Posts by BeemerBiker


log in
1) Questions and Answers : GPU applications : CUDA_FERMI ok to run on non fermi? (Message 1111348)
Posted 1178 days ago by Profile BeemerBiker
I just noticed that a system that has gtx280 and gtx460 is being assigned CUDA_FERMI and both GPUs are getting work. I assume that is OK since I don't see any errors. As I recall, gtx280 is not a fermi.

http://setiathome.berkeley.edu/results.php?hostid=5933773


    setiathome_CUDA: Found 2 CUDA device(s):

    Device 1 : GeForce GTX 280
    totalGlobalMem = 1013776384
    sharedMemPerBlock = 16384
    regsPerBlock = 16384
    warpSize = 32
    memPitch = 2147483647
    maxThreadsPerBlock = 512
    clockRate = 1296000
    totalConstMem = 65536
    major = 1
    minor = 3
    textureAlignment = 256
    deviceOverlap = 1
    multiProcessorCount = 30

    Device 2 : GeForce GTX 460
    totalGlobalMem = 1008271360
    sharedMemPerBlock = 49152
    regsPerBlock = 32768
    warpSize = 32
    memPitch = 2147483647
    maxThreadsPerBlock = 1024
    clockRate = 1350000
    totalConstMem = 65536
    major = 2
    minor = 1
    textureAlignment = 512
    deviceOverlap = 1
    multiProcessorCount = 7

    setiathome_CUDA: CUDA Device 1 specified, checking...
    Device 1: GeForce GTX 280 is okay
    SETI@home using CUDA accelerated device GeForce GTX 280



from WU

http://setiathome.berkeley.edu/result.php?resultid=1922397675

2) Message boards : Number crunching : cuda23 never showed progress on BM (Message 1107449)
Posted 1190 days ago by Profile BeemerBiker
Different system, dual opteron, same problem: no progress unless the task is restarted:



This system has two cards: gtx460 (d0) and a 9800gtx (d1) and is also running that 275.27 and 6.12.26

the above is from boinctasks otherwise the progress would have shown 0.0 %

[EDIT]
I suspect a driver problem. I just logged onto that system, vista 64, and had to dismiss two nvidia driver reset warnings that popped up. A driver reset during crunching possibly is causing this to happen.
3) Message boards : Number crunching : cuda23 never showed progress on BM (Message 1107409)
Posted 1190 days ago by Profile BeemerBiker
Here is another one that hung (never progressed) until I restarted boinc:

From BoincTasks (note the 847% complete as shown by BoincTasks) and runtime of 1 hour, 22 minutes.



This is the same image but from Boincmgr 6.12.26



I restarted boinc and immediately progress was shown under boinc. The task completed in 8 minutes as shown here



I updated the project and the result was posted here

This job ran AFTER I upgrade to that beta nvidia driver 275.27

From looking at the work I see no indication of the over 1 hour run time that unaccountably occured. Something is amiss. This system has only a single gtx280 and a q6700 cpu and is used exculsively for boinc (no games)
4) Message boards : Number crunching : cuda23 never showed progress on BM (Message 1107269)
Posted 1190 days ago by Profile BeemerBiker
This message board was down while I had the same problem with a milkyway job. The milkway jobs were open_cl as opposed to cuda_fermi whatever that is or cuea23. I posted the same problem over at milkyway where the suggestion was to upgrade to the beta driver 275.whatever

Anyway it seems that if I stop boinc and restart then the job either completes immediately with success or restarts at 0.0 and runs the normal 15 or so minutes before completing successfully. Since milkyway deletes results almost as fast as they are posted I can only guess that it was validated. I had a 3 hour 30 minute job (3:25 cpu time) that completed within 20 minutes from scratch (0.0 %) after restarting boinc.

