RAM, runtime and credits
Message boards :
Questions/Problems/Bugs :
RAM, runtime and credits
Message board moderation
Author | Message |
---|---|
Send message Joined: 6 Sep 09 Posts: 7 Credit: 1,308,528 RAC: 0 |
Hello! Since quite some days all 14e-tasks I get do need more RAM than stated in the settings, some almost need 700 MB per task! This gave me a bit of trouble on one of my hosts (with an AMD FX 8320) in the last two days since another program also had heavy RAM usage and my 16 GB memory was almost full, resulting in an endless harddisk access storm and a slowed down system... I suggest you update the requirements for this if it's going to go permanently over 500 MB per task! Also, since I crunched all three types of WUs in the last time and observed how they were doing I have a question about it: Why are credits given so inconsequently? 14e runtimes can vary from around 6000 to 10000 secs (currently), but give only 36 credits (regardless of runtime). 15e runtimes can vary from around 4000 to 6000 secs and give 44 credits. 16e runtimes can vary from around 5000 to 7000 secs and give 130 credits. These runtimes are taken from my two hosts (AMD FX 8350 and AMD FX 8320). It's strange to me that the 8350, which is faster than the 8320, sometimes needs longer to complete a task than the 8320, although I did simultaneously crunched only one of the apps. Given the above mentioned varied runtimes it surely would more benefit the users when CreditNew would come to use here than this fixed credit system, especially on 14e you get way too less credits on long tasks! Life is Science, and Science rules. To the universe and beyond Proud member of BOINC@Heidelberg My BOINC-Stats |
Send message Joined: 26 Sep 09 Posts: 218 Credit: 22,841,893 RAC: 220 |
The only thing we can do is change the specs at NFS@Home preferences which I've already notified the admin to do so. We won't change the credit system since it is related to the difficulty of the number you're sieving. |
Send message Joined: 6 Sep 09 Posts: 7 Credit: 1,308,528 RAC: 0 |
We won't change the credit system since it is related to the difficulty of the number you're sieving. Well, this is not really logical to me. How do you define difficulty? If a 14e task needs twice as longer as an 16e the processor had to calculate more, so that was rather difficult than - but due to the fixed credits it gets punished the longer a tasks needs... On a side note, I just have 14e tasks which need almost 900 MB RAM (!) and are "only" at 28% with 1h 20 min. Talk about difficulty here... Life is Science, and Science rules. To the universe and beyond Proud member of BOINC@Heidelberg My BOINC-Stats |
Send message Joined: 5 Sep 09 Posts: 18 Credit: 63,574 RAC: 0 |
14e C231_134_79 snfs 257 lbp 32 20-400 M 680.05 MB -c 16000 -f 104976000 2h 42min 15e F1327 snfs 276 lpb 32 50-500 M 552.37 MB -c 4000 -f 109520000 1h 6min It ' s likely that C231_134_79 should use -c 8000 but not -c 16000, that is why it takes 2 hours but not 1 hour. In principle, F1327 should use 16e, C231_134_79 should use 15e. NFS@Home ' s main purpose is focus on Cunningham numbers, (a^n + 1, a^n - 1). 16e is full, so F1327 use 15e. 14e is empty, so C231_134_79 use 14e. |
Send message Joined: 23 May 11 Posts: 23 Credit: 25,231,022 RAC: 47,513 |
We want people to run the 15e and 16e siever for the (leading edge) big factorizations, the 14e is not really that interesting. Also the 14e requires a lot of effort/time from the post-processors people (relatively speaking). The extra credits are for the higher memory use and hopefully bias people towards the 16e. |
Send message Joined: 30 Dec 13 Posts: 5 Credit: 2,162,810 RAC: 75 |
The only thing we can do is change the specs at NFS@Home preferences which I've already notified the admin to do so. A year on and the NFS@Home Preferences section on the website still says this: lasieved - app for RSALS subproject, uses less than 0.5 GB memory: I guess the admin is just too busy ensuring there is some available disk space ;-) regards Tim |
Send message Joined: 1 Aug 19 Posts: 3 Credit: 5,071,518 RAC: 0 |
New to this project but long time BOINC... VERY confused by the credits. Results from one host: 14e - 4,945 seconds, 36 credits 15e - 7,243 seconds, 44 credits 16e - 4,956 seconds, 130 credits This is seriously broken... I have no major problem with the 14 being "low" or the 16 being "high" at around the same execution times (given you're trying to get people to do 16), but the 15 is absurd. 50% more work for the lower-end credit? At the very least, it should be >130, or maybe 100 if you REALLY want people to only do 16. I've also had RAM limit issues, but (having read the forum before signing up) I expected those, and can compensate. |
Send message Joined: 17 Mar 15 Posts: 4 Credit: 5,053,792 RAC: 4,083 |
The only thing we can do is change the specs at NFS@Home preferences which I've already notified the admin to do so. lasievee now allocates more than 1.5GB |
Send message Joined: 17 Mar 15 Posts: 4 Credit: 5,053,792 RAC: 4,083 |
lasieved now allocates more than 0.8GB and has multiplied run-time - https://escatter11.fullerton.edu/nfs/workunit.php?wuid=180680592 |