Author |
Message |
|
Hello,
Boinc 6.3.11 for Windows is released and can be downloaded here.
I installed it and now my dual-core runs three tasks. CPDN, EAH and GPUGrid simultaneously.
Of course it is a development version too!
Use it on your own risk!
Nowi |
|
|
JKuehl2Send message
Joined: 18 Jul 08 Posts: 33 Credit: 3,233,174 RAC: 0 Level
Scientific publications
|
Not working on Quadcore. Still 4 simultaneous WUs (3x Spinhenge, 1x Ps3grid).
At least i hope the developers were not dumb enough to just change source code from NCPU to NCPU+1.. That would give the same result as changing cc_config.xml and therefore also the huge slowdown in computation time of PS3grid WUs. |
|
|
|
Not working on Quadcore. Still 4 simultaneous WUs (3x Spinhenge, 1x Ps3grid).
At least i hope the developers were not dumb enough to just change source code from NCPU to NCPU+1.. That would give the same result as changing cc_config.xml and therefore also the huge slowdown in computation time of PS3grid WUs.
As far as I'm concerned, I didn't experience ANY slowdown. Check my WUs, the last 4-5 ones were done with the NCPU+1 option enabled
|
|
|
GDFVolunteer moderator Project administrator Project developer Project tester Volunteer developer Volunteer tester Project scientist Send message
Joined: 14 Mar 07 Posts: 1957 Credit: 629,356 RAC: 0 Level
Scientific publications
|
Not working on Quadcore. Still 4 simultaneous WUs (3x Spinhenge, 1x Ps3grid).
At least i hope the developers were not dumb enough to just change source code from NCPU to NCPU+1.. That would give the same result as changing cc_config.xml and therefore also the huge slowdown in computation time of PS3grid WUs.
As far as I'm concerned, I didn't experience ANY slowdown. Check my WUs, the last 4-5 ones were done with the NCPU+1 option enabled
You had a 10% slowdown from 71ms to 77ms. Considering the different performance of the GPU vs CPU, it is more rewarding to have 1 CPU unused (used to control the CPU) and top GPU speed.
As I said we will soon make use of the CPU for a part of the calculation which is suitable for the CPU. In the meanwhile you are free to use ncpus to override it if you think that it is more convenient. I don't think that it is.
GDF
|
|
|
|
6.3.11 is no working for four hours without any change on the cc_config.xml.
I notice that the scheduler is running all of my tasks. So I have five task running simultaneously (2xCPDN, EAH, QMC, GPUGrid). The process explorer shows different amount of processor-time for each task. I don´t know how the resources are shared and whether it depends on the resource share I set in the project preferences.
Any suggestions to this?
|
|
|
|
.... Just joined the GPU effort with one host - I can only get 4 tasks in total running on my Quad.
I have found that on my systems 6.3.11 has the same annoying screen flicker as 6.3.10 - 6.3.8 is better on (my) eyes.... |
|
|
|
As far as I'm concerned, I didn't experience ANY slowdown. Check my WUs, the last 4-5 ones were done with the NCPU+1 option enabled
You had a 10% slowdown from 71ms to 77ms. Considering the different performance of the GPU vs CPU, it is more rewarding to have 1 CPU unused (used to control the CPU) and top GPU speed.
As I said we will soon make use of the CPU for a part of the calculation which is suitable for the CPU. In the meanwhile you are free to use ncpus to override it if you think that it is more convenient. I don't think that it is.
GDF
A) That slowdown happened BEFORE I switched the option on.
My times are:
Without NCPUS+1 = 71.040; 70.454; 70.112; 73.247; 75.679; 79.532; 76.774; 91.687 (strange one, but I switched to NCPUS+1 DURING this WU)
With NCPUS+1 = 80.055; 77.799; 77.203; 78.304;
As you can see, the last ones WITHOUT NCPUS+1 are 75-79-76. The ones WITH NCPUS+1 are 80-77-77-78. I wouldn't say a 10% decrease, more likely a 3%.
B) Was it 10% or 3%, it still pales down before the full load of a core which would otherwise be unused.
When the Linux one will be available, I'll try the 6.3.11 version of BOINC and see if there is any change.
And when GPUGRID will be able to use my CPU as well, I will decide if I want to give its power to the project. Until then, my CPU is better used for others.
|
|
|
|
I see no immediate change with 6.3.11. It is NOT running 2 CPU tasks + 1 GPU task, only 1 CPU task and 1 GPU task, same as6.3.10. Same on both identical hosts.
