Welcome to MilkyWay@home

High frequency of "Too many results (may be nondeterministic) Too many total results"

Message boards : Number crunching : High frequency of "Too many results (may be nondeterministic) Too many total results"
Message board moderation

To post messages, you must log in.

AuthorMessage
Anthony Ayiomamitis

Send message
Joined: 25 Dec 14
Posts: 8
Credit: 1,007,280
RAC: 0
Message 62907 - Posted: 29 Dec 2014, 19:16:37 UTC

Four every four validated results, I am getting one invalid result due to "Too many results (may be nondeterministic) Too many total results" and which represents an approximate invalid rate of 20%.

Are others experiencing similar invalidated rates and consequent loss of credits since these invalid results are not credited.
ID: 62907 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3315
Credit: 519,950,748
RAC: 21,733
Message 62910 - Posted: 30 Dec 2014, 11:47:51 UTC - in response to Message 62907.  
Last modified: 30 Dec 2014, 11:58:01 UTC

Four every four validated results, I am getting one invalid result due to "Too many results (may be nondeterministic) Too many total results" and which represents an approximate invalid rate of 20%.

Are others experiencing similar invalidated rates and consequent loss of credits since these invalid results are not credited.


First stop running the n-body units on your pc, they are doing this:
Completed, validation inconclusive 5,258.36 9,181.88 pending MilkyWay@Home N-Body Simulation v1.46 (mt)

while the regular units are doing this:
Completed and validated 7,462.27 6,429.66 106.88 MilkyWay@Home v1.00

and the modified fit units are doing this:
Completed and validated 1,756.22 1,476.64 26.74 Milkyway@Home Separation (Modified Fit) v1.36

The n-body units use ALL of your cpu cores to process each workunit and it appears your pc can't handle that. That could be because your pc also does other things at the same time or for some other reason. As you can see from the credits awarded the n-body units are giving you 65.64 credits for 5300 seconds of work, while the regular units are giving you 106.88 credits for 7462 seconds of work. BUT more of your n-body units are failing than either the mt units or the regular units, so your time is spent more wisely NOT doing the n-body units. You can change what kind of units you are getting by going into Your Account on the webpage, then clicking this line "Preferences for this project MilkyWay@Home preferences" and then 'edit your preferences' and then unchecking the n-body units in this section:
Run only the selected applications
[]MilkyWay@Home
[]MilkyWay@Home N-Body Simulation
[]Milkyway@Home Separation
[]Milkyway@Home Separation (Modified Fit)

Make sure you click 'Save' after you are done editing or nothing will be changed. The next time your pc contacts the website to return or get more workunits it will stop getting the n-body units until you recheck that box again. You could abort any n-body units you currently have on your pc, which should force the communication quicker, and then you could spend your time crunching units that have a better likelihood of validating for you.

As for the "too many results" question it appears those units were just bad units, if you look at each unit under the 'workunit' column you will see this:
max # of error/total/success tasks 3, 9, 6
errors Too many results (may be nondeterministic) Too many total results

The 3 means 'max # of errors', the 6 means 'total' and the 9 means 'successful tasks'. You can see that there re 9 total pc's those units were sent out to in an effort to validate it but it just wasn't successful. It happens, some units just can't be crunched due to something in the workunits data itself. Obviously the project tries to not let that happen, but the only way to 100% ensure that is for them to crunch the units themselves, and then they don't need us. The one thing you might be able to do is to lower your cache size to a day or so, that way you aren't the LAST person to return the unit, but even if you are the first person it still doesn't guarantee it will work.
ID: 62910 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Anthony Ayiomamitis

Send message
Joined: 25 Dec 14
Posts: 8
Credit: 1,007,280
RAC: 0
Message 62911 - Posted: 30 Dec 2014, 12:38:06 UTC - in response to Message 62910.  

Many thanks for the advice and feedback.

I will modify my preferences as you suggest!
ID: 62911 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Anthony Ayiomamitis

Send message
Joined: 25 Dec 14
Posts: 8
Credit: 1,007,280
RAC: 0
Message 62912 - Posted: 30 Dec 2014, 12:47:21 UTC

I have another duo core machine arriving this week and which I plan to dedicate to MilkyWay@Home.

Thanks again for advice!
ID: 62912 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3315
Credit: 519,950,748
RAC: 21,733
Message 62918 - Posted: 31 Dec 2014, 12:18:53 UTC - in response to Message 62912.  

I have another duo core machine arriving this week and which I plan to dedicate to MilkyWay@Home.

Thanks again for advice!


Anytime, there are several of us who will help if you just ask.
ID: 62918 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Anthony Ayiomamitis

Send message
Joined: 25 Dec 14
Posts: 8
Credit: 1,007,280
RAC: 0
Message 62929 - Posted: 2 Jan 2015, 10:45:11 UTC

Mike et al,

Happy New Year!

Following the suggestion above, I am delighted at the crunching which is taking place at the moment.

I have a couple of questions for the group. I have two laptops with the same processor (Intel Duo Core 2.67 GHz P9600) but there is a notable difference in the "Measured Integer Speed" and in contrast to the "Measured Floating Point Speed" which is virtually identical between the two machines.

The machine with the slower integer speed (5542.79 million ops/sec) is running Windows Vista 32-bit with 2 Gb RAM whereas the machine with the faster integer speed (8323.87 million ops/sec) has Windows 7 64-bit with 4 Gb RAM.

Question 1: is the difference due to operating system and/or RAM? I do plan to clone the hard disk from my faster machine anyway for use in the slower machine but the intentions were originally different (to save reinstallation of various bits of other software)

Question 2: for the purposes of the MW project, I would assume the floating point speed is what really matters and which is virtually identical between the two machines. Is this correct?

Question 3: does the graphics card play a role in the measured integer speed (NVIDIA Quadro 160MVS vs Intel Graphics card with accelerator)?

Thanks in advance for the invaluable assistance!
ID: 62929 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 8 May 09
Posts: 3315
Credit: 519,950,748
RAC: 21,733
Message 62930 - Posted: 2 Jan 2015, 11:59:56 UTC - in response to Message 62929.  

Mike et al,

Happy New Year!

Following the suggestion above, I am delighted at the crunching which is taking place at the moment.

I have a couple of questions for the group. I have two laptops with the same processor (Intel Duo Core 2.67 GHz P9600) but there is a notable difference in the "Measured Integer Speed" and in contrast to the "Measured Floating Point Speed" which is virtually identical between the two machines.

The machine with the slower integer speed (5542.79 million ops/sec) is running Windows Vista 32-bit with 2 Gb RAM whereas the machine with the faster integer speed (8323.87 million ops/sec) has Windows 7 64-bit with 4 Gb RAM.

Question 1: is the difference due to operating system and/or RAM? I do plan to clone the hard disk from my faster machine anyway for use in the slower machine but the intentions were originally different (to save reinstallation of various bits of other software)


Yes. Software that is running while the crunching is going on will slow down Boinc, but of course not when that other software stops running. 64 bit machines use memory 'better' than 32bit machines, and Win7 is 'better' than Vista.

Question 2: for the purposes of the MW project, I would assume the floating point speed is what really matters and which is virtually identical between the two machines. Is this correct?


I don't know

Question 3: does the graphics card play a role in the measured integer speed (NVIDIA Quadro 160MVS vs Intel Graphics card with accelerator)?


No not unless you are using it for crunching, or have something using it in the background while the 'benchmarks' are running.

Thanks in advance for the invaluable assistance!


You are welcome.
ID: 62930 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : High frequency of "Too many results (may be nondeterministic) Too many total results"

©2024 Astroinformatics Group