This may now be outdated, but for now, see this:
https://forums.anandtech.com/threads/ab ... e.2504696/
Wow, there are tons of info in that thread, but it may be dated. I'll run some quick tests to see if there are any major changes.
It appears you need to optimize the task's kernel size. EDIT: optimizing caused immediate Computation failure on 2 machines
OCN has an extensive 51 page thread on Collazt. http://www.overclock.net/t/1414835/coll ... thread/450
There is also a thread on the Collatz forum on the subject: https://boinc.thesonntags.com/collatz/f ... hp?id=8#69
A quick test un-optimized:
RTX 2080Ti - 220.68s GPU - 7.41s CPU (AMD 3800X) - Linux, nvidia-smi gives a reading of 94% GPU usage - 37,750.87 points
GTX 1080 - 708.16s GPU - 10.07s - CPU (AMD 3800X) - Linux, percent of GPU usage unknown -35,981.74 points
GTX 1070 - 845.15s GPU - 0.41s CPU (i7-5820k) Windows, 97% GPU usage - 36,033.20 points
R9-280X - 1,663.66s GPU - 68.45s - CPU (FX8320E) Windows, 93% GPU usage - 35,894.69 points
Optimized:
GTX 1070 - crashed
R9 280X - crashed
Collatz Conjecture
- 10esseetony
- TAAT Member
- Reactions:
- Posts: 473
- Joined: Sun Sep 15, 2019 7:02 pm
Collatz Conjecture
Last edited by 10esseetony on Sun Apr 26, 2020 8:51 am, edited 1 time in total.
Re: Collatz Conjecture
I haven't run Collatz in a long time, though I remember it paying well.
Re: Collatz Conjecture
You're a prophet, Tony! Collatz is the Sprint!