Message boards : Wish list : reasonable deadlines
Author | Message |
---|---|
so i downloaded the client and got my first work unit yesterday. it has 75 hours to go and a deadline in 4 days. now you see, the only way i could meet that deadline on the default settings - that is, boinc only running when i'm not using my computer - is if i left my computer running 24/7 and never used it. | |
ID: 21728 | Rating: 0 | rate: / Reply Quote | |
so i downloaded the client and got my first work unit yesterday. it has 75 hours to go and a deadline in 4 days. now you see, the only way i could meet that deadline on the default settings - that is, boinc only running when i'm not using my computer - is if i left my computer running 24/7 and never used it. You should unselect the "ACEMD for long runs (8-12 hours on fastest GPU)" in your profile under the GPUGRID preferences. i'm sorry but my computer is not a dedicated compute node for you. i have to pay power builds and i did not purchase my graphics card just for gpugrid.net. there is no way in hell mu machine will ever make that deadline, and if the deadlines continue to be like that, there is no way in hell my computer will ever make ANY of them. so what, does all the work get thrown away then? that sounds downright retarded. In this situation, maybe GPUGRID is not the right project for you. PrimeGrid, SETI, MilkyWay@home has shorter WUs. GPURID is intended particularly for cruncher enthusiasts. deadlines should be more on the order of months. their only purpose is to be able to reassign a work task on the very rare chance that one gets orphaned, by say someoen reformatting their hard drive or something like that. Every GPUGRID simulation is split into 10-20 workunits. If the deadline was on the order of months, a simulation would take a year (or more) to complete in worst case. This is too long time for the research they do. deadlines DO NOT effect the rate of processing results, with the exception of they are too short, in which case they an enormous effect on it - it can bring it to a practical standstill in short order. While you are right about the rate, in this project a simulation cannot be continued without the result of a wu. So the majority of the simulations would get stuck on slow hosts if the deadline was long. so make the deadline months out as they should be and which your effective computing power instantly jump an order of magnitude at no cost to anyone. This project's aim is not just maximizing computing power, this project is also about quick completion of every simulation they do. BTW your workunits are failing, maybe your screensaver, or your power saving should be turned off. | |
ID: 21730 | Rating: 0 | rate: / Reply Quote | |
kevin, | |
ID: 21731 | Rating: 0 | rate: / Reply Quote | |
Also, there is a 50% credit bonus for returning a result with 12 hours. Actually, the time limit for 50% bonus is 24 hours. | |
ID: 21733 | Rating: 0 | rate: / Reply Quote | |
Reading this post made me aware about: You should unselect the "ACEMD for long runs (8-12 hours on fastest GPU)" in your profile I have a Laptop with a NVIDIA NVS 4200M card. This card is unsuitable for the long runs and a deadline of 5 days. Unfortunately I learned the hard way, i.e., after 3 days of computing I made 10%. Maybe the GPUGRID application or scheduler can take the particular graphic card on the computer in question into consideration when deploying workloads. Thank you. | |
ID: 25779 | Rating: 0 | rate: / Reply Quote | |
People can read the FAQ - Recommended GPUs for GPUGrid crunching | |
ID: 25935 | Rating: 0 | rate: / Reply Quote | |
Message boards : Wish list : reasonable deadlines