Hello all,
Petri mentioned during WoW that his 1080s starve for a few hours during the weekly maintenance with his special app.
Assuming the daily output will be very similar (his i7 12core + 4x GTX 1080 = ~180K RAC), I've done some quick calculations.
Based on those #s, a new VR-ready rig with a powerful i7 and one GTX 1080 will get ~50K/day, which is ~500tasks/day.

I'm wondering if there's any talk with the project staff about raising the 100task limit/gpu by the time Petri's app-in-dev is ready for S@h Beta testing.
Otherwise, once it is available thru Lunatics for S@h main, some might head to greener Boinc pastures when they find out that their new gaming and/or VR rig will starve during the weekly S@h maintenance.

If the 100tasks/gpu limit isn't raised by then, would it create a "situation" if I released a script so that we don't loose those who sort of "complain" about the weekly maintenance?
The script would allow the transfer of tasks from the CPU queue to the GPU queue and I already have a working Win7+ proof-of-concept script for Cuda50 only.
It allows me to stash up to the Boinc Client limit of 1,000tasks/project (or the 10 day limit, which ever comes first).
On my ~23k/day rigs, that's ~4 days worth of cache.

So before I spend any time in turning my proof-of-concept into a working prototype (ready for a restricted beta-type of release), I am wondering:
Will the pros outweight the cons?
...if such a script was distributed only to those with "power rigs" (daily output >25k/day) who would only be able to stash <4day's worth of work.
It certainly would give Boinc power users some peace of mind of not having to check their rigs once or twice per day.

I know for some it can be a touchy subject and I don't know much about the history other than past S@h server crashes that haven't happened in many years.
From my perspective, this script would likely put less load on the servers than raising the limit for all GPUs.

If the discussion gets lively, please keep it civil.
Cheers,
Rob :-)

---

Hello all,
Petri mentioned during WoW that his 1080s starve for a few hours during the weekly maintenance with his special app.

The script would allow the transfer of tasks from the CPU queue to the GPU queue and I already have a working Win7+ proof-of-concept script for Cuda50 only.
It allows me to stash up to the Boinc Client limit of 1,000tasks/project (or the 10 day limit, which ever comes first).
On my ~23k/day rigs, that's ~4 days worth of cache.

So before I spend any time in turning my proof-of-concept into a working prototype (ready for a restricted beta-type of release), I am wondering:
Will the pros outweight the cons?
...if such a script was distributed only to those with "power rigs" (daily output >25k/day) who would only be able to stash <4day's worth of work.
It certainly would give Boinc power users some peace of mind of not having to check their rigs once or twice per day.

I know for some it can be a touchy subject and I don't know much about the history other than past S@h server crashes that haven't happened in many years.
From my perspective, this script would likely put less load on the servers than raising the limit for all GPUs.

If the discussion gets lively, please keep it civil.
Cheers,
Rob :-)

. . I think it likely there will be reservations about such a script. And there will be those who are dead set against it. But what is life without a little controversy. :)

. . But I can see how it would be nice knowing that there is enough work in your caches to sustain productivity during a prolonged network hiatus.

Stephen

.