http://setiathome.berkeley.edu/forum_thread.php?id=52968

• > BOINC V6.6.11 can't support WU cache > ~ 2,500 WUs
  
  
• after deinstallation of BOINC V6.6.x - BOINC V6.4.x show 'funny' nVIDIA driver versions
  
  
• CPU work request, although the CPU should idle
  

---

Hello,

I could reach the developers with this thread?

Or the developers don't look here around?

Is this 'only' a forum for user to help other user?

How it's possible for me to get in contact with the developers?

If they are here around also, it would be nice if they would get a sign, that they had noticed BUG-related threads.

Thanks a lot.

Best wishes.



PS.
It's not a DEV-V of BOINC - it's a 'final'.. a recommended version.
Although after .20, it came .21, .22 and current V6.6.23.

---

The developers hardly read these forums, your better bet is to reach them through Trac, which is the BOINC bug database. Register there, then make a new ticket.

---

http://setiathome.berkeley.edu/forum_thread.php?id=52968

• > BOINC V6.6.11 can't support WU cache > ~ 2,500 WUs
  
  
• after deinstallation of BOINC V6.6.x - BOINC V6.4.x show 'funny' nVIDIA driver versions
  
  
• CPU work request, although the CPU should idle
  

The 1st one (WU cache) has been addressed in 6.6.24. See this message thread.

---

MarkJ

Thanks to you both! :-)

The 1st one (WU cache) has been addressed in 6.6.24. See this message thread.

Are you sure? ;-)


I didn't found this change in V6.6.24

Message 24503

Under which point/name is this listed?


Now V6.6.28 is the recommended version.
Big WU cache will work also with this version?


Thanks! :-)

---

You will need to be a lot more specific than that if you want developers to listen.

> BOINC V6.6.11 can't support WU cache > ~ 2,500 WUs

Why not? What happens if you have that many WUs? Or do you mean BOINC isn't requesting that much work and you think it should?

after deinstallation of BOINC V6.6.x - BOINC V6.4.x show 'funny' nVIDIA driver versions

Define "funny". Do the driver versions make you laugh?

Isn't this basically reporting a bug on 6.4, which is already old?

CPU work request, although the CPU should idle

What do you mean with "the CPU should idle", you want it to be idle?

Post a log from the Message tab showing exactly what happens, and explain what you think should be happening.

---

If you followed the URL in my first post, you could read my whole experiences.

BTW - you are a dev?


>BOINC V6.6.11&1/2 CUDA performance
http://setiathome.berkeley.edu/forum_thread.php?id=52968


I don't wanted to overSPAM the ticket and this thread.


O.K., here my experiences:

> BOINC V6.6.11 can't support WU cache > ~ 2,500 WUs

Why not? What happens if you have that many WUs? Or do you mean BOINC isn't requesting that much work and you think it should?

> BOINC V6.6.11 have problems to manage high cache with > ~ 2,500 WUs [~ 900 MB]!

And because of this BOINC have 1/2 and more CUDA performance lost.


I made this update from BOINC V6.6.11 to V6.6.14 and V6.6.20 when I had a cache of ~ 3,700 WUs.


Then I tested BOINC V6.6.20 with ~ 2,500 WUs and I had 100 % CUDA performance.
Performance like it's normal for =< BOINC v6.6.11 and ~ 3,700 WUs.


~ 3,700 WUs was the 10 day cache for 2 GPUs.
Now I added two more GPUs.. this would mean with better driver and CUDA .dll's ~ 8,000 WUs.
Soon I will test BOINC V6.4.7 and ~ 8,000 WUs..
I don't know if this will be possible.. I will make the test.

after deinstallation of BOINC V6.6.x - BOINC V6.4.x show 'funny' nVIDIA driver versions

Define "funny". Do the driver versions make you laugh?

Isn't this basically reporting a bug on 6.4, which is already old?

BOINC V6.4.x don't post the driver version to the Berkeley-overview-side.
BOINC V6.6.x do this.

But, after deinstallation of BOINC V6.6.x and latest .20 - BOINC V6.4.7 show now 'funny' driver versions.
For example:
driver: 925775673, 1667316016, -1, or others.. and sometimes nothing.. after every update a change..



O.K., if this is already reported.. nobody said it to me in my old threads..

CPU work request, although the CPU should idle

What do you mean with "the CPU should idle", you want it to be idle?

Post a log from the Message tab showing exactly what happens, and explain what you think should be happening.

Only CUDA app_info.xml..

I unchecked CPU usage in the preferences. -> request for new CPU work, and 4 idle CPUs [every minute], [regardless also to GPU work request]....

Then additional..
I insert the <ncpus>0</ncpus> to cc_config.xml. -> request for new CPU work, and 4 idle CPUs [every minute], [regardless also to GPU work request] - and get answer "no work sent - check preferences" [this would mean high Berkeley server load!]

Then additional..
I made my <avg_ncpus>2.00x -trick in app_info.xml. -> request for new CPU work only simultaneously with GPU request [double CPU- as GPU- seconds], but get only CUDA work.. *happy*

[Quad-CPU + 2 GPUs - ONLY BOINC/SETI@home - no other project - ONLY GPU CRUNCHING !]



And.. for a small test..
I reduced the CPUs in the computing preferences.
Simultaneously MB/CPU + CUDA/GPU.
I chose only use 2 CPUs - but 4 MBs/CPU were running.

If this because of the app_info.xml - I could limit the CPUs with <ncpus> in cc_config.xml ?

---

If you followed the URL in my first post, you could read my whole experiences.

Okay, so a developer who sees your ticket on Trac has to follow the link to this thread, then from here to the (long) SETI thread, in order to know what's the problem you found? I should close that ticket as invalid then, because it provides almost no information.