9/23/2008 8:43:56 AM||Starting BOINC client version 6.3.11 for windows_intelx86
9/23/2008 8:43:56 AM||This a development version of BOINC and may not function properly
9/23/2008 8:43:56 AM||log flags: task, file_xfer, sched_ops
9/23/2008 8:43:56 AM||Libraries: libcurl/7.18.0 OpenSSL/0.9.8e zlib/1.2.3
9/23/2008 8:43:56 AM||Data directory: E:\Keiths\HP301 BOINC 6 Data
9/23/2008 8:43:56 AM||Running under account Keith
9/23/2008 8:43:56 AM||Processor: 2 GenuineIntel Intel(R) Pentium(R) 4 CPU 3.00GHz [x86 Family 15 Model 4 Stepping 9]
9/23/2008 8:43:56 AM||Processor features: fpu tsc pae nx sse sse2 mmx
9/23/2008 8:43:56 AM||OS: Microsoft Windows XP: Professional x86 Editon, Service Pack 2, (05.01.2600.00)
9/23/2008 8:43:56 AM||Memory: 2.00 GB physical, 3.85 GB virtual
9/23/2008 8:43:56 AM||Disk: 149.04 GB total, 138.57 GB free
9/23/2008 8:43:56 AM||Local time is UTC -4 hours
9/23/2008 8:43:56 AM||Not using a proxy
9/23/2008 8:43:56 AM||CUDA devices found
9/23/2008 8:43:56 AM||Coprocessor: GeForce 8800 GT (1)
9/23/2008 8:43:56 AM||Version change (6.3.10 -> 6.3.11)
...
9/23/2008 8:43:56 AM||General prefs: from malariacontrol.net (last modified 21-Aug-2008 07:41:54)
9/23/2008 8:43:56 AM||Computer location: work
9/23/2008 8:43:56 AM||General prefs: no separate prefs for work; using your defaults
9/23/2008 8:43:56 AM||Reading preferences override file
9/23/2008 8:43:56 AM||Preferences limit memory usage when active to 1023.24MB
9/23/2008 8:43:56 AM||Preferences limit memory usage when idle to 1841.83MB
9/23/2008 8:43:56 AM||Preferences limit disk usage to 0.93GB
9/23/2008 8:43:56 AM||Running CPU benchmarks
9/23/2008 8:43:56 AM||Suspending computation - running CPU benchmarks
9/23/2008 8:44:28 AM||Benchmark results:
9/23/2008 8:44:28 AM|| Number of CPUs: 2
9/23/2008 8:44:28 AM|| 1334 floating point MIPS (Whetstone) per CPU
9/23/2008 8:44:28 AM|| 2394 integer MIPS (Dhrystone) per CPU
9/23/2008 8:44:29 AM|PS3GRID|Restarting task nWu7048-GPUTEST3-5-10-acemd_1 using acemd version 645
9/23/2008 8:44:29 AM|Docking@Home|Restarting task 1d4j_mod0013sc_2194_377705_0 using charmm34 version 615
The problem is with prefs
9/23/2008 8:43:56 AM||Computer location: work
9/23/2008 8:43:56 AM||General prefs: no separate prefs for work; using your defaults
There are separate prefs for work, and always have been, It should not be using defaults. This is a problem as it is using the wrong preferences, now I know why it is not working correctly. I can view them on the website still, so they are there. Checking my local global_prefs file and the venue for work is there. Client is not functioning properly. |
|
|
|
Hello,
Boinc 6.3.11 for Windows is released and can be downloaded here.
I installed it and now my dual-core runs three tasks. CPDN, EAH and GPUGrid simultaneously.
Are you sure about that, Nowi?
I wonder if this is with or without ncpus+1.
All here (including me) see no effect with the 6.3.11, it still is the same like with 6.3.10 and doesn't work without the ncpus-variable.
I wonder what the developers changed on that version now...
____________
Member of BOINC@Heidelberg and ATA!
|
|
|
|
My result on 6.3.10
Without NCPUS+1: 19h on a 9600GT
With NCPUS+1 : 26h
The boinc 6.3.11 does not run three tasks on dualcore cpu, only two wu are calculating :( |
|
|
GDFVolunteer moderator Project administrator Project developer Project tester Volunteer developer Volunteer tester Project scientist Send message
Joined: 14 Mar 07 Posts: 1957 Credit: 629,356 RAC: 0 Level
Scientific publications
|
My result on 6.3.10
Without NCPUS+1: 19h on a 9600GT
With NCPUS+1 : 26h
The boinc 6.3.11 does not run three tasks on dualcore cpu, only two wu are calculating :(
This is expected.
