Z TDP is 183 watts.
STEP 2: increase the power limit
This has been one of the new features of the last few years, and is also common to NVIDIA cards. Increasing the power limit provides additional margin for the graphics card. In this case, we’ve increased the power limit by 10%. We forgot to include the screen, but the gain is very small: the overall score is 17814 points (+ 3.1%), while the GPU score is 17071 points (+ 1.2%). Boost frequency reaches a maximum of 3292 MHz.
STEP 3: Overclocking proves ineffective…
Normally, once the power limit has been pushed to its maximum, the Boost frequency is increased in increments to achieve maximum stability. But here, with our RX 9060 XT, we’re quickly limited and instead of seeing a score that continues to improve, it regresses.
On the left, the capture with a 100 MHz increase in Boost frequency and on the right, a 200 MHz increase. The effect is non-existent.
STEP 4: Voltage reduction
Still using the AMD drivers, we lower the voltage, step by step, until we find the lowest voltage that still allows us to run our benchmark. If at any point our card runs out of juice, the benchmark will crash and we’ll know we’ve reached the voltage limits.
We were able to lower the voltage to -125 mV and thus improve our score. Our overall score is 18267 points (+ 3.6%) and the GPU score is 17600 points (+ 4.3%), but that’s nothing to get excited about!
STEP 4: memory overclocking
Another element we’ll be able to influence is the frequency of our GDDR6 memory. By default, it’s 2518 MHz, but logically, from experience, it should be at least 2800 MHz. For the record, memory clocked too high causes artifacts on the screen.
In the end, we were able to obtain an overall score of 19056 points, giving us a gain of 8% on our base score and 18454 points for the GPU, i.e. 9.4% better than the original. Maximum frequency is 3483 MHz and power consumption is 200 watts, with a voltage reduction of -140 mV.
