AMD Radeon VII Linux Benchmarks - Powerful Open-Source Graphics For Compute & Gaming
Written by Michael Larabel in Graphics Cards on 7 February 2019. Page 8 of 11. 80 Comments

In some games the power efficiency of the Radeon VII is still well behind the latest NVIDIA graphics cards, but it's at least much better than the current Polaris and former Vega graphics cards.

Overall, it's a very nice improvement in performance-per-Watt for the Radeon VII relative to the Radeon RX Vega 64, but still in many instances the latest NVIDIA graphics cards are able to deliver greater power efficiency. We'll see how the situation plays out with Navi later this year.

For those wondering about the overall AC system power consumption numbers, here is a look over the course of all the OpenGL and Vulkan gaming benchmarks. The Radeon VII in the Core i9 9900K system had an average total power consumption of 292 Watts with a peak of 441 Watts, which is better than the 314 Watt average of the RX Vega 64. The TITAN RTX had the highest average power draw at 319 Watts, but peaked to 433 Watts, lower than the Radeon VII, The GeForce RTX 2080 meanwhile had an average system power draw of just 255 Watts with a peak AC system power usage at 342 Watts.

There is the GPU thermal results over the course of all the benchmarks run, but be aware those numbers for the Radeon VII are certainly inaccurate. This is the main but lone bug I've hit with the Radeon VII during testing. As of Linux 5.0, the Radeon VII temperature reporting appears well off with this Vega 20 graphics card while the other Radeon GPUs were reporting accurate thermal numbers from their hwmon interface on sysfs. The reported Radeon VII numbers from the driver are certainly too high. Had the Radeon VII really been running at 90~109C, there would have certainly been thermal throttling to kick in and would have severely handicapped the performance of this GPU. So it's likely not a power management issue per se since the performance appeared unaffected, but the reported temperature to user-space is most certainly incorrect. I've reported this to AMD but have yet to hear of any resolution to the issue or explanation of what went awry.

Over the course of all the graphics/gaming benchmarks, here is a look at the harmonic mean of all the gaming FPS benchmarks carried out. From this high-level metric we see the Radeon VII came in at 27% faster than the Radeon RX Vega 64. This boost with 7nm Vega was enough to put its harmonic mean right behind the GeForce RTX 2080: the RTX 2080 had just a 4% lead over the Radeon VII... Certainly if just looking at OpenGL-powered games, the Radeon VII average would come out ahead of the RTX 2080 while with Vulkan it was more of mixed results around this similarly priced graphics card.

Related Articles

Trending Linux News