Discussion here

I think the job that had the problem I posted about seems to have problems at all the wingmen. However, I have observed jobs completeing succesfully after restarting boinc, both SETI and MILKY

I have upgraded all my systems to that 275.latest&geatest and will see what happens.
5) Message boards : Number crunching : cuda23 never showed progress on BM (Message 1106446)
Posted 1193 days ago by Profile BeemerBiker
I aborted this task because it never showed any progress after 13 hours of runtime. BM showed 0.00% complete while BoincTasks percent complete varied around 1400.00% It was the 1400 that caught my attention.

This happens very rarely, but when it does it consumes a lot of run time before I notice the problem.

I would think that any project using a GPU for co-processing should implement some type of timeout for "no progress".

I do not know why BoincTasks reported non-zero progress of %1400 for this task while BM showed 0. This indicates that BoincTasks spotted something that BM didnt.

my 2c.
6) Message boards : Number crunching : Watts 'n watts (Message 987453)
Posted 1596 days ago by Profile BeemerBiker
You may find this links interesting:

VDDC Current ?
http://forums.techpowerup.com/showthread.php?t=68526

Power Consumption of Contemporary Graphics Cards
http://www.xbitlabs.com/articles/video/display/gpu-power-consumption-2010_3.html#sect0

.



Thanks for that link on power consumption. I had been looking for something like that.
7) Message boards : Number crunching : Watts 'n watts (Message 987398)
Posted 1596 days ago by Profile BeemerBiker
I scrolled the screen down to where the gpu voltage is displayed and wrote the image back out. It is only 1.1 or so volts, not 12. Hit CTRL-F5 if in IE to refresh the gpu-z image. So about 1.1 * 40 amps = just over 44 watts peak. 40 amps * 12 would be too high a wattage and that made me go back and re-examine gpu-z I assume there is significent efficiency loss converting from 12 down to 1.1. I am no engineer but I know ohms law. Since this unit is used only for boinc then the video load is 0.0 and I assume the current requirement is much more if gaming.
8) Message boards : Number crunching : Low power crunchin'? (Message 987378)
Posted 1596 days ago by Profile BeemerBiker
Ok, I have a question, it came to me while playing around with the idea of building one of these just for fun, to see how it performs. I see that the Atom 330 boards list an external power supply, something like 65 watts or such. It says it has a 16x PCI-E slot, and the question I had was isn't there some spec about how much power a 16x PCI-E slot is supposed to put out, because if one was thinking (just for fun) of putting in, let's say, a Nvidia GTX 285, obviously you would need an external PS to power the 6/8 pin connectors, but isn't there a certain minimum amount of power needed to power the PCI-E slot, and is a 60 some watt external PS enough to do it? Or hadn't they thought that someone like me would be crazy enough to attempt to put the 2 together? ;)


You could rig a powersupply like this one. I added a 12 volt relay that switched the power supply on when the primary powersupply was turned on. This saved me a few bucks as I had a nice 400 watt supply and didnt want to get a 750 watt for my two gts250s. I bought two 8 pin molex extension cables, cut the female ends off and connected them to the output of the power supply.

NOTE: If you do something like that, be sure not to connect any pci-e cables from the primary supply to the same card(s) that have that aux power or you can get reverse current flow transients. The remote control relay is also a must or the reverse current flow will not just be a transient.
9) Message boards : Number crunching : Watts 'n watts (Message 987375)
Posted 1596 days ago by Profile BeemerBiker
I do not have a wattmeter tho I do get a bill at the end of the month.

However, I do have gpu-z and it looks like I am drawing between 21 and 41 amps while crunching seti. I recall milkyway gave a higher average like 32 - 40. Collatx did not pull as much as milkway I think.



The above current is for my gtx280 with its 64nm core size. I would be interested in what your gtx285's are drawing as they have a smaller die.