The 6.3.11 improves GPU performance when running 5 tasks on a 4 cores.
It is up to the server (so us) to request or not the extra CPU. So far, we are requesting it until we see that there is no difference in GPU performance, a slower GPU could cause to reduce instead to increase the number of credit/day of your machine even summing up the extra credits coming from the CPU core.
gdf |
|
|
|
Hello,
Boinc 6.3.11 for Windows is released and can be downloaded here.
I installed it and now my dual-core runs three tasks. CPDN, EAH and GPUGrid simultaneously.
Are you sure about that, Nowi?
I wonder if this is with or without ncpus+1.
All here (including me) see no effect with the 6.3.11, it still is the same like with 6.3.10 and doesn't work without the ncpus-variable.
I wonder what the developers changed on that version now...
My ncpus value is 0 in the configfile. Now I´m crunching five tasks on a dual-core.
When I installed and started 6.3.11 only three tasks were active as shown above. After about 30 minutes the two additional parts followed resulting in processing five tasks.
I tried a boinc-restart but the processing of five tasks still remains.
To make it sure: I never dealed with the ncpus-option in the configfile.
At time I have no completion time for my GPUGrid-WU, but it seems that it works slower with five tasks active because the GPU-Coretemp decreased. |
|
|
STE\/ESend message
Joined: 18 Sep 08 Posts: 368 Credit: 4,173,502,885 RAC: 27,150,888 Level
Scientific publications
|
BOINC View works like it's supposed to now with the GPU Clients, so that was a Fix ... :) |
|
|
JKuehl2Send message
Joined: 18 Jul 08 Posts: 33 Credit: 3,233,174 RAC: 0 Level
Scientific publications
|
I tested some other things regarding the ncpus question
a) ncpus to 5, WINDOWS scheduler running at itself. Massive increase in computation time, temperature of GPU drops, fan slows down.
b) ACEMD gets one core for itself, 4 tasks of spinhenge get cores 2+3+4. ACEMD speed back to normal. (But of course one processor in idle).
using boinc.exe instead of boincmgr.exe: no effect - this was a suggestion from the developer mailing list, as boincmgr.exe seems to take up a lot of ram and cpu ressources under certain conditions.
further improvement in computation time: MANUALLY setting every single process on the pc to cores 2+3+4 and setting ACEMD to core 1 (instantly more stressed GPU - can be seen in further increase of temperature and fan speeds).
Using the program CPU-Control for this. Maybe i´ll bench some wus overnight |
|
|
STE\/ESend message
Joined: 18 Sep 08 Posts: 368 Credit: 4,173,502,885 RAC: 27,150,888 Level
Scientific publications
|
I'm going back to 6.3.10, I have as many as 20 Wu's running @ once with the 6.3.11, unless your actually runing a ps3grid wu the v6.3.11 client wants to run every wu in your cache from the other projects. |
|
|
|
I also switched back to 6.3.10. Maybe a further release will work in a way I can understand.... ;-) |
|
|
|
I'm going back to 6.3.10, I have as many as 20 Wu's running @ once with the 6.3.11, unless your actually runing a ps3grid wu the v6.3.11 client wants to run every wu in your cache from the other projects.
Yes, I have the same. 6.3.11 is usefull only if you have run PS3GRID. I think they didn't any testing with the new versiom. It is a big bug. |
|
|
|
My result on 6.3.10
Without NCPUS+1: 19h on a 9600GT
With NCPUS+1 : 26h
That is a bit strange.
I have the same card processor and with the 6.3.10 I did run several WUs in about 22 hours with the ncpus+1.
But now with the 6.3.11 I detected a time of 26h for the first WU...
I'm going back to the 6.3.10, too, I have the feeling, that the client slows down the processing.
____________
Member of BOINC@Heidelberg and ATA!
|
|
|
|
so far I like the new version. At least on server 2008 x64with the 64bit version of BOINC. Memory usage for boincmgr.exe is 6mb, down from 150-250 v10 took. Starting up is far quicker, and the lag when moving windows around is reduced on my 800gts512. I dont know yet about the speed, I will run a few WUs before deciding. I do now have ncpus at 5, and don't see any problems yet. I also don't see some of the screen anomalies I had, like it incorrectly organizing things. I see it still flickers on occasion as it updates itself, but the old one did that to. I have gpugrid set to a priority of 10,000, so it is not likely to not run this project. |
|
|
|
Is it crazy?
