Posts by Zydor

\n studio-striking\n
21) Message boards : Number crunching : Forcing work on one GPU only (Message 2449)
Posted 30 Jan 2012 by Profile Zydor
Post:
Thats what Mikey pointed you at, that app_info puts one per gpu.

This is the one I am using with 2x5970s, they run one Moo WU per GPU

<app_info>
<app>
<name>dnetc</name>
<user_friendly_name>Distributed.net Client</user_friendly_name>
</app>
<file_info>
<name>dnetc_wrapper_1.3_windows_intelx86__ati14.exe</name>
<executable/>
</file_info>
<file_info>
<name>dnetc518-win32-x86-stream.exe</name>
<executable/>
</file_info>
<file_info>
<name>dnetc-gpu-1.3.ini</name>
</file_info>
<file_info>
<name>job-ati14-1.00.xml</name>
</file_info>
<app_version>
<app_name>dnetc</app_name>
<version_num>103</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.1</avg_ncpus>
<max_ncpus>0.5</max_ncpus>
<plan_class>ati14</plan_class>
<flops>1157115231469.729200</flops>
<file_ref>
<file_name>dnetc_wrapper_1.3_windows_intelx86__ati14.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>dnetc518-win32-x86-stream.exe</file_name>
<copy_file/>
</file_ref>
<file_ref>
<file_name>dnetc-gpu-1.3.ini</file_name>
<open_name>dnetc.ini</open_name>
<copy_file/>
</file_ref>
<file_ref>
<file_name>job-ati14-1.00.xml</file_name>
<open_name>job.xml</open_name>
<copy_file/>
</file_ref>
<coproc>
<type>ATI</type>
<count>1</count>
</coproc>
</app_version>
</app_info>

Create it inside Notepad (no other editor - use Notepad), and save file as a txt format file called app_info.xml, into the Moo Project Directory. Dont forget to run down the cache before using it else you will trash existing WUs. BAM must be closed down and restarted for the app_info to take effect.

EDIT:
In an app_info, the part that contrrols number of WUs running on the gpu is this bit:

<coproc>
<type>ATI</type>
<count>1</count>
</coproc>

A count of one means one per GPU. If it was put as 0.5, it would be 2 per GPU .... and so on, so for example(s)

Count of ...

1 is 1 per GPU
0.5 is two per GPU
0.33 is three per GPU
0.25 is 4 per GPU.

Do the Maths you'll get the idea

(Dont run more than two Moo WU per GPU, utterly pointless, just creates other issues for zero purpose, and usually one per GPU in a multi GPU setup solves the fragmentation / utilisation problem.)

Watch the heat carefully when first starting this, expect to have to downclock GPUs by about 30 on previous setting to keep heat under control - more work flying around, hence the need to reduce GPU speed slighly to compensate - be careful ... you will overheat if you dont watch this the first time you use the app_info.

Regards
Zy
22) Message boards : Number crunching : Utilization of HD6950 GPU (Message 2419)
Posted 28 Jan 2012 by Profile Zydor
Post:
Just confirming what you said already .... the 12.1 drivers are good to go if anyone contemplating a move away from the dubious ridden later 11.XX drivers.

Murphy may surface, but, I dont think so as a general propostion. There will always be some issue somewhere, but 12.1 really does seem solid after a few days use, and I have not seen excessive bad press about it (given some gaming issues, but they are always ongoing)

For BOINC they do seem fine .... give them a go.

Regards
Zy
23) Message boards : Number crunching : Back Up To Speed - Thanks Teemu :) (Message 2415)
Posted 27 Jan 2012 by Profile Zydor
Post:
:)

Ya still get the plaudits for the time spent though .... :)

Regards
Zy
24) Message boards : Number crunching : Back Up To Speed - Thanks Teemu :) (Message 2412)
Posted 27 Jan 2012 by Profile Zydor
Post:
Well ..... Murphy had a fun time :)

All seems good this end, back to its normal response et al.

Its times like this when efforts of Admin's like Teemu are appreciated.

