I'm not actually interested in gaining credits (I cannot buy anything with them), but I see many people are.

So: if I run one 1000 events ATLAS task with 28 cpus, it can end in about 1 hour (elaboration time), and I get about 500 credits for it. Every day, that's about 12000 credits.

If I run 10 task at the same time, 3 cores each (total 30 cores instead of 28, who cares), they will end in about 9 hours. I get 4500 credits for each task, so in total 45000 task in 9 hours. That's more than 100000 credits a day. Almost ten times more.

And the second solution looks less efficient: the CPU time used by a task is about 90000 second with 28 cores, and more than 95000 with three.
What's the logic in it? I mean: a bit of variation is IMHO negligible... but TEN times?
--
Bye, Lem

I'm running 1000-event Atlas tasks too, but mine are earning about 3,000 credits each.

---

Am I the only one getting a gateway timeout on the server status page? Everything else seems to be loading properly (with some latency, as usual nowadays). The dev server status page also loads just fine. Tried different browser, different laptop/phone, same result.

Oops, poste in the wrong thread, sorry.

---

nope, we get it all ;)

---

---------- 24/7 Crunching since 2011 -----------
DSKAG Austria: http://www.dskag.at

In reply to Lem Novantotto's message of 22 Nov 2025:

What's the logic in it? I mean: a bit of variation is IMHO negligible... but TEN times?

Could it simply be that:
1. ATLAS project uses CreditNew – which is basically a random number generator, additionally programmed to generate the lowest possible numbers?
2. The Credit system looks at the "Run time" column, when it should be looking at the "CPU time"?

P.S. Maybe this was not the original goal of your message, but I suspect all the credit farmers will now use this hint and start running ATLAS tasks strictly on a single CPU core. :)

---

LHC project uses CreditNew this has been discussed before.

I think you are also correct about the the numbers used, in calc are Run time.