A few questions about Blender Benchmark Launcher and errors/bugs management

Hi there!

I’m about to change my pc so I want to have full Blender benchmarks with old and new config. So I just downloaded Blender Benchmark Launcher from opendata, but I have two issues.

  1. Where are the downloaded content?
    Surely I have that 2.90 build on my computer, as well as the scenes, why not tell the tool to use them instead of redownloading everything. And if I want to re-do the benchmark later, it seems that it wants to re-download everything? Unless that’s because of issue 3, surely there’s a better way?

  2. Downloads are insanely slow!


    I could download these 200MB within a minute, but at the time I took this screenshot it already spend about thirty minutes and still counting. I actually left my computer alone for the day at this point, hoping to enjoy the benchmark results once back home after work. But nope:

  3. I can’t benchmark all the files at once, I always end up like this:


    937e48b682bb414fb2c7afe04474e50b
    Which raises more issues on the UX side:

    1. I think users should have more than a “something went wrong” error, because maybe he can do something about it instead of either spending a huge amount of time reporting bugs - waiting for a fix and redoing everything again or give up for time & sanity 's sake.
    2. Sure, facing an error is not good UX on itself, but managing it could be.
      So where am I supposed to report a bug? What am I even supposed to post besides the error ID and “it says something went wrong”
      I could post on Blender’s tracker or here or in some Blender related website or somewhere else I don’t know, and I have to guess. I might get it wright or wrong, and not every Blender user knows how to do such things. Why not have a link right here in that screen? Why not make that link prefill the bug report with the needed info (including that beautiful error ID) like Blender’s report a bug button does?

I post here first since to me it’s also more of a discussion and feedback rather than a strict bug report., but I’ll happily make a bug report on the tracker if needed.

– Blendy regards.

1 Like

Perhaps I can speak to this a bit.

Our connection to the rack is only 1 gigabit, so it can be slow at times due to throttling and QoS. Worse, traffic from Amsterdam to certain places, such as LA or Australia, has been known to be a maximum of about 15mbit, which could be from any number of reasons.

In addition to potentially slow intercontinental bandwidth, we also throttle bandwidth to download.blender.org due to the sheer volume of people we have downloading Blender and the open movies, basically 24/7. Mostly likely this is the limit you are hitting if the speed is limited to around 100KB/sec.

Unfortunately I can’t do a whole lot from my end in terms of our maximum bandwidth; we have what we have, and thus it must be rationed! It is possible that maybe @fsiddi or someone could place the download on the CDN to help with this.

Ideally we could redirect people to mirrors, but this would require a script on our end to dynamically send users to nearby mirrors for blender, as this would not be something you would want to hard code into a binary such as the launcher.

3 Likes

Right, so there are technical limitations, I can understand. Though I still can download from download.blender.org under a minute, or two in rush hours. Is there much of a difference between the benchmark and the website download?

Concerning mirrors, is P2P possible? If the BF provides its official homemade torrents, I bet some people will gladly use them, and it only needs a few scattered around the world to give everyone a great download speed.

GIMP does that, and it works well.

I honestly don’t know anything at all about how they designed their downloader tool. For all I know they throttle it themselves. If it downloads from builder.blender.org though, it’s gonna get a bandwidth cap different from download.b.o

Anything is possible! I did some tests of the same torrent software that Ubuntu used and tests went somewhat okay, but there were some tricky issues relating to our internal setup that made it less than ideal, and I eventually threw out the test. Perhaps PTP could be resumed again though.

I believe that Blizzard has been doing this as well for quite a while. I think we also do it for some of our videos for something in Cloud.blender.org or the community site or something, I forget which.

Thanks for the feedback. There are several aspects of the UX for Blender Benchmark that could be improved, but the project hasn’t been actively looking for contributors to address this. It’s an item in my todo list.

When it comes to download performance: we use the same mirrors that blender.org/download uses, so it might have been a mirror issue. Let me know if the issue persists.