Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
![]() |
World Community Grid Forums
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
No member browsing this thread |
Thread Status: Active Total posts in this thread: 822
|
![]() |
Author |
|
erich56
Senior Cruncher Austria Joined: Feb 24, 2007 Post Count: 295 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
What I have noticed: there is much less GPU work available today, compared to the days before.
|
||
|
zdnko
Senior Cruncher Joined: Dec 1, 2005 Post Count: 229 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
What I have noticed: there is much less GPU work available today, compared to the days before. Today I received only 3 wu for nvidia, no one for intel. ![]() |
||
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
I started to get work units again as i got like 12 in the last couple of hours
|
||
|
kittyman
Advanced Cruncher Joined: May 14, 2020 Post Count: 140 Status: Offline Project Badges: ![]() ![]() ![]() |
There seems to be just enough work coming this way for the last few days, but the kitties' needs are modest, with a single GTX980 munching from the kibble bowl. I can easily see how users with multiple or higher powered GPUs would not be getting nearly enough to keep their rigs busy.
----------------------------------------But the kitties have been satisfied as of late. Meow! ![]() |
||
|
bozz4science
Advanced Cruncher Germany Joined: May 3, 2020 Post Count: 104 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
After the stress test I paused GPU work for a while and just resumed computing those GPU WUs recently. Even after crunching through only ~50 WUs, the writes are still pretty high and the intermittent dumps in the GPU compute utilization combined with the yet unsolved write amount issue have me stumped. Is there any ongoing effort to look into these two open issues. Not only is it an annoyance to run 7-8 WUs concurrently on even a 1660S card to have GPU CUDA compute util stay relatively constant at >= 99% and therefore the need to give up 7-8 CPU threads, but it is consuming my SSD's hardware lifespan unncessarily. 20 TB written for the last 5 weeks for intermittent OPNG support only.
----------------------------------------Surely, there must be technical tricks to tackle that write tsnumami... Thx ![]() AMD Ryzen 3700X @ 4.0 GHz / GTX1660S Intel i5-4278U CPU @ 2.60GHz |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Surely, there must be technical tricks to tackle that write tsnumami... If you are on Windows, use PrimoCache (use the write-cache only): https://www.romexsoftware.com/en-us/primo-cache/index.html If you are on Linux, use their built-in write cache: https://lonesysadmin.net/2013/12/22/better-li...rformance-vm-dirty_ratio/ I have not bothered to looks at the writes here, since they aren't sending out many, but I would think that a cache size of 2 GB and a latency (write-delay) of 30 minutes should work. I use them on all of my machines to protect the SSDs. |
||
|
nanoprobe
Master Cruncher Classified Joined: Aug 29, 2008 Post Count: 2998 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
Surely, there must be technical tricks to tackle that write tsnumami... If you are on Windows, use PrimoCache (use the write-cache only): https://www.romexsoftware.com/en-us/primo-cache/index.html If you are on Linux, use their built-in write cache: https://lonesysadmin.net/2013/12/22/better-li...rformance-vm-dirty_ratio/ I have not bothered to looks at the writes here, since they aren't sending out many, but I would think that a cache size of 2 GB and a latency (write-delay) of 30 minutes should work. I use them on all of my machines to protect the SSDs. I assume you're using the paid version on windows? I ask because the the only delayed write settings I could find were 10,60 and 300 seconds or infinite.
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.
----------------------------------------![]() ![]() [Edit 1 times, last edit by nanoprobe at Jun 8, 2021 6:13:13 PM] |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
I assume you're using the paid version on windows? I ask because the the only delayed write settings I could find were 10,60 and 300 seconds or infinite. Yes. I don't know of any free ones. Primo spent years developing it to eliminate all the errors. I have never seen such a long beta. You could try a ramdisk instead, but you have to size them to hold the entire BOINC data folder. I think there are some free ones, but I use Primo Ramdisk (the best monitoring tools) or Dataram. [Edit 1 times, last edit by Jim1348 at Jun 8, 2021 7:28:35 PM] |
||
|
erich56
Senior Cruncher Austria Joined: Feb 24, 2007 Post Count: 295 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
You could try a ramdisk instead, but you have to size them to hold the entire BOINC data folder. I think there are some free ones, but I use Primo Ramdisk (the best monitoring tools) or Dataram. I am using the Primo Ramdisk Professional version, which provides a Ramdisk up to 32GB (the free version has only 16GB, as far as I remember). Price was USD 49. My total RAM is 64GB, so I used 32GB for the RAMdisk (good part of it is being used by the ATLAS tasks from LHC, which are also known as SSD killers). Everything works perfect and very fast. |
||
|
Jim1348
Veteran Cruncher USA Joined: Jul 13, 2009 Post Count: 1066 Status: Offline Project Badges: ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
My total RAM is 64GB, so I used 32GB for the RAMdisk (good part of it is being used by the ATLAS tasks from LHC, which are also known as SSD killers). Everything works perfect and very fast. That reminds me of the good old days, when I had a large RAMdisk for the CEP2 project here on WCG. It had a horrendously high write rate, over 1 TB/day for 8 cores. Also, CPDN had a high write rate back then, though less now. The limitations on the write rate for mechanical disks caused errors too. I was able to eliminate them with the ramdisk. With enough memory, you can do a lot of things. |
||
|
|
![]() |