Not all nvidias have current sensors. I cant get a valid reading on a gts250 or a 9800gtx+
10) Message boards : Number crunching : BOINC 6.10.43 version released (Message 987341)
Posted 1596 days ago by Profile BeemerBiker

=========try the following====

From the cmd prompt of the windows system with the old boinc version, or from the terminal prompt of the older linux system (what.ever is the ip address of the newer system)

telnet 192.168.what.ever 31416

if you see "unable to connect to the remote .." then you got a firewall problem

if you see a bunch of strange characters and cant seem to exit then you connected but are not authorized


Well, I don't have telnet available as a DOS command with this Vista 32


It can be enabled in windows components.
11) Message boards : Number crunching : BOINC 6.10.43 version released (Message 987319)
Posted 1596 days ago by Profile BeemerBiker
I believe this is a firewall problem.

I can connect from linux 6.10.17 to any linux (6.10.43, 6.10.44) or windows xp (6.10.43)

All of the above were launched using -allow_remote_gui_rpc

I can connect to all of my vista 64 and win 7 systems (6.10.43) if I explicitly put the ip address of the 6.10.17 system into

\ProgramData\Boinc\remote_hosts.cfg

*AND* I either turn off the firewall or configure the exclusion properly.

Excluding "boinc.exe" with no path to the executable will not work.

Excluding "\Program Files (x86)\boinc\boinc.exe" will not work if you are running the 64 bit boinc. I have no idea how that 32 bit location got into my firewall.

======turn off the firewall and try the following====

From the cmd prompt of the windows system with the old boinc version, or from the terminal prompt of the older linux system (what.ever is the ip address of the newer system)

telnet 192.168.what.ever 31416

if you see "unable to connect to the remote .." then you got a firewall problem

if you see a bunch of strange characters and cant seem to exit then you connected but are not authorized
12) Message boards : Number crunching : your least liked motherboard brand (Message 973176)
Posted 1640 days ago by Profile BeemerBiker
Asus K8N-DRE - This board was too good to be true. It only cost $12.99 at compgeeks and came with a PCI-E slot. I had a spare pair of dual core opteron's and a bunch of old PC2100 R ECC that I was not using so I ordered it to find that it would only take a 2/3 size card. I could not find an ATI card that was 188mm or under that did double precision. I did manage to get a GTS-250 to fit in the slot and it cleared by 4mm.



a writeup is here and it does run ubuntu 9.1 well.
13) Questions and Answers : GPU applications : where can I get help for cuda on ubuntu? (Message 969977)
Posted 1650 days ago by Profile BeemerBiker
Does BOINC see the drivers/libraries (CUDA 2.2) too?

Has BOINC access rights for the video device files in /dev?

Gruß,
Gundolf


/dev/nvidia(dont remember exact name) showed "video" for a group. I went to users and groups and there was no user "video" nor even a group "video". I did see a user "boinc" but there was no video group to put him/her into. I went and changed ownership of /dev/nvidia(dont remember) to root:root since the other nvidia entry in /dev had root:root and not root:video

that didnt work nor did a reboot. I quickly discovered that a simple reboot lost the nvidia driver and I had to rerun the install on each reboot.

what did work just fine was 9.1 tho it took me 4 hours to download on my slow dsl. I am currently crunching gpugrid but not seti nor collatz for some reason. I have seen that before and I assume they will start a gpu task eventually.

I installed the ALT-CD 9.1 x64 from the iso download, got that "ia32" library, got that "build-essential" stuff that nvidia needed, then installed the nvidia driver 190.53, then ran the boinc 6.10.32 which all it did was unpack and left me to figure out how the service worked. Fortunately, getting the 6.4.5 thru apt-get sets up the service so I copied the new version 10.32 ones over the old ones and stuffed that cudalibart somewhere that google suggested as I knew it would not do me any good on my desktop.

anyway, it is working :-)
14) Questions and Answers : GPU applications : where can I get help for cuda on ubuntu? (Message 969938)
Posted 1651 days ago by Profile BeemerBiker
I put a GTS 250 in the latest hardy (8.04) and installed that 190.53 from nvidia.

The nvidia settings work, display looks great, etc. All seems to be working and I see the cuda libs where they are supposed to be but boinc 6.10.17 x64 can find no gpu.

