Posts by S@NL - eFMer - www.efmer.eu/boinc


log in
21) Message boards : Number crunching : Optimize your GPU. Find the value the easy way. (Message 1328412)
Posted 587 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor

So ... what am I not doing correctly? (Yes, I stopped BOINC Manager before executing the performance utility).

Look here:
C:\Users\username\AppData\Roaming\eFMer\SetiPerformance\slot0

stderr.txt should give you a clue.

The GT 520 may not be able to run the g version.
22) Message boards : Number crunching : Changing GPU - But which will be better? (Message 1326387)
Posted 594 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
I like to make some comments.
The 6xx series has a lot more cudas, but more isn't always better.
6xx series is more power efficient,that's a good thing.
But the cuda cores are a lot lot less powerful.

As you can see some 5xx cards outperform the 6xx series.
But I expect there will be some gain in optimizing for the 6xx series
23) Message boards : Number crunching : "Sleepy" NV GPU - how to "wake up" ? (Message 1309218)
Posted 643 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
I`m using MSI Afterburner.
Works on nvidia also.

Thanks, Mike. Downloading...
EDIT: looks like this utility is successor of RivaTuner too.

MSI Afterburner works as a service? With out need to log on any user?

You need to log in.
GPU access is only allowed for a logged in user. So no service.

Windows Server 2003 (see opening post) is comparable to Windows XP - and Windows XP allows services to access the GPU, even when the service logs on using a different account than the foreground user.

OK Thanks for the info, almost forgot that XP still exists.
But anyway Afterburner is a program, so normally that should only start at login.

24) Message boards : Number crunching : "Sleepy" NV GPU - how to "wake up" ? (Message 1309209)
Posted 643 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
I`m using MSI Afterburner.
Works on nvidia also.

Thanks, Mike. Downloading...
EDIT: looks like this utility is successor of RivaTuner too.

MSI Afterburner works as a service? With out need to log on any user?

You need to log in.
GPU access is only allowed for a logged in user. So no service.
25) Message boards : Number crunching : "Sleepy" NV GPU - how to "wake up" ? (Message 1308376)
Posted 644 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
Is it possible to flash required fan settings directly in GPU own BIOS memory?


Maybe you can try this instead:
http://www.nvidia.com/object/nvidia_system_tools_6.02.html

It's got a fan speed setting and the tools are tied into the NVIDIA Control Panel and can be made to work at startup, but I'm not sure about your Windows version.

Works nice on older cards.
It's a legacy program, they stopped developing a while.
26) Message boards : Number crunching : "Sleepy" NV GPU - how to "wake up" ? (Message 1308366)
Posted 644 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
Thanks all a lot!
As I understood this problem worth some consideration indeed. Auto-logon would be not good cause it's not home PC indeed.
Is it possible to flash required fan settings directly in GPU own BIOS memory ?

You should be able to use manual control with the slider. (Disable the auto mode)
This should stick, at least it did on my system.

The last driver seems to down clock some cards. Seems to happen close to 80C.
Probably the keep the temps in a safe zone.
27) Message boards : Number crunching : Switched to Einstine (Message 1307735)
Posted 646 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
I can't say I am happy with the current state of things here on my favorite project, but I do know that they shall get better. After all, with over 13 years under my belt here, I have seen a lot of ups and downs.

Helps is on its way, the simple trick. Make the workunits a lot longer to process, by a factor of 4 or so.
This will give us a year or so.......
28) Message boards : Number crunching : Calculation time GTX 580 vs GTX 680 (Message 1307709)
Posted 646 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
Hello,

I did an interesting finding on the calculation time reported by SETI. I have now in the same computer both GTX 580 and GTX680.

The finding is that for GTX 580 the reported calculation time for one WU is roughly 1100 seconds while for the GTX 680 the calulation time averages around 250 (4,2 minutes) seconds.

When checking the Boinc Manager Tasks tab on active work Units the average calculation time for a WU is roughly 20 minutes (=1200s) (3 WU concurrently per card). So the calculation time reported is more in line with the GTX 580.

