I am unable to upload again... and when this happens and the transfer que grows.. my Q6600 starts to act strangley with other apps. The only thing I can do is suspend BOINC/SETI untill things start to flow again.
Sad for my Wingmen I guess.

---

When we finally figure it all out, all the rules will change and we can start all over again.

Yea.............know what you mean.

I am seriously considering dumping Boinc and reduce my stress level a bit.

With over 4000 MB work units and can't upload or download who cares??? Boinc/Seti keep stating they need more crunchers but they can't support the ones they have properly.

I just don't get Berkeley. I see new crunchers come, install Boinc, Boinc fails to meet their expectations due to whatever. They remove Boinc and dump their work into the void. Happens hundreds of times a day I would bet. You only need to look at some of the work that is about to be reissued and see it. They sign on, download work, never to be heard from again.

I can't believe that after all the time I have spent on this project that now I would really consider removing Boinc and trashing all the work I have. But why not? Can't upload or report it!

Yea..........somebody else can have my measly bit of the bandwidth.

[edit] With tomorrow being the "scheduled shutdown for maintenance" it will probably be Thursday or later before I could possibly get through to Berkeley.

Yea..........somebody else can have my measly bit of the bandwidth.

---

Boinc....Boinc....Boinc....Boinc....

Hi,
What kinds of symtoms are going on when your "Q6600 starts to act strangley with other apps"?

Did you try and suspend network activity, and if so did that have any effect? Can you tell at what point the number of tasks trying to upload seem to effect the pc?

Edit: Also have you tried shutting down and restarting Boinc, or Windows?

---

I just don't get Berkeley. I see new crunchers come, install Boinc, Boinc fails to meet their expectations due to whatever. They remove Boinc and dump their work into the void.

For new crunchers that quit right away, I suspect that BOINC simply doesn't entertain the way the're expecting. Maybe they want it to do something, and mostly it is supposed to hide and crunch.

For those who have been crunching for a while, and have studied BOINC some, they expect to see it do exactly what they think it should do, exactly when they think it should do it.

They see "Temporarily failed upload of " and focus on the word failed.

I see "failed" and say "oh, interesting" and to see how BOINC handles a failed upload.

I've even restricted network activity to 3 hours/day.

Nothing here that is due before Sunday, so if things don't work until Thursday, that's okay -- but I rarely see that.

---

Other Apps start to freez for a few seconds and then resume running normally.
This starts to occur after I have 30 or more Uploads & or Downloads backed up in the Transfer Que.
I also note that when I enable CUDA, One processor drops to on 5% utilization or so and the other three drop from 25% to about 21 or 22. When I do not allow CUDA, all four Processors hover around 25%.
Memory utilization seldom exceeds 2GB and Memory diagnostics show no problems.

The system is running at 3.1Ghz 4Gigs Ram, 1TB Sata 300 disk and a NVIDA 9600 on an EVGA MB.
There is no unusual net activity during these problems.

Oh this is on Windows 7 RC, [Tons better than XP/X64 or VISTA Pro X64.}
BOINC 6.6.36
SETI Enhanced 6.08

Go Fugure:)

---

When we finally figure it all out, all the rules will change and we can start all over again.

!@#$%^&* More Information....!!!
I turned off CUDA and hanging/freezing problem for other apps went away!!!
It's not related to Backlog after all I think.

My error Sorry for poor observation and inaccurate information.

---

When we finally figure it all out, all the rules will change and we can start all over again.

!@#$%^&* More Information....!!!
I turned off CUDA and hanging/freezing problem for other apps went away!!!
It's not related to Backlog after all I think.

My error Sorry for poor observation and inaccurate information.

Sounds like VLAR (very low angle range) tasks which do tend to have symptoms you've described when using CUDA. You could try the settings not to use the gpu when the pc is in use thru the Boinc Manager advanced menu, under processor usage to see if that helps.

---

Boinc/Seti keep stating they need more crunchers but they can't support the ones they have properly.

Odd, when I got home tonight and noticed the project was having issues again/still/whatever. A variation of that exact thought was the first thing that came to mind. :/


In the long run, it would seem like less might equal more (stability, anyway)?



Queue the Parrot: The project doesn't offer 99% uptime blah, blah, blah... ;)

I Have and it does help!!
Thanks

---

When we finally figure it all out, all the rules will change and we can start all over again.

Would a faster GPU help? Can I justify a GeForce GTX 295 ;)

---

When we finally figure it all out, all the rules will change and we can start all over again.

Would a faster GPU help? Can I justify a GeForce GTX 295 ;)

LOL....

The only thing a GTX 295 would do at this point is make you as frustrated as I am....with my 2 GTX 260's and nearly 200 WU's that are 100% and can't be uploaded because of bandwidth issues...... ;)

On a more serious note, what are the specs of your current card ? 9600-what...GT, GS, GSO etc. ? and how many shaders/cores ? how much memory ?