I am guessing the card is too new for 8.04 and I edited nv.pci and put in the 0x615 id for gt-250 and reran the install, but that didnt help with the "no GPUs".

I would like to know if anyone has the gts 250 (or any of the newer nvidia cards) working with 8.04.

Do I need to upgrade from 8.04 to 9.1?

thanks for looking
15) Questions and Answers : GPU applications : ?Checked but no consensus? (Message 969264)
Posted 1654 days ago by Profile BeemerBiker
I see this rarely as at here where my result is the lunatic mod shown at the top (the anonymous platform). It seems that the only important difference is the estimated credits. Note that the other two are both cuda systems and use different CoProcessors. One would expect all 3 systems to be slightly different.

here is another one but there was no need for a consensus check and both wu's show the same claimed credit. Note that neither system used the CUDA but the cpu's are different.

So what causes a 3rd computer to be called on for a consensus? Is it because the claimed credits are different?

If I had used the standard (non-lunatic) version would there have been a need for a 3rd consensus?

16) Questions and Answers : Web site : server status incomplete for CUDA (Message 926270)
Posted 1831 days ago by Profile BeemerBiker
In looking at the server status page, I cannot tell how many cuda work units are available to be downloaded. Neither can I spot any info about CUDA tasks.

If there is no work available at all, should not there be something like

WU's ready to be download (SETI / SETI-CUDA): 0 / 0

and something similar for astro pulse?

I do see the row item "Results ready to send" and 0 for both seti and astro. I assume Results are what is to be downloaded by users and Workunit to be stuff that is uploaded when the user finished processing the Result. From a user point of view, that seems backwards does it not?

I have 3 systems running seti cuda, all 6.6.36, and the status messages seem inconsistent. Note that only the last system goes on to request work from the GPU. The other two systems seemingly only ask for CPU work and I did not find any further reference to getting GPU work further down in the message log. i.e.: I did not selectively cut and paste.

System 280 Vista64
8/15/2009 8:15:27 AM SETI@home Sending scheduler request: To fetch work.
8/15/2009 8:15:27 AM SETI@home Requesting new tasks for CPU
8/15/2009 8:15:32 AM SETI@home Scheduler request completed: got 0 new tasks
8/15/2009 8:15:32 AM SETI@home Message from server: (Project has no jobs available)


System 9800#1 Vista64
2009-08-14 4:46:11 AM SETI@home Sending scheduler request: To fetch work.
2009-08-14 4:46:11 AM SETI@home Requesting new tasks for CPU
2009-08-14 4:46:16 AM SETI@home Scheduler request completed: got 0 new tasks
2009-08-14 4:46:16 AM SETI@home Message from server: No work sent
2009-08-14 4:46:16 AM SETI@home Message from server: CPU jobs are available, but your preferences are set to not accept them


System 9800#2 Windows-7
8/13/2009 10:15:46 AM SETI@home Sending scheduler request: To fetch work.
8/13/2009 10:15:46 AM SETI@home Requesting new tasks for CPU
8/13/2009 10:15:51 AM SETI@home Scheduler request completed: got 0 new tasks
8/13/2009 10:15:51 AM SETI@home Message from server: (Project has no jobs available)
8/13/2009 10:25:07 AM SETI@home Sending scheduler request: To fetch work.
8/13/2009 10:25:07 AM SETI@home Requesting new tasks for GPU
8/13/2009 10:25:12 AM SETI@home Scheduler request completed: got 0 new tasks
8/13/2009 10:25:12 AM SETI@home Message from server: (Project has no jobs available)
17) Questions and Answers : GPU applications : Windows 7 CUDA no go (Message 924022)
Posted 1841 days ago by Profile BeemerBiker
Got it working. I uninstalled, then reinstalled and the cuda device was recognized. Go figure!
18) Questions and Answers : GPU applications : Win7 and CUDA - 9800GX2 (Message 924021)
Posted 1841 days ago by Profile BeemerBiker
I cannot get win7 RC 7100 to work with my gtx9800+ and boinc 6.6.36. It simply does not see the cuda device. Two other vista-64 systems work fine.


