A new boinc project called trailertcrack has been deployed, it aims at cracking the seed for the Thumbnail 2 of the Trailer video of Minecraft (https://www.youtube.com/watch?v=MmB9b5njVbA), this is currently the last one missing so we can fully reproduce the original trailer within Minecraft in higher quality right in time for the 10th anniversary of this iconic piece of Minecraft history. (we do already have most of the seeds and this is an active project).

This app is GPU only and will be rewarded 3333 credits We currently only support Windows and Linux 64 bits for NVIDIA GPUs (we use cuda).

Happy crunching !
Cheers, Neil

Cool, nice to have a gpu app again here.

I'm curious how long this will actually take, but enjoy having your GPUs loaded up (though, not as heavily power-wise compared to Kaktwoos-cl!). I think I estimated around a month with 500 'GTX 1650 (1060)' level GPUs. Though having a couple of faster ones in the mix shouldn't change that too much.

Will be great after all these months to have the final seed for the trailer too.

OK, there is an issue with the new trailertcrack app. It is running two instances on the same card. There was a similar issue over at GPUGrid a month ago. BOINC is improperly assigning the two instances of an application and task to the same gpu even though it says it is assigned to separate gpus.
Here it is running two tasks on gpu:0 even though BOINC is reporting that a tasks is running on gpu:0 and gpu:1.

So currently in our code we do account for a --device arg that will set the device used by nvidia https://github.com/MrSpike63/TrailerTCrack/blob/master/main.cu#L667.
However somehow your client doesn't pass that argument to the command line. I am trying to reproduce it but so far with 2,3 and 4 gpus no issues. Do you mind sending either the link to the issue on gpugrid or the version of your boinc client?

First reported by Ian here. https://www.gpugrid.net/forum_thread.php?id=5246&nowrap=true#57261
And more information by ServicEnginIC. https://www.gpugrid.net/forum_thread.php?id=5248&nowrap=true#57272
You can read the diagnosis in the thread.
Was resolved by the release of the 2.18 application to replace the faulty 2.17 app.
No information about what was actually done in the app by the dev/admins AFAIK, at least publicly in the forums.

I am running a custom BOINC version based on the 7.17 branch. Based on TBar's All-in-One client with further GPUUG binary modifications known as the Pandora client.

I'll try and get Richard Haselgrove to look at your code snippet and see if he can figure out why my client is tripping up on the device argument and gpu enumeration like what appeared to have happened at GPUGrid.

But one thing different is the new app is not a wrapper app like at GPUGrid.

In the meantime I have reconfigured the pandora_config and app_config files to only allow one gpu task onboard at any time. That will work around the issue until I better understand what is happening.

We found the issue, a fix is being deployed, this should be fixed before 02/10 00PM UTC

That's great, Neil. Glad to have pointed out the issue so soon in the beta app deployment and a fix was found quickly.

That is why we do beta deployments against the installed base. I am great at finding corner cases.

In case this helps future projects, here was the issue:
BOINC docs mention the client passing "--device (device number)" as an argument, but for newer versions of the API (7.5+) - which pretty much any BOINC project should be on now, it isn't passed anymore:
https://github.com/BOINC/boinc/blob/73a7754e7fd1ae3b7bf337e8dd42a7a0b42cf3d2/client/app_start.cpp#L729
With that said, thanks for the report. It should be fixed soon.

Either the website server stats are VERY wrong (524192 available tasks) about new app., or I have a different problem, because I can't seem to download any tasks on all 4 of my Windows 10 machines. "Run Test Apps" - checked, "Use NVidia GPU" - checked. Messages in BOINC client -
9/30/2021 7:40:07 PM | minecrafthome | Sending scheduler request: Requested by user.
9/30/2021 7:40:07 PM | minecrafthome | Requesting new tasks for NVIDIA GPU
9/30/2021 7:40:09 PM | minecrafthome | Scheduler request completed: got 0 new tasks
9/30/2021 7:40:09 PM | minecrafthome | Project requested delay of 7 seconds

Ideas?

---

They may have stopped sending out tasks till the new updated application is deployed.

What is 00PM?

amd radeons don't work with the project?

What is 00PM?

I am confused also. Think that the time and date was Euro format. So October 2nd 0000 UTC is likely what was meant for the new app release date and time.

amd radeons don't work with the project?

Not with the new trailertcrack gpu project. CUDA only. For now.
The old kaktwoos cactus stacking project was an OpenCL application so worked with any gpu with OpenCL support.

Already out of work?

---

Reno, NV
Team: SETI.USA

They have found a bug and disabled work distribution untill they fix it.

They have found a bug and disabled work distribution until they fix it.

Yes, they did that earlier. Since then, they have released an updated app, and new work. Now we are out of work.

Ah, they just added more work. Nice!

---

Reno, NV
Team: SETI.USA

Hello,
As with all beta applications, some hiccups happened, notably two major ones, credits awarded were not in line with our previous work and multi gpu were not working correctly.
The good news is: We just fixed those 2 issues and now you are able to run the beta workunits as intended.
Current GPU unit will reward 3333 credits (fixed) which is roughly equivalent to our previous jobs for Kaktwoos and in line with the current rewards for onechunk.

As a side note, we are open to discussion about our current rules with credits and we are also making sure to stay in line with the grcpool guidelines.

Cheers,
Neil