*Tips' Hat* .... Thank You Teemu :)

Regards
Zy
25) Message boards : Number crunching : Runtimes with 1.3 version (Message 2393)
Posted 24 Jan 2012 by Profile Zydor
Post:
Overnight the results on the HD5850 are run times @ 1200 seconds (was 2,300) and credit 1750, was @6.5K to 7.0K.

Weider and weirder.


Have a look at file sizes, I cant check, PCs are hidden - particularly the numbers of Stat Unit Groups etc. Chances are you are getting the smaller WUs because you reset the Project, it will eventually reset back to bigger ones when the adaptive mechanism it uses gets used to the fact you have a card capable of bigger WUs.

Regards
Zy
26) Message boards : Number crunching : Runtimes with 1.3 version (Message 2392)
Posted 24 Jan 2012 by Profile Zydor
Post:
I have been having the same prblm w/my 5870's, Lg wu's? fragmentation? 32 min wu times. Ran thru CCC drivers, setteled on 11.9 for both. Where is this file? or do I need to try the app_info? On the Moo preferance page I changed ATI core to 3, but I dont think thats whats being talked about?


Change Moo preferences back to what they were - if forgotten, change to defaults (values are shown there).

What is being talked about is a value inside the project ini file in the Project Directory under "Program Data/Boinc/Projects/ .... moo

In there you will see a file called "dnetc-gpu-1.3ini"

Look just under halfway down and you will see


[processor-usage]
priority=9

If you want to test to see if a difference, change 9 to a 4. I have done this in the past, and its being reset to 9, seems not possible to stay at 4, not sure .... anyway, test it at 4 see what happens to RAC. Keep an eye on it to see if it resets itself to 9.

The purpose of the ini file is to present specific generic graphics card options to moo WU, that one inside the file you are changing, specifies the level of priority between GPU & CPU. It was 4 in the past, there is now a suspicion 9 is too much and 4 may have been ok.

Caution needed on rushed deductions, need evidence, but the more try the 9 to 4 change, the more solid the deduction. So at present its a suspicion only, need more to see if it makes a difference.

Regards
Zy
27) Message boards : Number crunching : Runtimes with 1.3 version (Message 2388)
Posted 24 Jan 2012 by Profile Zydor
Post:
Pretty well seals it I reckon, the default of 9 needs revisiting, its not looking to be a good option as a default value.

Regards
Zy
28) Message boards : Number crunching : Runtimes with 1.3 version (Message 2386)
Posted 23 Jan 2012 by Profile Zydor
Post:
Whoa .... thats set the cat amongst the pigeons ...... is there anyone else can reproduce that ?

Regards
Zy
29) Message boards : Number crunching : Runtimes with 1.3 version (Message 2382)
Posted 23 Jan 2012 by Profile Zydor
Post:
Last night I crunched a bunch of Wus in 1 hour 20 minutes and see the dnetc-gpu-1.3 file has the processor priority at 9, so I cannot do anything.


There's definitely something wacky going on - thats similar to Dan's problem in the OP - and he has been through nothing short of heroic efforts to try and find the issue, so far to no avail. He has 5870s, you are reporting it on 5850s, and I dont get the issue on twin 5970s.

This one is baffling.

Try the app_info file - even for single GPUs [just put 1 as the number of GPUs in the app_info] - and see if explicit instructions from app_info overcomes the problem

EDIT: Just had a thought, bit wacky but ..... ,I did have an issue with it being set to 9 when the option first came out. Many reported it being quicker, but I found it slower, so I always had it set to 4, which was the old default. On change to V1.3, I went in and changed mine from the new default of 9 back to 4. Its a long shot .... but .... I did have slow down issues at 9 wa y back when it first came in, although nothing like reported at the OP - try it at 4 not 9 and see what happens - bit of a wild card, but you never know, and easy to check.

Regards
Zy
30) Message boards : Number crunching : 7970s Dont Run (Message 2381)
Posted 23 Jan 2012 by Profile Zydor
Post:
There's a fair chat developed over on MW in their "7970 Dont Run" thread, due, I suspect, to the NVidia OpenCL app running there on 7970s, a steady flow of tidbits starting to appear in it from other Projects.

