I finally retired a 13 year old GPU that had not crunched a WU in many years. I have a new one (mid-level?), but cannot find any GPU work units.
- Einstein no WUs
- GPU Grid no WUs
- Milky Way no WUs
- Prime Grid yes this has many WUs, but sorry, that doesn't "trip my trigger."
- Asteroids used to have GPU WUs but no longer?

Am I missing anything or have any of these projects started to issue work?
Is it time to rejoin Folding@Home?
Thanks! Ni!

Of that list, I'd consider Einstein to be the most reliable. I'm running 6 tasks at the moment - all using NVidia GPUs, some under Windows, others under Linux.

Can you be more specific about your hardware and software? Projects tend to like mostly modern GPUs, with plenty of GPU memory. And you do need to study each project's website, to make sure you have checked all the appropriate options for running GPU work.

OK this system includes:
NVIDIA GeForce RTX 3070 with 8G GDDR6 memory
AMD Ryzen 7 3700X CPU with 8 cores / 16 threads
16 G DDR4-3200 memory clocked at 1600 MHz.
Thanks for any suggestions...

Ok both Einstein and Asteroids have NVIDIA tasks available ... I am getting on both with my little GTX 1660

Assumptions you have enabled NVIDIA GPU work on your Project Profile for both projects

And ... that you have looked at the Application lists for both programs and selected applications using Nvidia GPUs'

Best of Luck

Respectfully
Bill F

---

In October 1969 I took an oath to support and defend the Constitution of the United States against all enemies, foreign and domestic;
There was no expiration date.

In reply to Sir LanDroid's message of 15 Mar 2026:

Am I missing anything or have any of these projects started to issue work?

The new AI project, Axiom, has gpu work

Tweaked some settings and I'm getting closer on Einstein. However it says:
Not requesting tasks: don't need (CPU: blocked by project preferences; NVIDIA GPU: job cache full)
I see no WUs from Einstein even though "job cache full." ???
I have it set to accept GPU, but not run CPU. Do I have to enable CPU in order to get GPU work?

Will look into Axiom...

Update: I received an Einstein WU for All-Sky Gravitational Wave search on 04 2.03 (GW-cuda). Thank you!
Now to check details for Axiom, GPUGrid, and Milkyway@Home.

Be aware that Axiom is not on the BOINC projects list, and not vetted for safety and scientific legitimacy. https://boinc.berkeley.edu/projects.php

Text below from: https://linustechtips.com/blogs/entry/2405-march-boinc-blog/

- New BOINC project: AXIOM DISTRIBUTED AI

https://axiom.heliex.net/
https://boincsynergy.ca/wiki/Axiom_Distributed_AI/

The project is still in development and aims to train a neural network. According to the site, "An LLM (Claude) serves as the principal investigator: designing experiments, deploying them to volunteer hardware, reviewing results, and awarding credit based on scientific quality." heliex.net is a GridCoin and CureCoin cryptocurrency exchange.

As I always say, before joining any project (especially new ones) read its web site and consider: Does the project clearly describe its goals? Are these goals important and beneficial? Do you trust the project to use proper security practices? Has it been vetted by the community? Who will own the results of the computation? I'm personally going to pass on this project, at least for now.

Many projects aren't on the list.

MW - CPU only
GPUGrid - No work ATM

GPU projects
E@H, NF, PG, A@H, Moo, Minecraft, Amicable (Ending soon), SRBase, Gerasim, Private GFN

In reply to Grumpy Swede's message of 16 Mar 2026:

The project is still in development and aims to train a neural network. According to the site, "An LLM (Claude) serves as the principal investigator: designing experiments, deploying them to volunteer hardware, reviewing results, and awarding credit based on scientific quality." heliex.net is a GridCoin and CureCoin cryptocurrency exchange.

From Axiom admin:

Not crypto. Every cycle of compute on Axiom goes to running scientific experiments — things like Kuramoto oscillator models, sandpile dynamics, contact processes, Langton ant variants. The workunit names reflect the actual experiments being run. You can read the experiment scripts directly at https://axiom.heliex.net/api/experiments.php and the scientific findings at https://axiom.heliex.net/api/findings.php.

Regarding the PoUC document makracz found — that was a concept paper I wrote exploring how a science-based credit system could connect to a token model in the future. It is not implemented and no token exists. Right now the project is 100% focused on the science platform: designing experiments, collecting results, and publishing findings. The first paper came out last week on reactive mode localization in ecological networks.

HeliEx (the exchange) and Axiom are separate codebases. Axiom volunteers are not mining anything — you are contributing to distributed scientific research.

Scientific details: Experiments

In reply to boboviz's message of 17 Mar 2026:

In reply to Grumpy Swede's message of 16 Mar 2026:

The project is still in development and aims to train a neural network. According to the site, "An LLM (Claude) serves as the principal investigator: designing experiments, deploying them to volunteer hardware, reviewing results, and awarding credit based on scientific quality." heliex.net is a GridCoin and CureCoin cryptocurrency exchange.

From Axiom admin:

Not crypto. Every cycle of compute on Axiom goes to running scientific experiments — things like Kuramoto oscillator models, sandpile dynamics, contact processes, Langton ant variants. The workunit names reflect the actual experiments being run. You can read the experiment scripts directly at https://axiom.heliex.net/api/experiments.php and the scientific findings at https://axiom.heliex.net/api/findings.php.

Regarding the PoUC document makracz found — that was a concept paper I wrote exploring how a science-based credit system could connect to a token model in the future. It is not implemented and no token exists. Right now the project is 100% focused on the science platform: designing experiments, collecting results, and publishing findings. The first paper came out last week on reactive mode localization in ecological networks.

HeliEx (the exchange) and Axiom are separate codebases. Axiom volunteers are not mining anything — you are contributing to distributed scientific research.

Scientific details: Experiments

Yeah, Yeah, and I'm Santa Claus. If they were using volunteers for mining, you really think they would admit it? That project is a big NO NO for me at least. Maybe I'll join in a year or two, when the project is really vetted by trusted people.

In reply to Grumpy Swede's message of 17 Mar 2026:

Scientific details: Experiments

Yeah, Yeah, and I'm Santa Claus. If they were using volunteers for mining, you really think they would admit it? That project is a big NO NO for me at least. Maybe I'll join in a year or two, when the project is really vetted by trusted people.

The mining project (GridCoin) said it clearly.
And they lost their time to describe every single simulation in details if they want only mining?
Are you interested in it?

And the Axiom code is open (unlike many historical boinc projects)

As a project Axiom appears to have used BOINC as a vague structure, then imposed its own set of validation and credit rules/systems - both of which are integral parts of BOINC. This is the only part of Axiom's code that has been made Open Source, the rest, the really interesting stuff, remains hidden behind a wall of secrecy. (btw. It is very common for projects not to publish their source code, so this is not a criticism per-se)

Looking at the nine current projects (aka live experiments) the top level subject descriptor shows one experiment to be "machine learning", six to be "science", and two"ecology". Then reading next level of description we find that two of the "science" experiments are actually "machine earning".
So, continuing to read, most of the experiment descriptors are all encompassed in lots of buzz-words with very little actual scientific meaning.

So, all in all - a nice idea, but they really do need to get their game in order to attract and keep users.

Given the level of inaccuracy in the level of inaccuracy in the top & second level descriptors for the experiments I would put this as an AI generated attempt at machine learning (best), or an outright con (worst).


(Having now read all the text of the descriptors I would classify ALL the current experiments as "machine learning", in other words every single one of these nine experiments is wrongly described.)