---

I am TCP JESUS...The Carpenter Phenom Jesus....and HAMMERING is what I do best!
formerly known as...MC Hammer.

GE Force 9600 GT
512K
64 Stream Processors
675 Core Clock
1700 Shader Clock
900 Memory Clock

I have Three PCIE slots on my MB.

All I need now is some Stimulus money:)

---

When we finally figure it all out, all the rules will change and we can start all over again.

GE Force 9600 GT
512K
64 Stream Processors
675 Core Clock
1700 Shader Clock
900 Memory Clock

I have Three PCIE slots on my MB.

All I need now is some Stimulus money:)

With my GTX295, I can watch streamed video without affecting machine performance so, yes it would improve your situation. With your MoBo, though, a pair of GTX275's could do more work and may be cheaper than a 295. Remember to check for PSU compatibility (I had to upgrade mine to supply the 295) and both solutions chuck out a LOT of heat so case cooling might need looking at too.

F.

---

I can tell you that if you have a upload cache of nearly 6000 wu's boinc stops to a crawl on a Phenom II 3.5 Ghz machine.

If boinc tries to upload 2-3 results to s@h it does that, when that is done boinc.exe spikes for around 24 seconds leaving boincmgr out of control, lets go and then uploads 2-3 more and pauses etc.

It took me around 6 hours of suspended activity to upload all work and boinc.exe was spiked at 25% cpuload nearly 95% of that time, if i had boinc running with both the manager and with work in the background there was always the "boinc is trying to communicate window" in boincmgr. I couldn't do anything at all because boinc was so busy in itself so it couldnt measure the heartbeats from the clients so they stopped and restarted every 30 seconds so it couldn't process work at all because it restarted the .exe's all the time.

Gonna write some suggestions to boinc devs if my thoughts are correct with what i'm about to propose to ease boinc workload for clients..

Interesting times though because this had to happen sooner or later and if you have built a system that nearly produces 2000 results in a day you tend to know the bottlenecks in boinc and how it manages its work.

Boinc was not meant to process this amount of work with client_state files that grew to beyond 30 mbytes, it simply takes too long time for itself.

Well enough of my babbling now :) , this was a sidetrack and my view of a borked s@h connection or ul/dl problems.

Kind regards Vyper

---

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group

Probably BOINC is like a road designed for horse-drawn carriages at the time when it had to support only the traffic due to crunching with "unused cycles". Now that people are using fast CPUs and faster GPUs it must support the traffic of a modern highway and cannot cope with it. This is another example of the unforeseen consequences of technical progress.
Tullio

---

Probably BOINC is like a road designed for horse-drawn carriages at the time when it had to support only the traffic due to crunching with "unused cycles". Now that people are using fast CPUs and faster GPUs it must support the traffic of a modern highway and cannot cope with it. This is another example of the unforeseen consequences of technical progress.
Tullio

Couldn't agree more :)

//Vyper

---

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group

Funny how if you take your pills and get a good nights sleep that things look much better in the morning. Especially if Boinc did it's job and uploaded all the pending work and got you more work to do. All while I was sleeping.

So I am trying to figure out why others can get through to Berkeley and max out the bandwidth but I can't even get an occasional hit on Berkeley at the same time. It must be something in the way the internet is constructed or the way my ISP is routed to Berkeley. Anyone know for sure? I'm certain there is nothing I can do about it except maybe change my ISP and maybe that wouldn't help.

---

Boinc....Boinc....Boinc....Boinc....

What I'm about to say may be regarded as heresy, but here goes.....

I have a suspicion that old, slow hosts - like my single-core P4s, and especially my 400 MHz MMX Celeron - have a better chance of hitting the scheduler, and having their meagre caches refilled after an outage, than multi-core, mega-GPU behemoths (not that I run any of those, but I can dream.....)

I'm not suggesting any conspiracy, just an observed trend. One possible mechanism would be to observe that the transaction involved, say "request new work", involves transmitting a sched_request....xml file. The last one from my Celeron was 9 KB: from my 8-core Xeon (no CUDA) 39 KB: and from a quad+CUDA 123 KB. I've seen reports from mega-crunchers that their request files can be over 1 MB.

The request files have to fight their way to Berkeley through all the other data traffic. It doesn't surprise me that the biggest ones have the least chance of arriving intact and in a timely manner.

So the pipes are full of little guys getting their one or two tasks, and wondering what all the fuss is about, while the battleships stand out to sea and have to wait for the tide to turn....

Richard it makes sense to me.

---

Could be but when the pipe is full I can't even get an upload and report to go through. Usually have hundreds of uploads backed up so of course no requests for downloads are being done at this time.

---

Boinc....Boinc....Boinc....Boinc....