Why is the reported calculation time for GTX 680 differing so much?
Is GTX 680 idling somewhere and this is not reported in correctly?


http://setiathome.berkeley.edu/results.php?hostid=3299266&offset=0&show_names=0&state=2&appid=

Here is a list:
http://www.efmer.eu/forum_tt/index.php?topic=981.0
Try the program and report back the best time.
29) Message boards : Number crunching : M$ pushing nVida drivers as an 'Important' update... (Message 1307364)
Posted 647 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
The new driver has the adaptive mode switched on and whatever you do it stays on.
Down clocking the card all the time.

Switched back to 301.42 and the adaptive mode can be switched off as it should.
30) Message boards : Number crunching : M$ pushing nVida drivers as an 'Important' update... (Message 1307351)
Posted 647 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
Hmm that resulted in setting SLI mode.
Pushing my GPU's back to 33% use instead of 98%
31) Message boards : Number crunching : Optimize your GPU. Find the value the easy way. (Message 1307064)
Posted 649 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
V 1.7 with a small bug fix, now stops as soon as the best result has been reached.

http://www.efmer.eu/forum_tt/index.php?topic=974.0
32) Message boards : Number crunching : Optimize your GPU. Find the value the easy way. (Message 1307062)
Posted 649 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
Hello Fred, thank you very much for the program i use it from the first version you release.
I want to ask you which driver you use to complete 130sec for WU for GTX 295 and which WU from the list because my top time with 3 drivers tested same clock everything the same the best time for Core755/Shaders1655/Mem1255 is 188-190 sec with x41g for wu.
and my max settings are 805/1733/1255 the time for PG0444_v7 unit need 179 sec.

Driver 280.26
The Wu is the one in zip\x41g
Using default settings 576 / 1008 / 1242 "Factory shipped frequencies"
33) Message boards : Number crunching : Separate work for GPU and CPU? (Message 1296121)
Posted 680 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
Agree my situation is simple with one host and one gpu. The project preference "use NVIDIA (or ATIa) gpu" only tells BOINC to not ask for gpu work. Turning it off will cause BOINC to only ask for cpu work. Definately not an automated approach - hope you find a better way.

I tried to automate something like that.
But even though the computer asks for CPU work it gets noting.
Another computer that request in the same period, gets work ever single time.
But just enough to keep the GPU buffer filled.....

Probably there is something wrong with the BOINC scheduler. At least for computers with more GPU's
34) Message boards : Number crunching : Separate work for GPU and CPU? (Message 1296119)
Posted 680 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
I've noticed that 7.0.28 will keep asking for GPU work until it gets to about 1.25 days more than my min + additional setting, and even at that level it will ask for a few thousand seconds of GPU work. I use local preferences of 6 and .01. I don't get much CPU work until it gets to 7.25 days of GPU work per BOINCTasks' sum of estimated times. You can intervene and temporarily turn off GPU fetch in the SETI@Home project preferences (website). I have do this about once a week to rebalance the cache. Of course, that would affect all of your hosts unless you are using venues (home, school, etc).

That's what I've been trying to automate. That's when you don't get any work at all.
No CPU that way.

But I expect you get no problems if you have one GPU.
35) Message boards : Number crunching : Separate work for GPU and CPU? (Message 1296116)
Posted 680 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
And set at least the <sched_op_debug> logging flag in cc_config.xml so that you can see exactly what's going on when work is requested/received.

Done:

68079 SETI@home 17-10-2012 12:19 [sched_op] estimated total NVIDIA task duration: 7555 seconds
68078 SETI@home 17-10-2012 12:19 [sched_op] estimated total CPU task duration: 0 seconds
68077 SETI@home 17-10-2012 12:19 Project requested delay of 303 seconds
68076 SETI@home 17-10-2012 12:19 [sched_op] Server version 701
68075 SETI@home 17-10-2012 12:19 Scheduler request completed: got 22 new tasks
68074 SETI@home 17-10-2012 12:18 [sched_op] NVIDIA work request: 6868.66 seconds; 0.00 devices
68073 SETI@home 17-10-2012 12:18 [sched_op] CPU work request: 2080512.00 seconds; 8.00 devices

BOINC request work and gets GPU work just above the value needed.

