Thanks for the report. We have recently removed the dependency on the GLU library entirely, so future versions of the benchmark will work fine without it.

Assigning to Sybren for now. We have to discuss further how to handle the case, since it can't be easily reproduced on our machines here AFAIK.

Yes:

Could you make a screenshot, please?

Thanks for updating us; marking as solved.

Unfortunately my system changed since this bug report, I switched back from linux to windows 10.
I just ran the quick CPU benchmark and publishing resulted in the same error. BUT when I refreshed the browser tab I was met with some nice bar graphs and the shocking realisation that my render times are twice the average.

This should be resolved now, can you try again?

Experiencing the same error here.
macOS High Sierra 10.13.6 with Blender Benchmark 1.0beta2.
Error log is attached.

There is an issue with the macOS benchmark, will be fixed in the next update.

Sorry.
I didn't know I couldn't.

Please don't change priority yourself.

This is now partially implemented (when viewing a result, we can compare with the average).

same for the Windows64-bit version:

if changed in the script manually to the right value, all works fine:

Same here. We're seeing this on our production builds at 10.12.5 and our beta builds at 10.13.6 tested across iMacs, Mac Pros and MBPs and iMac Pros. error report attached.

Same thing for me on MacOS 10.13.6

same thing just happened to me
MacOS 10.12.6

Addressed feedback on the implementation in the wip-compare-results. Can we have another round of feedback and merge this @Sybren A. Stüvel (sybren)?

Resolved in My Data 5250b653f34cb168afa77e51f32e6f70141dae28, Open Data 02f37fae5a7b69d53cda349e10e7b2ef13195c08, and Benchmark Bundle 4726ee0c9c058d755acb2bbbbac997f1d88c3c90

The data is accessible via https://opendata.blender.org/raw-data/.

Sergey says it's important to have the display/non-display info in there, so simply stripping is not an option.

Hi Sybren,

Thank you very much for your answers Francesco :-)

We definitely plan to provide some more in-depth metrics for certain results.
Check the rest rest of the work board to get an idea of what is coming next. In particular T56478 and T56323 will (at least partially) address your question. For the rest, there's T56320.

Ok, thanks for the answer, so... Why to show only the 5 fastest GPUs and CPUs and not all the tried ones ?
This could be very useful for all users, and more motivating to everybody trying the benchmark software, you don't think so ?

This actually the expected behaviour of the "view benchmark" page. The code that you see is the actual result of the benchmark.
We plan to add a description that better explains that, as well as some charts that show the result in context of the other results on the platform. Thanks for the feedback!