Changing the cache to 0.2 did not get an immediate work fetch.
I'm sorry, but I wanted you to increase the setting by 0.2 days. So, for example, if it had been showing 0.3 days previously you would change that to 0.5. I also should have mentioned that because the change is on the website you will need to make your BOINC client contact the website to be informed immediately. You do this in BOINC Manager - Advanced view by clicking the Einstein project on the 'Projects' tab and clicking the 'update' button. This 'forces' your client to become aware of the change and it will then ask for extra work. I'm not sure these days what the default value for that setting was. Perhaps it was as low as 0.1 days. To protect against temporary project outages, it's a good idea to set that value even a bit higher, say 1.0 days, so that you will have several tasks waiting to start crunching even if the project can't immediately supply a new one when asked.
Quote:
Should I create a home profile and use that for my other computer so it works on both Archibo and Parkes?
Your other computer has a NVIDIA GT 720 GPU which is a lot less powerful than the AMD HD7850/7870. So yes, it would be a good idea to put each computer in a separate 'venue' or 'location' so that you can use different preference settings for each computer. I very much doubt your GT 720 would show any benefit from running concurrent GPU tasks but it perhaps will show a benefit from running the beta test app which uses a more up-to-date version of CUDA. Without worrying too much about the details, if you make a further change to the project preferences and tick the box to allow test applications to run, you will get a new app for your NVIDIA card which should run a bit faster, hopefully. It's certainly worth a try.
In summary, you could make this list of changes
* Add the 'home' venue so you have both 'default' and 'home' available.
* The existing 'default' preferences should be suitable for the AMD GPU so you should be able to leave them as they are.
* Edit the new 'home' preferences to be suitable for the NVIDIA (eg. GPU factor of 1 and allow test apps).
* Parkes (BRP6) tasks will be best for your GT 720 so continue to exclude Arecibo GPU tasks (BRP4G) for the 'home' venue.
* Change the venue (location) for your NVIDIA host from 'default' (---) to 'home'. Account page -> View computers -> Details page. Change and update the location near the bottom of the page.
* 'Update' the Einstein project on each machine so that each becomes aware of the changes.
Ask, if any of this is not clear.
Quote:
On another note, my page still doesn't show WCGrid stats.
It's still early days. Just give it time. You could double check that each project has precisely the same email address :-).
Ok, Did it all. Check up on me later or when you can to make sure it is running optimally. Please keep an eye on it. Appreciate your help as I know this took some time for you to type it all out.
Oh, both Einstein and WCGrid were on one email after I made the change yesterday, but I just changed them both to another identical email address to see if that makes a difference. Best Regards
Oh, both Einstein and WCGrid were on one email after I made the change yesterday, but I just changed them both to another identical email address to see if that makes a difference. Best Regards
You should use the same email address and username at ALL of the BOINC projects, those two factors are what's used to determine your account CPID. That's what is used to identify you across each of the projects.
Check up on me later or when you can to make sure it is running optimally.
I think you should be very pleased when you look at the latest Parkes (BRP6) results for your i7-2600K with the AMD Pitcairn series GPU. Notice that the results returned on Jan 6 or earlier were all taking around 40K seconds elapsed time, running single tasks. As from Jan 7, you are running 2 GPU tasks concurrently and the elapsed time has dropped to around 11K seconds, and you get 2 tasks done in that time. Effectively you've gone from around 40K secs per task to 5.5K secs - a speedup factor of between 7 to 8 times!! I believe the crucial factor was that you were running 8 CPU tasks on all 8 virtual cores and the change has just removed one CPU task and allowed the GPU to show its full potential.
Quote:
Appreciate your help as I know this took some time for you to type it all out.
You're most welcome! I hate to see good hardware not living up to its full crunching potential. I sincerely thank you for following the advice given.
Quote:
Oh, both Einstein and WCGrid were on one email after I made the change yesterday, but I just changed them both to another identical email address to see if that makes a difference.
I wish you hadn't done that :-). The upshot is that your account page here at Einstein now shows no other projects. That means that this new email address you have selected is NOT known at the other projects that were previously showing here. You HAVE to have the same email address at ALL projects if you want to see ALL your contributions on each different project's account page. That includes older projects that perhaps you are no longer currently active at. You should make sure your email address is the same there too. You don't need to download work. Projects can even be suspended. But you have to (manually) force contact on a machine (or group of machines) that have access to all projects so that the synchronising of the CPID can work.