Regards
Zy
31) Message boards : Number crunching : Runtimes with 1.3 version (Message 2372)
Posted 22 Jan 2012 by Profile Zydor
Post:
I know this will sound wacky .... but hey we are at that stage .... go try a memory test on main system memory, an extended one not just a quickie. I know its unlikely, but different apps use memory in different ways. I have seen it before that an app will run, but another will not, turned out to be a memory stick had gone bad.

I am starting to go more and more towards hardware somewhere, albeit I know a Logic shows it cant be. However you've changed just about every bit of software there is, and no go - apart from a Windows reinstall, and I guess there is a far fetched logic that says try a windows reinstall ..... but its a hell of a lot of work for something so unlikely.

Give the memory test a whirl, doesnt take much, and at least that would be eliminated if nothing else. (The Windows built in memory checker is not bad, run it as a first check, but make sure you retest as well with a good third party test software as well)

Regards
Zy
32) Message boards : Number crunching : App_info..xml (Message 2370)
Posted 22 Jan 2012 by Profile Zydor
Post:
Keep it there for a while - the mechanism used for sizing WUs is adaptive. It will (or should ...) end up detecting the card capability, and be back to previous sizes, just takes time to adapt.

Additional Note to my Post on 5970s:
5970 Cards should stick to one per GPU - two per is usually too much for it. Might get away with two per on single 5970s, but those running twin 5970s will grind to a halt with 8 WUs running. One per GPU on single 5970 cards is fine and alleviates nearly all the problem.

Regards
Zy
33) Message boards : Number crunching : 7970s Dont Run (Message 2367)
Posted 22 Jan 2012 by Profile Zydor
Post:
OoooooO :)

Well spotted ..... had a look, looks interesting. Will be good to see how that sits with the PrimeGrid Devs, they are a good bunch, theyll nibble that to death and get it going. Be interestiung to run it once the Devs give the go ahead with a PG cleared version.

Things are starting to pick up now re 7970s as the problems are identified and software tweeked. Shouldnt be too long before we can unleash these things at full speed on a good number of Projects, at present they run on OpenCL apps - and even thats impressive, speeds are incredible for OpenCL even running NVidia OpenCL compiled apps.

Usual story with new genration cards .... takes time :)

Regards
Zy
34) Message boards : Number crunching : Runtimes with 1.3 version (Message 2365)
Posted 22 Jan 2012 by Profile Zydor
Post:
.... He is still running version 11.11 of the drivers, aren't they still not good for crunching and shouldn't he be at 11.9 or less right now.


As a general proposition, yup, best stay away from the later ones. However its not absolute, they will run, I am running 11.12 on my main box. He is getting massive slowdown - its huge - that cant be the driver, too big for that (assuming the driver install is a good one). Hence thoughts over maybe a bit left from V1.2 somewhere.

Your second point may have mileage in it though ... its a good thought .... the new options should be used with care. If they have been played with, I'd reset them all to -1 (the default) and see what happens on the WU. Same goes for the .ini file, the default setting of 9 for CPU doesnt work for everyone (I reset mine back to V1.2 level of 4).

Its worth a shot as you indicate .... would be interesting to know if its was that in the end, because this one is real, real, unusual, something wacky is going on.

Regards
Zy
35) Message boards : Number crunching : Runtimes with 1.3 version (Message 2355)
Posted 21 Jan 2012 by Profile Zydor
Post:
Bit of a loss as to what to go for to be honest - anyone any ideas fire away :)

Due to other projects running fine, Hardware is ok, BOINC is ok, driver is ok. You've detatched/attatched, so project files are ok. The application is fine - run thousands of times by now on 58XX cards, and in any case you've crunched it 50 times successfully, so its not the app.

The only thing I can thing of is there is lurking a file from V1.2 still hanging around causing a clash, or the registry is messed up.

