Benchmarking CPU and GPU across Windows 10 and CentOS 7 with Blender Benchmark

Free AI web copilot to create summaries, insights and extended knowledge, download it at here

1083

Abstract

/a></li><li>48 GB (6x8 GB) DDR4–2666 ECC Registered RAM</li></ul><p id="2b81">The boot drives are identical M.2 NVME SSDs. On Windows, I’m running 10.0.17763, with the <a href="https://www.nvidia.com/download/driverResults.aspx/145178/en-us">419.67 Creator Ready Driver</a>. On Linux, I’m running CentOS 7.6–1810 with kernel 3.10.0–957.10.1.el7.x86_64 and the <a href="https://www.nvidia.com/Download/driverResults.aspx/145182/en-us">418.56 driver for Linux</a>.</p><p id="d6af">I ran full CPU and GPU tests, and the results surprised me — mostly because <a href="https://www.pugetsystems.com/labs/articles/DaVinci-Resolve-14-Performance-Windows-vs-Linux-1126/">Puget Systems had already deemed Linux “meh” with regards to performance, compared to Windows</a>.</p> <figure id="4f7e"> <div> <div>

            <iframe class="gist-iframe" src="/gist/sethgoldin/e1a393cda096f37b140fc4ea595a7ae6.js" allowfullscreen="" frameborder="0" height="undefined" width="undefined">
          </div>
        </div>
    </figure></i

Options

frame></div></div></figure><p id="f04b">My results show that Linux outperforms Windows drastically for CPU usage, and slightly for GPU usage. This is <a href="https://en.wikipedia.org/wiki/Internal_validity">internally valid</a>, but I’m not sure if this is <a href="https://en.wikipedia.org/wiki/External_validity">externally valid</a> for DaVinci Resolve.</p><p id="d899">I used Blender Benchmark because it’s a nice little cross-platform, self-contained test, and didn’t require managing a bunch of different clips and setting up all sorts of different renders on the two different operating systems. I don’t know if Resolve consistently uses CPU and GPU in a similar enough way to infer that Linux has the edge for all workflows. The test really just seems to confirm my priors that Linux is more powerful than Windows, but hasn’t swayed my belief very much, as per proper Bayesian updating. I will continue to recommend running DaVinci Resolve on Linux as the preferred platform for stability and performance.</p><p id="1175">I also welcome further analysis.</p></article></body>

Linux seems to outperform Windows, but external validity for other post-production applications is unproven

I dual boot Windows and Linux, so I figured this tool would be a great way to compare the two operating systems, apples to apples. System specs:

The boot drives are identical M.2 NVME SSDs. On Windows, I’m running 10.0.17763, with the 419.67 Creator Ready Driver. On Linux, I’m running CentOS 7.6–1810 with kernel 3.10.0–957.10.1.el7.x86_64 and the 418.56 driver for Linux.

My results show that Linux outperforms Windows drastically for CPU usage, and slightly for GPU usage. This is internally valid, but I’m not sure if this is externally valid for DaVinci Resolve.

I used Blender Benchmark because it’s a nice little cross-platform, self-contained test, and didn’t require managing a bunch of different clips and setting up all sorts of different renders on the two different operating systems. I don’t know if Resolve consistently uses CPU and GPU in a similar enough way to infer that Linux has the edge for all workflows. The test really just seems to confirm my priors that Linux is more powerful than Windows, but hasn’t swayed my belief very much, as per proper Bayesian updating. I will continue to recommend running DaVinci Resolve on Linux as the preferred platform for stability and performance.

I also welcome further analysis.