Unfortunately it shows WCGrid with only 8 million points instead of the 57 million I actually have etc.
One possible reason that springs to mind - multiple accounts at WCG. Is there any possibility that your 57M points is spread over different 'accounts'? I have no idea how things work at WCG.
Another thought. Are there different sub-projects that perhaps have different stats update intervals and so you may have to wait until all of these are updated?
Another thing to try. Go to Rosetts and see if your WCG total shows differently there.
Failing all this, perhaps you should ask at WCG if there is any reason why your full score isn't showing on your account pages at other projects.
Quote:
But I'm happy. I prefer that computer is running more efficiently.
Have you taken a look at the Parkes PMPS XT (BRP6) tasks list on your other computer with the nvidia GT-720 GPU lately? Before the changes, task times were over 38K seconds. Briefly, tasks were being crunched 2x - you have 2 results around 74K secs. They were still with the old cuda32 app. Now that you have allowed beta test apps, you have some cuda55 tasks (running singly) and taking 30K secs. That's a nice improvement as well.
Quote:
And I really appreciate your efforts.
Thank you.
No worries, it was a pleasure to see the improvement for both machines but particularly for the AMD Pitcairn series GPU.
EDIT: One other thing you should do. Your GT-720 host has only a single 'in progress' task. You should increase the work cache size for the venue (location) it's in (say to 1.0 days) just like you have done for the other host. That way you can have a couple of tasks 'ready to run' in case of a temporary network outage.
Unfortunately it shows WCGrid with only 8 million points instead of the 57 million I actually have etc.
57 million according to whom? WCG does not use the standard BOINC cobblestone, they use their own point system because they have a legacy system that started issuing credit before they used BOINC clients. To convert to BOINC credit (as you would see on a stats site) you divide WCG points by 8, which yields... 8.1 million.
Wooh. According to whom? I'm sure you message didn't mean to come across like that. When us common folk log into WCGrid it says 57 million. I don't know anything about a legacy system or divide by eight and have never heard of any of that. All I know is what I see.
Thank you for explaining it.
RE: Changing the cache to
)
I'm sorry, but I wanted you to increase the setting by 0.2 days. So, for example, if it had been showing 0.3 days previously you would change that to 0.5. I also should have mentioned that because the change is on the website you will need to make your BOINC client contact the website to be informed immediately. You do this in BOINC Manager - Advanced view by clicking the Einstein project on the 'Projects' tab and clicking the 'update' button. This 'forces' your client to become aware of the change and it will then ask for extra work. I'm not sure these days what the default value for that setting was. Perhaps it was as low as 0.1 days. To protect against temporary project outages, it's a good idea to set that value even a bit higher, say 1.0 days, so that you will have several tasks waiting to start crunching even if the project can't immediately supply a new one when asked.
Your other computer has a NVIDIA GT 720 GPU which is a lot less powerful than the AMD HD7850/7870. So yes, it would be a good idea to put each computer in a separate 'venue' or 'location' so that you can use different preference settings for each computer. I very much doubt your GT 720 would show any benefit from running concurrent GPU tasks but it perhaps will show a benefit from running the beta test app which uses a more up-to-date version of CUDA. Without worrying too much about the details, if you make a further change to the project preferences and tick the box to allow test applications to run, you will get a new app for your NVIDIA card which should run a bit faster, hopefully. It's certainly worth a try.
In summary, you could make this list of changes
* The existing 'default' preferences should be suitable for the AMD GPU so you should be able to leave them as they are.
* Edit the new 'home' preferences to be suitable for the NVIDIA (eg. GPU factor of 1 and allow test apps).
* Parkes (BRP6) tasks will be best for your GT 720 so continue to exclude Arecibo GPU tasks (BRP4G) for the 'home' venue.
* Change the venue (location) for your NVIDIA host from 'default' (---) to 'home'. Account page -> View computers -> Details page. Change and update the location near the bottom of the page.
* 'Update' the Einstein project on each machine so that each becomes aware of the changes.
Ask, if any of this is not clear.
It's still early days. Just give it time. You could double check that each project has precisely the same email address :-).
