Why are the RTX Graphics Cards now really Poor at running this Project.

Author	Message
bcavnaugh Send message Joined: 26 Jan 17 Posts: 21 Credit: 272,313,122 RAC: 0	Message 1059 - Posted: 18 Jan 2019, 23:19:19 UTC Why are the RTX Graphics Cards now really Poor at running this Project. I went from less that 2 min to over 19 min. Crunching@EVGA The Number One Team in the BOINC Community. Folding@EVGA The Number One Team in the Folding@Home Community. ID: 1059 · Rating: 0 · rate: / Reply Quote

Sergei Chernykh Project administrator Project developer Send message Joined: 5 Jan 17 Posts: 513 Credit: 72,451,573 RAC: 0	Message 1060 - Posted: 18 Jan 2019, 23:22:04 UTC - in response to Message 1059. The bottleneck is prime sieving at this stage because it's done on CPU. The only solution is to run as many tasks per GPU as possible - reduce kernel size to fit more tasks. ID: 1060 · Rating: 0 · rate: / Reply Quote

bcavnaugh Send message Joined: 26 Jan 17 Posts: 21 Credit: 272,313,122 RAC: 0	Message 1062 - Posted: 19 Jan 2019, 4:09:03 UTC - in response to Message 1060. The bottleneck is prime sieving at this stage because it's done on CPU. The only solution is to run as many tasks per GPU as possible - reduce kernel size to fit more tasks. Thanks and good to know, but a little odd. Why are the tasks not just ran on the CPUs? Crunching@EVGA The Number One Team in the BOINC Community. Folding@EVGA The Number One Team in the Folding@Home Community. ID: 1062 · Rating: 0 · rate: / Reply Quote

Sergei Chernykh Project administrator Project developer Send message Joined: 5 Jan 17 Posts: 513 Credit: 72,451,573 RAC: 0	Message 1063 - Posted: 19 Jan 2019, 15:08:53 UTC - in response to Message 1062. Why are the tasks not just ran on the CPUs? Because they are GPU tasks? The actual search is still done on GPU, it just needs a list of prime numbers in given interval to proceed, and this is done on CPU. ID: 1063 · Rating: 0 · rate: / Reply Quote

Azmodes Send message Joined: 5 Mar 17 Posts: 20 Credit: 3,017,841,792 RAC: 0	Message 1064 - Posted: 20 Jan 2019, 13:53:43 UTC I tried 8 tasks on my 2080, each with a full CPU core allocated, but core load was barely over 50% IIRC. Adding more caused a BSOD, so I assume I should reduce the kernel size, but I can't experiment with that right now. Anyone have a working configuration they'd like to share? ID: 1064 · Rating: 0 · rate: / Reply Quote

Jozef J Send message Joined: 24 Jan 17 Posts: 20 Credit: 1,187,449,660 RAC: 398,776	Message 1070 - Posted: 22 Jan 2019, 0:13:10 UTC Hi, last time i run on 2080Ti , ,16 tasks with 21 kernel.. what alocated nearly 10 000 mb , no problem with run or stability.. but system was starting bit l"lagy" ... Todays is crunching on RTX card limitation .. gpu ram.. only 11 Gb.. but problem is also when you restart computer.. i try stoping project in boinc manager and after whole boinc m. but always I lost running task.. wich is problem now when lose was about 16 task at "once".. now i run tasks only on laptop 960M nvidia 4Gb i can try run more task on lower kernel just for test..becouse gpu ram limitation..but system will very lagy i think and imposible to do somethink on.. i got idee also run whole boinc from ram disc.. next days maybe.. to decrease that irq impact -lag but i will back to project on 100% at spring,now i need heat from run-) ID: 1070 · Rating: 0 · rate: / Reply Quote