Frequency, Temperature & Noise
Temperature & Clock Frequency
AMD's fan controller is fairly conservative. Consequently, the Radeon Vega FE reaches its maximum temperature of 84°C (or, briefly, 85°C) relatively quickly. At that point, performance has dropped by approximately 10 percent, which is almost exclusively due to automatic frequency throttling. For each 5°C increment, we computed an average clock rate that took into account the peaks and valley during our benchmark run. The resulting range went from 1440 MHz for a cold card all of the way down to 1269 MHz at Vega 10's peak temperature.
The following graph shows that performance scales well with frequency. Using 1269 MHz as a basis for comparison, a 12% increase to 1440 MHz yields approximately 10% more gaming performance.
Temperature & Power Consumption
With the Radeon Vega FE running at 1269 MHz due to temperature-imposed throttling, it consumes an average of 266W. At 1440 MHz, the card hits almost 300W. This means that a 12% clock rate increase necessitates 13% more power consumption. Together, these figures enable the aforementioned 10% frame rate boost.
Such impressive scaling suggests that there's room for even better performance. Of course, you'd be looking at power consumption well beyond the 300W mark, which seems a bit dubious.
Losses to leakage don't seem to play much of a role in our measurements. The days of saving 40W or more by simply keeping temperatures low enough are gone, it appears.
GPU vs. HBM2 Memory Temperatures
Unless the sensors are lying to us, the GPU’s maximum temperature is 84°C (85°C peak), while the HBM2 gets up to 95°C (96°C peak). That latter figure seems fairly high, but it does end up close to the ceiling for GDDR5X. We'll keep an eye on both temperatures as we get our hands on Radeon RX Vega 64; it's important to ensure the sensor data is 100% accurate.
What jumps out to us is that the board just below the GPU reads ~5°C cooler than the inside of Vega 10! The obvious question is why, and we already gave the answer earlier in our review. The Radeon Vega FE’s GPU and HBM2 sit on an interposer attached to a package substrate positioned on top of the PCB. Furthermore, the interposer doesn’t seem to make full contact with this substrate, causing a so-called underfill issue. Air between the layers acts almost like insulation.
The following time-lapse video shows the Radeon Vega FE warming up. It runs at 10x, showing the original 20-minute sequence in two minutes.
During the stress test, temperatures are a little lower due to the fan spinning faster and Vega 10's lower (throttled) frequency.
Let’s not beat around the bush: this is the best reference cooler we've seen from AMD since its Radeon HD 2900. A much-improved radial fan supplied by Delta allows the Radeon Vega Frontier Edition to compete on even footing with Nvidia’s GeForce GTX 1080 Founders Edition. According to our measurements, the card’s maximum noise level is 44 dB(A), which almost completely drowns out the coil whine present at very high frame rates. Our video demonstrates this well, especially towards the end:
The frequency analysis doesn’t reveal any low-frequency growling that would be a sign of bad bearings or cheap motors. Meanwhile, the coil whine can be spotted around 8 kHz on the following graph:
Most of the noise is generated by the fan, but it's certainly acceptable. An Nvidia GeForce GTX 1080 Ti, with its ~250W of waste heat, isn’t significantly quieter, even though it's a lower-power card. If you push the 1080 Ti to a more comparable 270W by increasing its power target, noise rises to the same level as AMD's Radeon Vega FE.
This particular competition ends in a draw, with both manufacturers probably hitting the physical limits of what can be achieved. Moving to more expensive cooling solutions would only yield diminishing returns.
MORE: Best Graphics Cards
MORE: All Graphics Content