Cheers,
Gary.
Ok, Did it all. Check up on
)
Ok, Did it all. Check up on me later or when you can to make sure it is running optimally. Please keep an eye on it. Appreciate your help as I know this took some time for you to type it all out.
Oh, both Einstein and WCGrid were on one email after I made the change yesterday, but I just changed them both to another identical email address to see if that makes a difference. Best Regards
RE: Oh, both Einstein and
)
You should use the same email address and username at ALL of the BOINC projects, those two factors are what's used to determine your account CPID. That's what is used to identify you across each of the projects.
Click Here to see My Detailed BOINC Stats
RE: Ok, Did it
)
That's great!! Well done!!
I think you should be very pleased when you look at the latest Parkes (BRP6) results for your i7-2600K with the AMD Pitcairn series GPU. Notice that the results returned on Jan 6 or earlier were all taking around 40K seconds elapsed time, running single tasks. As from Jan 7, you are running 2 GPU tasks concurrently and the elapsed time has dropped to around 11K seconds, and you get 2 tasks done in that time. Effectively you've gone from around 40K secs per task to 5.5K secs - a speedup factor of between 7 to 8 times!! I believe the crucial factor was that you were running 8 CPU tasks on all 8 virtual cores and the change has just removed one CPU task and allowed the GPU to show its full potential.
You're most welcome! I hate to see good hardware not living up to its full crunching potential. I sincerely thank you for following the advice given.
I wish you hadn't done that :-). The upshot is that your account page here at Einstein now shows no other projects. That means that this new email address you have selected is NOT known at the other projects that were previously showing here. You HAVE to have the same email address at ALL projects if you want to see ALL your contributions on each different project's account page. That includes older projects that perhaps you are no longer currently active at. You should make sure your email address is the same there too. You don't need to download work. Projects can even be suspended. But you have to (manually) force contact on a machine (or group of machines) that have access to all projects so that the synchronising of the CPID can work.
Cheers,
Gary.
I've just noticed that all
)
I've just noticed that all your projects are back, including WCGrid.
A satisfactory conclusion for all problems :-).
Cheers,
Gary.
Unfortunately it shows WCGrid
)
Unfortunately it shows WCGrid with only 8 million points instead of the 57 million I actually have etc.
But I'm happy. I prefer that computer is running more efficiently.
And I really appreciate your efforts.
Thank you.
RE: Unfortunately it shows
)
One possible reason that springs to mind - multiple accounts at WCG. Is there any possibility that your 57M points is spread over different 'accounts'? I have no idea how things work at WCG.
Another thought. Are there different sub-projects that perhaps have different stats update intervals and so you may have to wait until all of these are updated?
Another thing to try. Go to Rosetts and see if your WCG total shows differently there.
Failing all this, perhaps you should ask at WCG if there is any reason why your full score isn't showing on your account pages at other projects.
Have you taken a look at the Parkes PMPS XT (BRP6) tasks list on your other computer with the nvidia GT-720 GPU lately? Before the changes, task times were over 38K seconds. Briefly, tasks were being crunched 2x - you have 2 results around 74K secs. They were still with the old cuda32 app. Now that you have allowed beta test apps, you have some cuda55 tasks (running singly) and taking 30K secs. That's a nice improvement as well.
No worries, it was a pleasure to see the improvement for both machines but particularly for the AMD Pitcairn series GPU.
EDIT: One other thing you should do. Your GT-720 host has only a single 'in progress' task. You should increase the work cache size for the venue (location) it's in (say to 1.0 days) just like you have done for the other host. That way you can have a couple of tasks 'ready to run' in case of a temporary network outage.
Cheers,
Gary.
RE: Unfortunately it shows
)
57 million according to whom? WCG does not use the standard BOINC cobblestone, they use their own point system because they have a legacy system that started issuing credit before they used BOINC clients. To convert to BOINC credit (as you would see on a stats site) you divide WCG points by 8, which yields... 8.1 million.
Click Here to see My Detailed BOINC Stats
Wooh. According to whom?
)
Wooh. According to whom? I'm sure you message didn't mean to come across like that. When us common folk log into WCGrid it says 57 million. I don't know anything about a legacy system or divide by eight and have never heard of any of that. All I know is what I see.
Thank you for explaining it.