I don't wanted to overSPAM the ticket and this thread.

Saying what the problem is wouldn't be overspamming, it would make your ticket remotely useful.

BOINC V6.6.11 have problems to manage high cache with > ~ 2,500 WUs [~ 900 MB]!

There goes vagueness again. What is "have problems"?

Anyway, BOINC has never worked right with that many WUs. Closing the manager will help (GUI uses a lot of CPU, and client uses a lot of CPU when talking to the GUI giving it information for that many WUs). Bug has been on Trac for years but it's not easy to fix.

O.K., if this is already reported.. nobody said it to me in my old threads..

I didn't say it was already reported. If it's a bug in 6.4, upgrade to 6.6. There won't be more bug fixing in 6.4, unless the bug is really serious, like deleting the user's documents...

---

I said all the time with > BOINC V6.6.11 I have 1/2 CUDA performance lost.

An AR-WU 0.4x go up from ~ 8 min. to ~ 30 min. or more . [this are nearly 1/4 performance]
ONLY with changing V6.6.11 to V6.6.14 or .20. [with a cache of ~ 3,700 WUs]


You think if I will close the Manager the time will be again normal?
Normally I don't have the Manager open.. it's an only crunching PC.. only short time after an update I look to the Manager..

But this wouldn't be the prob.. if I use BOINC V6.6.20 and less than ~ 2.500 WUs.. the CUDA tasks [same AR] have the same ~ 8 min.
The 'wall' are this ~ 2,500 WUs in cache.
If I'm over this ~ 2,500 WUs and > BOINC V6.6.11 the CUDA performance go bad.


I don't know why, and don't know how..
But I see it in BOINC Manager..
The calculation time (I repeat me) go from ~ 8 min. to ~ 30 min. or more [same AR-WU!].


I thought if I uninstall V6.6.20 all function will be deleted.
BOINC V6.4.7 didn't showed a nVIDIA driver version, now after uninstall of BOINC V6.6.20 it show my 'funny' driver versions.


If you will delete my ticket.. it was my first time and nobody said me the rules.. nobody will know the BUGs..

---

You think if I will close the Manager the time will be again normal?
Normally I don't have the Manager open.. it's an only crunching PC.. only short time after an update I look to the Manager..

No, I don't know all the answers. I was suggesting closing the manager will definitely help, not that it will solve the whole problem.

But this wouldn't be the prob.. if I use BOINC V6.6.20 and less than ~ 2.500 WUs.. the CUDA tasks [same AR] have the same ~ 8 min.
The 'wall' are this ~ 2,500 WUs in cache.
If I'm over this ~ 2,500 WUs and > BOINC V6.6.11 the CUDA performance go bad.

So with all versions from 6.6.11 to 6.6.20, having so many WUs slows down CUDA performance? Does it still happen with the latest version, 6.6.28?

Does it still happen with a version older than 6.6.10? That would be useful to know, we could see if there was any change in .11 that could have introduced a new problem.

Do you keep the manager closed, or just minimized to the tray icon? I think if it's on the tray icon, it will still be using a lot of CPU with a long list of WUs.

If you open task manager, do you see boincmgr.exe or boinc.exe (or both) using too much CPU time?

If you will delete my ticket.. it was my first time and nobody said me the rules.. nobody will know the BUGs..

Keep the problem in the forum, once we get enough information to reproduce the problem, someone who is more experienced could write a proper ticket with all the details.

For all we know now, it could be a problem with your computer and not a bug.

---

BTW.
Thanks for nothing.

Now you deleted the ticket.

You make it not very interesting for people to support the devs.



It's a prob with my PC, I don't think so.

Like I said, BOINC V6.4.5 and V6.4.7 and V6.6.5 and V6.6.11 can manage/support ~ 3,700 tasks cache.

In the time I had ~ 3,700 tasks on the PC I updated to BOINC V6.6.14 and later V6.6.20.
Both can't support/manage ~ 3,700 WUs.

In TaskManager I could see that the CPU support dropped 1/2 and more, sometimes long time to 0 %.
Also the calculation time of the SAME AR-WUs increased much, from ~ 8 min. to ~ 30 min. .

BOINC Manager was opened in this test.

Only boinc.exe peaks.
No boincmgr.exe peaks.


Why I know this.. because of server probs at Berkeley I ran down to ~ 2,500 tasks on my PC.
Then I gave V6.6.14 and later V6.6.20 a last try.. and surprising.. they could manage/support ~ 2,500 tasks.
Well calculation time of the tasks.



You took my interest for to post tickets.. so if I could reach the devs with this post, it would be well.

---

Now you deleted the ticket.

I didn't. I only delete tickets if they have spam.

It's a prob with my PC, I don't think so.

We don't know if it's a problem with your PC or not. Until we know for sure, we shouldn't open a ticket.

Like I said, BOINC V6.4.5 and V6.4.7 and V6.6.5 and V6.6.11 can manage/support ~ 3,700 tasks cache.

In the time I had ~ 3,700 tasks on the PC I updated to BOINC V6.6.14 and later V6.6.20.
Both can't support/manage ~ 3,700 WUs.

In TaskManager I could see that the CPU support dropped 1/2 and more, sometimes long time to 0 %.
Also the calculation time of the SAME AR-WUs increased much, from ~ 8 min. to ~ 30 min. .

BOINC Manager was opened in this test.

Only boinc.exe peaks.
No boincmgr.exe peaks.

That's quite interesting. Does boinc.exe still peak if you close the manager? (boinc.exe uses CPU when sending the information to boincmgr)

And to get the version numbers straight: 6.6.11 manages 3600 tasks and 6.6.14 doesn't?

Does the problem still happen with non-CUDA workunits?

---