24.09.2008 0:06:50||Starting BOINC client version 6.3.11 for windows_x86_64
24.09.2008 0:06:50||This a development version of BOINC and may not function properly
24.09.2008 0:06:50||log flags: task, file_xfer, sched_ops
24.09.2008 0:06:50||Libraries: libcurl/7.18.0 OpenSSL/0.9.8g zlib/1.2.3
24.09.2008 0:06:50||Running as a daemon
24.09.2008 0:06:50||Data directory: C:\Documents and Settings\All Users\Application Data\BOINC
24.09.2008 0:06:50||Running under account boinc_master
24.09.2008 0:06:52||Processor: 2 AuthenticAMD AMD Athlon(tm) 64 X2 Dual Core Processor 5200+ [AMD64 Family 15 Model 67 Stepping 2]
24.09.2008 0:06:52||Processor features: fpu tsc pae nx sse sse2
24.09.2008 0:06:52||OS: Microsoft Windows XP: Professional x64 Editon, Service Pack 2, (05.02.3790.00)
24.09.2008 0:06:52||Memory: 2.00 GB physical, 3.88 GB virtual
24.09.2008 0:06:52||Disk: 9.53 GB total, 1.63 GB free
24.09.2008 0:06:52||Local time is UTC +7 hours
24.09.2008 0:06:52||Not using a proxy
24.09.2008 0:06:53||CUDA devices found
24.09.2008 0:06:53||Coprocessor: GeForce GTX 260 (1)
24.09.2008 0:06:54||Version change (6.3.10 -> 6.3.11)
24.09.2008 0:06:54|rosetta@home|URL: http://boinc.bakerlab.org/rosetta/; Computer ID: 889198; location: (none); project prefs: default
24.09.2008 0:06:54|Einstein@Home|URL: http://einstein.phys.uwm.edu/; Computer ID: 1028244; location: (none); project prefs: default
24.09.2008 0:06:54|lhcathome|URL: http://lhcathome.cern.ch/lhcathome/; Computer ID: 9640875; location: (none); project prefs: default
24.09.2008 0:06:54|QMC@HOME|URL: http://qah.uni-muenster.de/; Computer ID: 71074; location: (none); project prefs: default
24.09.2008 0:06:54|PS3GRID|URL: http://www.ps3grid.net/; Computer ID: 7721; location: (none); project prefs: default
24.09.2008 0:06:54|World Community Grid|URL: http://www.worldcommunitygrid.org/; Computer ID: 339073; location: (none); project prefs: default
24.09.2008 0:06:54||General prefs: from QMC@HOME (last modified 28-Jul-2007 09:07:47)
24.09.2008 0:06:54||Host location: none
24.09.2008 0:06:54||General prefs: using your defaults
24.09.2008 0:06:54||Reading preferences override file
24.09.2008 0:06:54||Preferences limit memory usage when active to 1023.69MB
24.09.2008 0:06:54||Preferences limit memory usage when idle to 1842.65MB
24.09.2008 0:06:54||Preferences limit disk usage to 1.16GB
24.09.2008 0:06:54||Running CPU benchmarks
24.09.2008 0:06:54||Suspending computation - running CPU benchmarks
24.09.2008 0:06:54||Suspending network activity - user request
24.09.2008 0:07:25||Benchmark results:
24.09.2008 0:07:25|| Number of CPUs: 2
24.09.2008 0:07:25|| 2333 floating point MIPS (Whetstone) per CPU
24.09.2008 0:07:25|| 5107 integer MIPS (Dhrystone) per CPU
24.09.2008 0:07:26|World Community Grid|Restarting task R00144_03ac858068a1cb394f76d5792e3cac6d_01_1 using rice version 617
24.09.2008 0:07:30|Einstein@Home|Restarting task h1_0500.55_S5R4__138_S5R4a_0 using einstein_S5R4 version 604
24.09.2008 0:07:30|PS3GRID|Restarting task Ars6767-GPUTEST3-5-10-acemd_1 using acemd version 645
24.09.2008 0:07:30|QMC@HOME|Restarting task one_bench20b_s22-ecp2-QZ.2276_0 using Amolqc-preRC1 version 501
24.09.2008 0:07:37|World Community Grid|Restarting task ly482_00101_7 using hpf2 version 603
24.09.2008 0:07:37|World Community Grid|Starting ly491_00015_14
24.09.2008 0:07:37|World Community Grid|Starting task ly491_00015_14 using hpf2 version 603
24.09.2008 0:07:37|World Community Grid|Starting faah4379_000014_MC_xMut_md16150_03_0
24.09.2008 0:07:38|World Community Grid|Starting task faah4379_000014_MC_xMut_md16150_03_0 using faah version 606
24.09.2008 0:07:38|rosetta@home|Starting hombench_mtyka_foldcst_boinc_test1_foldcst_simple_t302___4499_270_0
24.09.2008 0:07:45|rosetta@home|Starting task hombench_mtyka_foldcst_boinc_test1_foldcst_simple_t302___4499_270_0 using minirosetta version 134
24.09.2008 0:07:45|Einstein@Home|Starting h1_0500.55_S5R4__137_S5R4a_0
24.09.2008 0:07:45|Einstein@Home|Starting task h1_0500.55_S5R4__137_S5R4a_0 using einstein_S5R4 version 604
8 tasks are running (!!!), including PS3GRID WU........ So strange...