I uninstalled, then reinstalled and the cuda device was recognized. Go figure!
19) Questions and Answers : GPU applications : Win7 and CUDA - 9800GX2 (Message 923886)
Posted 1841 days ago by Profile BeemerBiker
I cannot get win7 RC 7100 to work with my gtx9800+ and boinc 6.6.36. It simply does not see the cuda device. Two other vista-64 systems work fine.
20) Questions and Answers : GPU applications : Windows 7 CUDA no go (Message 923744)
Posted 1841 days ago by Profile BeemerBiker
According to this thread my system should be working, but it is not. I just setup the release candidate, 7100, with all patches, downloaded the latest windows 7 from nvidia: 190.38_desktop_win7_winvista_64bit_english_whql and installed boinc 6.6.36 64bit under my account (not as a daemon) but I cant make it work with my 9800gtx+ It just does not see it.

8/5/2009 12:33:34 AM Starting BOINC client version 6.6.36 for windows_x86_64
8/5/2009 12:33:34 AM log flags: task, file_xfer, sched_ops
8/5/2009 12:33:34 AM Libraries: libcurl/7.19.4 OpenSSL/0.9.8j zlib/1.2.3
8/5/2009 12:33:34 AM Data directory: C:\ProgramData\BOINC
8/5/2009 12:33:34 AM Running under account jstateson
8/5/2009 12:33:34 AM Processor: 4 AuthenticAMD Dual Core AMD Opteron(tm) Processor 270 [AMD64 Family 15 Model 33 Stepping 2]
8/5/2009 12:33:34 AM Processor features: fpu tsc pae nx sse sse2 pni
8/5/2009 12:33:34 AM OS: Microsoft Windows 7: x64 Edition, (06.01.7100.00)
8/5/2009 12:33:34 AM Memory: 3.00 GB physical, 6.00 GB virtual
8/5/2009 12:33:34 AM Disk: 149.04 GB total, 131.67 GB free
8/5/2009 12:33:34 AM Local time is UTC -6 hours
8/5/2009 12:33:34 AM No CUDA devices found
8/5/2009 12:33:34 AM No coprocessors
8/5/2009 12:33:34 AM Not using a proxy
8/5/2009 12:33:34 AM AQUA@home URL: http://aqua.dwavesys.com/; Computer ID: 23130; location: (none); project prefs: default
8/5/2009 12:33:34 AM Milkyway@home URL: http://milkyway.cs.rpi.edu/milkyway/; Computer ID: 94599; location: (none); project prefs: default
8/5/2009 12:33:34 AM General prefs: from http://bam.boincstats.com/ (last modified 19-Jan-2009 16:18:29)
8/5/2009 12:33:34 AM Host location: none
8/5/2009 12:33:34 AM General prefs: using your defaults
8/5/2009 12:33:34 AM Reading preferences override file
8/5/2009 12:33:34 AM Preferences limit memory usage when active to 3070.60MB
8/5/2009 12:33:34 AM Preferences limit memory usage when idle to 3070.60MB
8/5/2009 12:33:34 AM Preferences limit disk usage to 74.52GB
8/5/2009 12:33:34 AM AQUA@home Restarting task 04aug09-128-5M-64-a_52_23_0 using AQUA version 235


what I have tried: Removed myself from administrator group but could not then run the manager. Tried installing from a non-admin but it wont install unless admin. I just put system together yesterday with XP (it worked fine) then decided to go for 64bits and windows 7.

There is nothing informative in the event log. I do not see any of the nvidia tools that normally get installed as on my two vista 64 systems. All I see is the nvidia's "3d vision" under All Programs.

thanks for looking


Next 20

Copyright © 2014 University of California