Force 3.1 to execute with 4.2 app

[Snow Crash]

For those of us running the LONG tasks on fermi and kepler GPUS.
I am in the process of testing a workaround posted by Retvari Zoltan on the GPUGrid forums to force WUs downloaded and assigned to the 3.1 app to execute with the 4.2 app. So far everything is executing nicely, we'll have to wait a bit more to see the final results.
http://www.gpugrid.net/forum_thread.php?id=3043

Aborting CUDA3.1 tasks are easy (also detrimental for the project), but what can you do if you have a CUDA3.1 workunit running for hours, and you don't want to waste it?
I've figured it out, that I can convert CUDA3.1 tasks to CUDA4.2 tasks.
So if you have some file management skills, you can do it too in no time:
(My instructions are for Windows only) V2.0

1. Pause all running CUDA3.1 tasks in BOINC manager
2. Locate the slot directory of the CUDA3.1 task by searching through the slots for the CUDA3.1 executable (acemd.win.2352)
---(on Windows XP the slots are located in c:\Documents and Settings\All Users\Application Data\BOINC\slots\)
---(on Windows 7 the slots are located in: c:\ProgramData\BOINC\slots\)
3. Copy the CUDA4.2 executable files to the slot directory of the CUDA3.1 task located in step 2.
---(the source of the CUDA4.2 client files (acemd.2562.cuda42, cudart32_42_9.dll, cufft32_42_9.dll) is
---(on Windows XP: c:\Documents and Settings\All Users\Application Data\BOINC\projects\www.gpugrid.net\)
---(on Windows 7: c:\ProgramData\BOINC\projects\www.gpugrid.net\
4. Delete the CUDA3.1 client (acemd.win.2352) from the directory located in step 2.
5. Rename the CUDA4.2 client executable file to the name of the CUDA3.1 client in the slot directory
---(acemd.2562.cuda42 -> acemd.win.2352)
---(do not rename the dll's)
6. Rename the CUDA3.1 client executable file in the project directory. (acemd.win.2352 -> acemd.win.2352.bak)
---(on Windows XP: c:\Documents and Settings\All Users\Application Data\BOINC\projects\www.gpugrid.net\)
---(on Windows 7: c:\ProgramData\BOINC\projects\www.gpugrid.net\
7. Make a copy of the CUDA4.2 client in the project directory, and rename it to the CUDA3.1 client's name. (copy of acemd.2562.cuda42 -> acemd.win.2352)
8. Restart the CUDA3.1 tasks in BOINC manager (they will be displayed still as a CUDA3.1 task)

I hope it will work for you too.
Windows Explorer should be set to show hidden and system folders if you want to browse for these directories, or you can paste the paths directly from my text.
Also the "hide extensions of known files" option should be disabled.
I recommend a file manager tool like Total Commander (step 4 and 5 can be done in a single step with this utility)
I ask a Linux expert to try a similar method, and share it here if it's working on Linux too.

[Snow Crash]

Forgot to mention this is Windows only.

OK ... Looks like you have to re-do this every time you get a 3.1task. Please see GPUGrid thread linked above for a details of a .bat file process to do this for you.

If I get ambitious I may write my own in vbScript just so it is easier to follow.

[Snow Crash]

Another update ... (is included in the post at GPUGrid)

If you update your config.xml as seen below this only needs to be one 1 time (no script or babysitting necessary)!
If you can copy a file and change it's name you have NO excuse for not doing running 3.1 tasks with the 4.2 app

<cc_config>
<options>
<report_results_immediately>1</report_results_immediately>
<dont_check_file_sizes>1</dont_check_file_sizes>
</options>
</cc_config>

[Snow Crash]

Another update ... (is included in the post at GPUGrid)

If you update your config.xml as seen below this only needs to be one 1 time (no script or babysitting necessary)!
If you can copy a file and change it's name you have NO excuse for not running 3.1 tasks with the 4.2 app

<cc_config>
<options>
<report_results_immediately>1</report_results_immediately>
<dont_check_file_sizes>1</dont_check_file_sizes>
</options>
</cc_config>

[SMTB1963]

Thanks for this post SC. I've made the changes and things are moving quite a bit faster.

[Snow Crash]

I noticed your points shooting up with the rest of us the last day or so ...

Whether the improvements were due to the CUDA 4.2 updates alone or if the GPUGrid team took advantage of something new in it I think a 2x performance increase is exciting (well as exciting as watching grass grow can be
We might even start to see see some movement from people running other GPU projects switch over. Anyone one have PPD stats from other projects we could compare?

[shadowwind]

hey snow what are the the temps running at with this new program im getting ready to build another computer so i think i will give nvid another chance.

[SMTB1963]

I noticed your points shooting up with the rest of us the last day or so ...

While I was away, 3 of my 4 GTX275s and my AMD 1055t system went down. Temps in the garage (~95F) were probably too much for them. But yesterday I added another GTX570...that combined with the new 4.2 app should give me >700K ppd. Thanks again for pointing to Retvari's thread.