Let's do rollback to previous version...
|
|
|
|
still not seeing any issues here... |
|
|
|
No issues here either. Mind you, every time I put ncpus to 5, once I start running boinc it goes back to 4.....I know I'm a newbie here, but I'm scratching me head on that one (yes, yes I stop boinc manager and stop the service before I edit client_state.xml).
I have seen some cosmetic bug fixes with 6.3.11 (date format, etc). |
|
|
|
ok, now that it has been running for a while, I see that progress indicator on the gpu grid hasn't changed, while other projects are. so I set the ncps to 4, and then restarted BOINC. Thats when it decided to run 20 work units at once!. Stopped BOINC and then reset the npus back to 5, and it no longer ran 20, but instead just 5. Now, however, it is running Ramsey in high priority, but didn't stop running the other projects. Strange. I am thinking of going back to 6.3.10... |
|
|
|
I was hoping the developers added a function to benchmark the GPU!
So far no difference in the number of tasks as everyone mentioned in this thread. |
|
|
|
For me, 6.3.11 have some improvements!
I can use Boincview normally, and the statistics tab works normally.
Maybe i can see an improvement in how long running the WU,but i will see this tomorrow.
This version have resolve minors bugs, but we can't not use the dedicated core for the GPU to another task.
Waiting for the next release!
Jim PROFIT |
|
|
|
well, downgrading to .10 was a disaster. I lost EVERYTHING! all work, all projects... man am I peeved now. |
|
|
STE\/ESend message
Joined: 18 Sep 08 Posts: 368 Credit: 4,173,502,885 RAC: 27,150,888 Level
Scientific publications
|
well, downgrading to .10 was a disaster. I lost EVERYTHING! all work, all projects... man am I peeved now.
I lost 2 Complete Directories even though I had Backups for them, going to 6.3.11 & back to 6.3.10 neither 1 would recognize I had Work in the Directories even using the Backups so I had to start from scratch again on those 2 Box's ...
|
|
|
|
I later found what happened. for some reason, when I uninstalled v11, it moved all the files data directory to the old boinc folder I used when I was running 5.10.45.... but when I reinstalled 10, it knew where it had been before and just created the data directory as though it was a fresh install. Really messed up. At least I figured out how to get all my work back and am no running on v10 again. |
|
|
GDFVolunteer moderator Project administrator Project developer Project tester Volunteer developer Volunteer tester Project scientist Send message
Joined: 14 Mar 07 Posts: 1957 Credit: 629,356 RAC: 0 Level
Scientific publications
|
I later found what happened. for some reason, when I uninstalled v11, it moved all the files data directory to the old boinc folder I used when I was running 5.10.45.... but when I reinstalled 10, it knew where it had been before and just created the data directory as though it was a fresh install. Really messed up. At least I figured out how to get all my work back and am no running on v10 again.
Usually, before suggesting a new release of BOINC, we test it over here.
There are plenty of things that can go wrong and some releases might even not work at all. If you want to be safe, stick to the BOINC version number that is on our home page, for now, I am afraid it is 6.3.10.
GDF |
|
|
JKuehl2Send message
Joined: 18 Jul 08 Posts: 33 Credit: 3,233,174 RAC: 0 Level
Scientific publications
|
should have tested 6.3.12. wow that version was a masterpiece of programming! Cuda runs fine, but no other tasks are startet - hilarious...
so my advice to everyone is do not install 6.3.11 or 6.3.12 use 6.3.10 instead
I am a software developer and 6.3.12 would not have been even suitable for a daily build for our internal SVN... Who did the QA on that one? |
|
|