So I have been using 6.12.34 for quite a while and been happy with how it fetches work but it seems to suffer badly from memory leaks and will always be using 1GB+ after 24 hrs. I have upgraded to 7.0.28 which I know has a new scheduler/work fetch system but I just can't get it to work as I want.
On my system with 4 GPUs I can have a maximum of 160 WUs which is about 30 minutes work, so obviously I want to keep as near to this maximum as possible. BOINC 6.x would connect to the project quite frequently and hence always keep near the maximum WUs. BOINC 7.x only seems to connect when it has nearly run out and so never builds up to the 160 WUs cached because the most it ever gets back from the server in one go seems to be 24-25 WUs.
Is there any way I can get the behaviour I want in BOINC 7?
On the positive side the memory usage issues seem to be fixed in 7.x
/ 
Reply
Quote
So I have been using 6.12.34 for quite a while and been happy with how it fetches work but it seems to suffer badly from memory leaks and will always be using 1GB+ after 24 hrs. I have upgraded to 7.0.28 which I know has a new scheduler/work fetch system but I just can't get it to work as I want.
On my system with 4 GPUs I can have a maximum of 160 WUs which is about 30 minutes work, so obviously I want to keep as near to this maximum as possible. BOINC 6.x would connect to the project quite frequently and hence always keep near the maximum WUs. BOINC 7.x only seems to connect when it has nearly run out and so never builds up to the 160 WUs cached because the most it ever gets back from the server in one go seems to be 24-25 WUs.
Is there any way I can get the behaviour I want in BOINC 7?
On the positive side the memory usage issues seem to be fixed in 7.x
The easiest thing is to reverse your ver 6 numbers, the first box is now how many you MUST have before Boinc asks for more, so if you have 0.1 in that box it is 0.1 days worth before it asks for more. The 2nd box is how many extra it should get when it does get some units. My settings are now LARGER in the 1st box and smaller in the 2nd box and it works for me. I like a smaller cache so 0.75 is in my 1st box and 0.10 is in my second box. This gives me just over 1 days worth of work.
/ Reply
Quote
I have it set to 2 days and yet it still doesnt ask for more. The maximum this project allows is 40 WU per GPU core which is only 20-30 mins work. So I would have thought since it is always going to have less than 2 days work it would always request work when it finishes as WU but this doesnt seem to happen. If I manually make it keep asking until it reaches the server limit which is 160 WU (4x40) then it runs 150 WU before it connects to server then server only gives 24 back. All i want is to try and keep near the maximum 160 WUs cached to cover short term loss of internet connectivity on my end which occurs now and then.
/ Reply
Quote
Where are these settings? I'm getting a max if 6 WUs for my computer (sometimes 5) and I'm running a quad on one computer turning WUs in 1hr 45 min. I only alllow 2 cores to get work so it doesn't drag my system down but every 3 hours I'm looking for work.
On my other computer I'm running a dual core that's turning WUs ever 3hrs 10 min and every 6 hours its looking for work. I'd like to at least get a days worth. Hell even 1/2 day is better than looking every 3 hours.
In my preferences I have it set for 1 day with an additional day and at most I still get 5 or 6 workunits.
And to make it even worse I have units that haven't verified yet and these people have 300 to 900 WU's and an timing out for 2 weeks ago and are still getting new work everyday. This is frustrating to say the least that I have to look for work every 3 to 6 hours and hope something is there while others are grabbing 100's and not even doing the work.
What do I have to do to get at least a days worth of work here?
Help would be greatly appreciated.
____________
/ Reply
Quote
If you click "your account" at the bottom of this page then go to "Computing preferences" and you have two settings in there called
"Maintain enough tasks to keep busy for at least" and "... and up to an additional"
I assume you are taking about CPU WUs? Im not sure what the limiti is on these because I use my CPU on other projects and only do MW on the GPU. The problem here is we are only allowed to cache 40 WUs and these are done in 30-60 seconds so the cache doesn't last long at all even when it is maxed out. When I can't even get it to fill the tiny cache we are allowed it will run out of work in like 5 mins if it can't connect.
/ Reply
Quote
Yes, I'm running CPU only and I have both of those boxes with 1.0 for number of days. Still only get 5 to 6 workunits...
I dunno..... If its frustrating for me then it has to even worse for you and others that are blasting through WU's even faster and are begging for workunits when someone has 300 to 900 workunits and is getting new ones every day and not even processing what they have already.
That's what sort of set me off. Why do they get them and I have to beg for 20 a day and mine get done without problems.
Oh well so it goes I guess.. Thanks for the reply.
____________
/ Reply
Quote
Yes, I'm running CPU only and I have both of those boxes with 1.0 for number of days. Still only get 5 to 6 workunits...
I dunno..... If its frustrating for me then it has to even worse for you and others that are blasting through WU's even faster and are begging for workunits when someone has 300 to 900 workunits and is getting new ones every day and not even processing what they have already.
That's what sort of set me off. Why do they get them and I have to beg for 20 a day and mine get done without problems.
Oh well so it goes I guess.. Thanks for the reply.
Could those people that have so many be using their gpu to crunch with? I too crunch with my gpu here and get a decent amount when I ask for units, but don't really care as my internet rarely goes down, I am knocking on wood right now!!!
/ Reply
Quote