The last shot in my bag, is do another atatch/detatch - and immediately afterwards, before running anything, do a registry clean. Virus check maybe, and make sure the project directory is an exception on the virus checker on standard running ..... but thats clutching at straws frankly.

Regards
Zy
36) Message boards : Number crunching : App_info..xml (Message 2353)
Posted 21 Jan 2012 by Profile Zydor
Post:
Suspending the test of this until I see what 1.03 can do, no point going ahead if 1.03 solves the ills, albeit the change log doesnt seem to get to these issues (not surprisingly).


V1.3 is solid on the PC so I restarted the test with twin 5970s.

Its worth doing with 5970s. The mileage you'll get - as always with infinite variation on individual setups - will vary, probably considerably given two gpus on one card. Overall it looks like around 30 secs to a min per WU, hard to be specific, some were greater savings, the fragmented nature of the WUs means its hard to estimate without any kind of reliable pole to revolve around. Other cards will likely see a better increase as they will not have to fight the VRM heat problem. But, yup, its worth doing with 5970s. Use the app_info above in this thread (dont forget to change version statement from 130 to 103).

Card temperatures will definitely rise, so if the card(s) were tuned to produce the maximum, already with single WUs, expect to have to reduce GPU clocks by circa 30, this is to keep the VRMs under control as they are now doing - crudely - twice the work. As always you are limited by the 5970 VRM design fault, so watch the second (or fourth if a twin 5970) like a hawk with GPU-Z. You'll find the max temps will occur around 60-80% done, so keep an eye on the percent done and run GPU-Z around those times - probably best to keep it open until WU finishes until you are sure you have got the new VRM levels nailed.

So ..... good to go .... just watch the VRM temps like a hawk for the first half dozen runs until you are sure you have VRM temps under control. Suggest you set cache for 0.1 until you are sure all is well and you want to continue with it longer term to save trashed WUs.

Regards
Zy
37) Message boards : Number crunching : Runtimes with 1.3 version (Message 2352)
Posted 21 Jan 2012 by Profile Zydor
Post:
Did you clean out the old drivers with Driver Sweeper and run a registry check ?

Regards
Zy
38) Message boards : Number crunching : Runtimes with 1.3 version (Message 2347)
Posted 21 Jan 2012 by Profile Zydor
Post:
You crunched around 50 WUs with V1.3 at your standard crunch time.

It started to go wrong 0040 UTC today. Did you change anything on the installation last night - driver, Win Update, new utilities, any new software?? If you did that may have corrupted something.

Run with no app_info, see what happens

Try a quick check that memory boards and cards have not worked slightly loose - unlikely as that would cause straight errors, but its a quick elimination.

I'd then try a reset first, and reload project files.

If still hassles, do a Project attatch/detatch

If that doesnt do it, I'd then suspect the AMD Driver had become corrupt (clean it totally with Driver sweeper and registry check before reloading fresh driver).

Post again if still running slow after that lot.

Regards
Zy
39) Message boards : Number crunching : 7970s Dont Run (Message 2345)
Posted 21 Jan 2012 by Profile Zydor
Post:
So it's not an OpenCL one..


*Perk*

That would put the cat amongst the pigeons if they got it going on CAL via a standard AMD driver for 7970s, the beast would be unfettered and gallop along :)

It'll be interesting to see what NVidia come up with, they lost their way for 4 years with the fermi 3XX and 4XX debacles - only pulled from the brink with 560/570/580, and seem to be getting back into their old hardware mode they used to have pre 2006.

I have a suspicion they may well come up with the goods this time .... and they need to, else the consumer side will rapidly decline after the Fermi 3XX & 4XX debacles - one bitten, twice shy, thrice you die as they say :)

Regards
Zy
40) Message boards : Number crunching : App_info..xml (Message 2342)
Posted 21 Jan 2012 by Profile Zydor
Post:
Sorry for your loss but it just means the rest of us have more units to crunch!


rofl :)

The main event is assured, its only the manner of your death thats in doubt :)

Regards
Zy


Previous 20 · Next 20


 
Copyright © 2011-2024 Moo! Wrapper Project