Gets 7555 needed 6868.
But there I expect it goes wrong. 687 is 76 for GPU to handle.
So it gets just about the GPU it needs and hoovers just below the threshold.

Is should get at least number of GPU * scheduler interval. + something extra. 9 * 6 * 60 = 3240 it gets 687.

I see this line:

68074 SETI@home 17-10-2012 12:18 [sched_op] NVIDIA work request: 6868.66 seconds; 0.00 devices
Does it know how many GPU tasks are running ?????
36) Message boards : Number crunching : Separate work for GPU and CPU? (Message 1296102)
Posted 680 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
After correction that problem in BoincTasks, the results are still the same.

No GPU work.

After changing the buffer from 2 -> 3 I got way over 1000 WU in a singe hour.
All GPU of course.

Now at 3 days.

67320 SETI@home 17-10-2012 10:56 Scheduler request completed: got 0 new tasks
67319 SETI@home 17-10-2012 10:56 Reporting 1 completed tasks, requesting new tasks for CPU
67318 SETI@home 17-10-2012 10:56 Sending scheduler request: Requested by user.

It is getting no CPU work.

The 6.12.33 has plenty of CPU work and gets is regularly.
37) Message boards : Number crunching : Separate work for GPU and CPU? (Message 1296091)
Posted 680 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
But I think I see what's going wrong.

Maybe that is....

V7 reused

Connect every
Additional work buffer

->

Minimum work buffer.
Additional work buffer.

So I set the Additional work buffer as in V6, but I probably have to use:

Minimum work buffer in V7.
38) Message boards : Number crunching : Separate work for GPU and CPU? (Message 1296089)
Posted 680 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
And what do you see in message log? Does BOINC client asks for CPU work and doesn't get it or it doesn't ask for CPU work?


As you can see, it asks for CPU work only.

I was building an additional in BoincTasks.
If GPU work > 2 day and CPU work < 0.5 day set buffer to 2 days.

That's when I noticed the buffer staying on 2 days for a couple of days and no CPU work except for some AP work.
39) Message boards : Number crunching : Separate work for GPU and CPU? (Message 1296086)
Posted 680 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
I've notices the CPU is not getting any work at all.
I know the GPU gets work first.

But I set a Buffer for 5 days, no CPU work, but the buffer is filled to 5 days.
Next I dropped the buffer to 2 days. BOINC is requesting CPU work.
Stayed there for a couple of days. The buffer fills to 2 days and stays there.

Now I set the buffer back to 3 days and I immediately get new work.
And a lot of it, of course all GPU.

This happens with 7.0.28 and 7.0.31.

At 2 days

64836 SETI@home 17-10-2012 09:03 Scheduler request completed: got 0 new tasks
64835 SETI@home 17-10-2012 09:03 Reporting 5 completed tasks, requesting new tasks for CPU
64834 SETI@home 17-10-2012 09:03 Sending scheduler request: Requested by user.

64874 SETI@home 17-10-2012 09:13 Scheduler request completed: got 42 new tasks
64873 SETI@home 17-10-2012 09:13 Requesting new tasks for CPU and NVIDIA
64872 SETI@home 17-10-2012 09:13 Sending scheduler request: Requested by user.


64977 SETI@home 17-10-2012 09:18 Scheduler request completed: got 72 new tasks
64972 SETI@home 17-10-2012 09:18 Reporting 3 completed tasks, requesting new tasks for CPU and NVIDIA
64971 SETI@home 17-10-2012 09:18 Sending scheduler request: To fetch work.
40) Message boards : Number crunching : Optimize your GPU. Find the value the easy way. (Message 1294078)
Posted 685 days ago by Profile S@NL - eFMer - www.efmer.eu/boincProject donor
V 1.6 with some minor fixes.

http://www.efmer.eu/forum_tt/index.php?topic=974.0

According to the developer, the AP application can't give a progress....
But I made some changes anyway.


Previous 20 · Next 20

Copyright © 2014 University of California