Message boards :
Number crunching :
very weak credit
Message board moderation
Author | Message |
---|---|
Send message Joined: 18 Aug 11 Posts: 10 Credit: 33,449,386 RAC: 1 |
with ubuntu AuthenticAMD 16 Dec 2019, 5:34:52 UTC 16 Dec 2019, 7:04:26 UTC Completed and validated 5,352.89 5,158.54 5.03 Get Decic Fields v3.00 (default) 16 Dec 2019, 5:34:52 UTC 16 Dec 2019, 6:35:47 UTC Completed and validated 3,634.37 3,519.06 3.33 Get Decic Fields v3.00 (default) 15 Dec 2019, 15:00:30 UTC 15 Dec 2019, 17:03:01 UTC Completed and validated 7,330.34 7,175.58 5.36 Get Decic Fields v3.00 (default) i have 44 wu like that. is this normal? thank you. |
Send message Joined: 8 Jul 11 Posts: 1346 Credit: 552,502,196 RAC: 657,688 |
That credit does look too low, by about a factor of 10. I thought maybe your benchmarks were off, so I looked at them. Your GFLOPS look reasonable but the GIOPS (Giga integer operations per second) look very high, which is actually the opposite of what I expected. So I am not sure what is going on here, and will need to think on it some more. This could be a side effect of "Credit New". I noticed that this machine has a high total credit but low RAC, meaning it's been away for a while. While you were away the credit system was changed to Credit New, so maybe that is contributing to the problem. I would try recomputing benchmarks from the client and resetting stats. |
Send message Joined: 18 Aug 11 Posts: 10 Credit: 33,449,386 RAC: 1 |
That credit does look too low, by about a factor of 10. very well I await a solution thank you |
Send message Joined: 8 Jul 11 Posts: 1346 Credit: 552,502,196 RAC: 657,688 |
That credit does look too low, by about a factor of 10. I noticed your credit per hour is slowly going up. So I have a guess as to what is happening. Normally we see the opposite effect when a user first connects - credit starts high and slowly comes down - this is because the fpops estimate is higher than it should be and this is the initial guess that Credit New uses. In your case, Credit New is starting from older stats from many months ago when run times were much quicker. So basically, in your case, Credit New is starting from an under-estimate, hence you get less credit. It will just take some time for Credit New to stabilize. If you don't want to wait for the credit to stabilize, you should reset your stats. I know it can be done, but I forget offhand how to do it. One way is to reset the project but this will also abort any currently running tasks, so be careful with that method. Hopefully someone else can chime in with a better method (Richard, are you still lurking around out there?). |
Send message Joined: 18 Aug 11 Posts: 10 Credit: 33,449,386 RAC: 1 |
good I start crunch again with this PC and over time the situation will recover? |
Send message Joined: 8 Jul 11 Posts: 1346 Credit: 552,502,196 RAC: 657,688 |
good Yes, but it could take a while. An instant fix would be to reset the project (If my hypothesis is true). |
Send message Joined: 18 Aug 11 Posts: 10 Credit: 33,449,386 RAC: 1 |
well I put 2 PC under ubuntu for 1 week. and I hope for rapid stabilization. |
Send message Joined: 3 Sep 11 Posts: 30 Credit: 10,789,219 RAC: 12,633 |
This credit issue still has not changed, I have recently come back and I am only getting 6 cr/h or less. 12 credits for more than 2 hours work is very much on the poor side. I ran a few and then aborted the rest as it was not worth the effort, credit wise. Sorry but I haven't stayed very long this time round, but I am just not seeing much of a reward to run the project. Conan |
Send message Joined: 8 Jul 11 Posts: 1346 Credit: 552,502,196 RAC: 657,688 |
This credit issue still has not changed, I have recently come back and I am only getting 6 cr/h or less. 12 credits for more than 2 hours work is very much on the poor side. I have been meaning to take care of the credit issue. I will do that as soon as the Pentathlon is over. I am considering using the credit per wu option, instead of creditNew. Does anyone know what a good credit/hour is on a single threaded CPU core (at 3.5 to 4 GHz)? |
Send message Joined: 28 Oct 11 Posts: 180 Credit: 259,702,441 RAC: 215,201 |
I am considering using the credit per wu option, instead of creditNew. Does anyone know what a good credit/hour is on a single threaded CPU core (at 3.5 to 4 GHz)?Let's be careful to research this properly - some people have, in the past, provided what must have been deliberate over-generous examples. I would personally nominate Einstein@Home as a well resourced, long-standing, thoughtful project that uses fixed credit and makes a conscientious effort to honour the cobblestone definition. I'll try to find some figures tomorrow. |
Send message Joined: 8 Jul 11 Posts: 1346 Credit: 552,502,196 RAC: 657,688 |
I am considering using the credit per wu option, instead of creditNew. Does anyone know what a good credit/hour is on a single threaded CPU core (at 3.5 to 4 GHz)?Let's be careful to research this properly - some people have, in the past, provided what must have been deliberate over-generous examples. No worries, it's at least 2 weeks away. And I've gotten better at reducing variance in run times. So I could base credit on the average run time (I run 100 cases offline for every batch to get a good estimate of the average). |
Send message Joined: 16 May 15 Posts: 10 Credit: 15,994,451 RAC: 0 |
Look to the PrimeGrid sieve sub-projects for a fair benchmark, both CPU and GPU. |
Send message Joined: 8 Jul 11 Posts: 1346 Credit: 552,502,196 RAC: 657,688 |
I am considering using the credit per wu option, instead of creditNew. Does anyone know what a good credit/hour is on a single threaded CPU core (at 3.5 to 4 GHz)?Let's be careful to research this properly - some people have, in the past, provided what must have been deliberate over-generous examples. I can use the credit that CreditNew computes to give me a good idea what I should use for fixed credit. On my one machine the credit fluctuates between 45 and 50 credits/hour. I will use the higher end of the range for the fixed credit. We can see how that works for now and then adjust accordingly. |