Hi everyone,

Eternity II is my first project, and I'm amazed by what you've made of it. I keep working on it every day to answer your requests, fix problems and improve the apps.

Thank you to all of you who give your computers' time to the search, and to everyone who takes a moment to send feedback. Your reports have helped me find real problems, often very quickly.

Every piece of feedback is noted, even when the answer is "not now". All your requests and ideas are on one list that I'll go through after the sprint. I can't promise that everything will be done, but nothing is forgotten.

During the Formula BOINC sprint, though, making changes is complicated. A new app version or a server change could disturb the computers of everyone taking part. So for now I'm mostly keeping things stable, and the bigger improvements will come afterwards.

Thanks for your patience, and enjoy the sprint!

Benjamin

As promised, I went through the list after the sprint. Done so far:
• Up to 8 GPU tasks per card, and you can choose your applications in the project preferences.
• More GPU models accepted (TITAN Xp, Tesla T4/V100, T500, Quadro RTX 6000, several Laptop GPUs).
• GPU tasks that dragged for hours near 98%: fixed in version 2.81.
• Bigger tasks for the RTX 4090/5090, shorter ones for the RTX 5070/5060 Ti, and a new "next bigger size" option.
• The "doesn't support computers of type" message and the 24-hour delays: fixed.
• Site security: thanks to the volunteer who reported it privately, three points are fixed and the site now switches browsers to https.
Still on the list, without dates: the AMD (OpenCL) version, a macOS version, a beta-tasks option, and a timestamp on the progress lines. Thanks again for all your feedback!

My 4080 and my 5070 are happily crunching 131.000v version instead of the 65.000v but strangely it doesn't take them much longer to finish them. The name of the task suggest it may be really a 65.000v task. "Eternity II NVIDIA GPU 272 (131072v tickets) v2.81 (cuda_e2_pubv_s2_ada_lin_65536v_57be346b_up) x86_64-pc-linux-gnu". Wuprop doesn't detect them either although version 131.000v is a known app there.

I get more credit, though!

In reply to Dargo75's message of 6 Nov 2026:

My 4080 and my 5070 are happily crunching 131.000v version instead of the 65.000v but strangely it doesn't take them much longer to finish them. The name of the task suggest it may be really a 65.000v task. "Eternity II NVIDIA GPU 272 (131072v tickets) v2.81 (cuda_e2_pubv_s2_ada_lin_65536v_57be346b_up) x86_64-pc-linux-gnu". Wuprop doesn't detect them either although version 131.000v is a known app there.

I get more credit, though!

@Dargo75: they are bigger tasks, and your credit shows it: credit is paid per ticket, and on your 5070 a full 131072v task gives 175,672 credits, exactly twice the 87,842 of a 65536v one. The "65536v … _up" in the version name is the class of your card (65,536) plus "_up", the "next bigger size" option you ticked.

About the time: be careful when comparing. When BOINC runs two GPU tasks at once on the same card, each one takes longer, and on 5 November your 5070 had two 65536v tasks running together. To compare fairly, look at the time of a task that ran alone on the card.

About WUProp: those tasks belong to the application eternity_gpu_272_131072v, like on every other host. I don't know how WUProp matches tasks on its side, so if it keeps missing them, their forum is probably the best place to ask.

Interesting. I just checked the runtimes of the last 15 tasks of my 5070 of the 65.000v version and you are right, the times are very inconsistent. They range from 2.100 - 13.900 seconds. It also finished 5 tasks of the 131.000v version ranging from 6.400 - 10.000 seconds. So if there is a difference it is not immediatly obvious. :-) My GPUs only crunch one task at a time because your tasks fill them up nicely so no need for that. No idea how you saw two tasks crunched concurrently in the results. Bottom line: I will aim for the biggest version from now on, thanks for offering that option! Didn't check my weaker GPUs yet. Wuprop sensed my tasks now, so all is good!

In reply to Dargo75's message of 6 Nov 2026:

Interesting. I just checked the runtimes of the last 15 tasks of my 5070 of the 65.000v version and you are right, the times are very inconsistent. They range from 2.100 - 13.900 seconds. It also finished 5 tasks of the 131.000v version ranging from 6.400 - 10.000 seconds. So if there is a difference it is not immediatly obvious. :-) My GPUs only crunch one task at a time because your tasks fill them up nicely so no need for that. No idea how you saw two tasks crunched concurrently in the results. Bottom line: I will aim for the biggest version from now on, thanks for offering that option! Didn't check my weaker GPUs yet. Wuprop sensed my tasks now, so all is good!

@Dargo75: thanks for checking, and you're right about the "two at a time": my mistake. I had looked at tasks whose download and return times overlapped, but that just means the second one was waiting in your queue, not running on the card. Sorry for the confusion.

The spread you see (2,100 to 13,900 s for the same size) is normal for this project. Every ticket is a different starting position: some die out after a few seconds, others go very deep, so two tasks with the same number of tickets can need very different amounts of work. Credit is paid per ticket, so it stays fair whatever the run time.

Going for the biggest size is a good choice for your cards: each task has one ending phase, where the last tickets slow down, and a bigger task has fewer of those for the same number of tickets.

Glad WUProp sees them now!