Message boards :
Number crunching :
Webside MV very slow
Message board moderation
Author | Message |
---|---|
Send message Joined: 12 Aug 09 Posts: 262 Credit: 92,631,041 RAC: 0 |
Hello, I encounter very slow responces of the MW websites. The Message boards is still not fully updayed after 5 minutes. I also find that the WU's ready to report are not sent automatically, they queued up and when I do a manual update not a lot seems to happen. My internet connection is 18.9MBbs affective and is working fine at the moment. Have other people the same? Greetings from, TJ |
Send message Joined: 12 Aug 09 Posts: 262 Credit: 92,631,041 RAC: 0 |
I found a part of the reason: 2/21/2011 1:31:59 AM Milkyway@home Started upload of de_separation_10_3s_fix_3_625854_1298240049_1_0 Greetings from, TJ |
Send message Joined: 12 Aug 09 Posts: 262 Credit: 92,631,041 RAC: 0 |
Now I have two pages with "client detached" so the ATI's run an hour for nothing. |
Send message Joined: 24 Dec 07 Posts: 1947 Credit: 240,884,648 RAC: 0 |
I'm not seeing any problems with the web page updating. |
Send message Joined: 28 Mar 09 Posts: 68 Credit: 1,003,982,681 RAC: 0 |
See Matt's answer in "aborted by project" thread. |
Send message Joined: 12 Aug 09 Posts: 262 Credit: 92,631,041 RAC: 0 |
See Matt's answer in "aborted by project" thread. I have read that post, however that were a bunch of old runs, and mine were new. I got them February 20th in the evening. So this is something else or new. Greetings from, TJ |
Send message Joined: 30 Dec 07 Posts: 311 Credit: 149,490,184 RAC: 0 |
Yes I have experienced spontaneous detaching on GPU projects. In my case it was not caused by BAM which apparently can do this sometimes as I have never used it. Happened to me a number of times on both Collatz and MilkyWay but never on DNETC. Used to happen much more often on my old HD 4890 which ran very hot, but only very rarely on my HD 5970 and not for a long time recently. It is some BOINC thing I suppose, perhaps the video driver gets stuck or crashes, BOINC then detects no video card so aborts all tasks in your cache of the GPU project you were crunching when it occurred. Possible solutions include fully removing and reinstalling the Catalyst driver or trying another driver version and making certain that your cards are not overheating especially if you are running unattended during the hotter part of the day. |
Send message Joined: 12 Aug 09 Posts: 262 Credit: 92,631,041 RAC: 0 |
Kashi, after reading your post you are right. But the strange thing is, it is only happening to a few of the WU, not all in the queque and only the ones that wait for sending (after uploading). I had 5 or 6 last week but Matt said they deleted old ones. So that was it. But yesterday, is or was a new "thing". The cards run with 91°C and a 48°C airflow so that is okay. Greetings from, TJ |
Send message Joined: 24 Dec 07 Posts: 1947 Credit: 240,884,648 RAC: 0 |
Kashi, after reading your post you are right. 91°C is very hot. I'd try cooling that down to under 85°C if I was you and even under 80°C. |
Send message Joined: 30 Dec 07 Posts: 311 Credit: 149,490,184 RAC: 0 |
When I experienced it, all tasks that were in the cache at the time it happened were aborted. This included tasks that had completed processing, had uploaded but had not yet reported and any unfinished tasks. Due to the small cache allowed in MilkyWay unless you notice it in BOINC Manager at the time it happens you would see a batch of "detached" tasks probably equal to the size of your allowed MilkyWay cache. The message in BOINC Manager for each aborted task would be in red and would say something like: Result de_separation_15_3s_fix_3_994759_1298286883_0 is no longer usable. Whereas on your tasks page on MilkyWay website for the aborted tasks it would say something like: Client detached Naturally this only happens after BOINC contacts the server. So the tasks that are aborted when the error happens are not marked as aborted until after BOINC contacts the server. So it is possible to complete the whole batch if it is a small batch after they have become unusable. Whether you would have any tasks not yet started included in the aborted batch depends on how you set your BOINC preferences for Connection and Additional work buffer. If you only notice it later you would just see the number of tasks in that batch that had aborted. Because the driver or card had recovered from the problem BOINC would download another batch of tasks when it contacted the server. In Collatz however there is a potential for a much larger number of tasks to be aborted because the allowable cache is much larger. |
Send message Joined: 12 Aug 09 Posts: 262 Credit: 92,631,041 RAC: 0 |
91°C is very hot. I'd try cooling that down to under 85°C if I was you and even under 80°C. Hi Gas Giant, How can I cool that so low. It is a Dell Alienware with liquid cooling on the processor. The 2 ATI's are in a "tube" or more a box with a dedicated fan blowing along, over and through the cards. In total the case has 4 fans. Ambient temperatture sensors of the systems are at 21°C and after a few hours 23°C. I don't believe them. CPUI Hardware monitor finds a GPU Core temp of 44°C at start of system and 91°C after a few hours, not hotter. This bix is only running at night when I have a cheaper power rate. So only 7-9 hours a day. Longer in the weekends. Do you have an idea for me? Thanks. Greetings from, TJ |
©2024 Astroinformatics Group