My new GTX570 hasn't gotten above 63C (others run at 76C with 35C ambient) while running at 99% utilization. IMO the 570s offer the best bang for the buck right now...I might just have to pick up a couple more since they'll also be good for WCG GPU tasks.

[Snow Crash]

@shadow - One of my cards is a GB GTX 670 from Newegg for $399 (still in stock I just checked)
This 'er runs quiet, cool (60c at 90F ambient), is a lightweight when it comes to electricity but is tearing through WUs like mad.
Heartily recommended

@SMTB -

<snip>that combined with the new 4.2 app should give me >700K ppd.

and I thought I had an outside chance of catching you but my 7 day avg is only 685k
now I just need to figure out how to sneek 1 more card past the wife

[shadowwind]

ty snow i added 2/ 670s to my cart im up to 1500. including shipping. now to try and sneak this pass my wife.

if needed i got a pair of 7870s folding in ubuntu helping out my old team for now, but i could all ways divert them to donate at home.
im waiting on wcg to go gpu for them hmmm seems like it taking forever for the gpu get going for wcg.

[shadowwind]

forgot to ask what is the average points per day on your 670 mecole,

[Snow Crash]

With running the GPU at 1259 on Win7x64 with 1 free CPU core the RAC on my 670 is currently at 223k and is still building strongly ... I'm going to guestimate it will peak around 375k.

[shadowwind]

good to know and ty for the info.

are you running WCG also.on donate at home i take a big performance hit on my gpu cards when i run wcg.
i wonder if its the same on the gpugrid ? on donate at home i lose about 200k if i run WCG.

[FaYt]

Yup Snowcrash a 670 is going like an .... got one here to and love it

[Snow Crash]

good to know and ty for the info.

are you running WCG also.on donate at home i take a big performance hit on my gpu cards when i run wcg.
i wonder if its the same on the gpugrid ? on donate at home i lose about 200k if i run WCG.

On my 920 rig I only have the 1 GTX 670 (sigh) and I manually set BOINC to use 99% of processors so I am only running 7 WCG tasks. On my 980X I have a GTX 480 and a GTX 295 so I'm running 3 GPUGrid tasks at a time and the cumulative CPU estimate per GPU task, as read by BOINC, forces a free core so I set the processor usage to 100% and run 11 WCG tasks.

Currently I'm doing a little cherry picking on the Rioja project with my 980X so I'm not actually running WCG (gasp) on that machine right now but the basic set up is exactly the same.

[SMTB1963]

@SMTB -
and I thought I had an outside chance of catching you but my 7 day avg is only 685k

What handle are you crunching under these days? I can't see you on BOINCStats.

now I just need to figure out how to sneek 1 more card past the wife

LOL, Good luck with that! (grats on the new card BTW)

[Snow Crash]

I always forget that I've changed my crunching name a few times now ... currently crunching under mecole ...
Maybe I should put it up for a poll

Undecided
Ambiguous
Dumbascii
Snow Crash (silly original choice but I thought it was pretty good at the time, why not just pick Neo you dork !!!)
SPM1964

I bet Otis and PG could have a field day with this topic

[werdwerdus]

so with the config file edit you only have to replace the files ONCE and then it will work for all new 3.1 tasks?

[SMTB1963]

I always forget that I've changed my crunching name a few times now ... currently crunching under mecole ...

HA! I was wondering who that new 'smitch "mecole" was... From the looks of things, you might just catch me anyways - although not as fast

Snow Crash (silly original choice but I thought it was pretty good at the time, why not just pick Neo you dork !!!)

I vote for you sticking with "Snow Crash". (I assume it's a reference to the drug/virus from the very influential book of the same title)

so with the config file edit you only have to replace the files ONCE and then it will work for all new 3.1 tasks?

Yep. Just remember to keep an eye out for new apps from the GPUGRID devs...my understanding is that <dont_check_file_sizes>1</dont_check_file_sizes> added to cc_config might keep new acemd apps from downloading.

[werdwerdus]

did it, lets hope it works (won't be around to check on it for 11 hours)

[Otis11]

I always forget that I've changed my crunching name a few times now ... currently crunching under mecole ...
Maybe I should put it up for a poll

Undecided
Ambiguous
Dumbascii
Snow Crash (silly original choice but I thought it was pretty good at the time, why not just pick Neo you dork !!!)
SPM1964

I bet Otis and PG could have a field day with this topic

But I shall refrain...

(Getting my card up btw - was supposed to be up Friday, just emailed the guy to see what's up.)

[Snow Crash]

The workaround should no longer be necessary

http://www.gpugrid.net/forum_thread.php?id=3057

I have made some changes to the server to add some debugging code and some other smaller changes.

Let me know if you have been given a cuda3.1 workunit and you should not have receive that.

gdf

The only reported issue is that if your machine has been marked as "unreliable" on the 4.2 tasks it will be sent the 3.1 instead. If you keep throwing errors you will eventually stop getting work altogether for the day which is no different than in the past.