Moo!

Gilthanis

[H]ard|DCer of the Year - 2014
Joined
Jan 29, 2006
Messages
8,718
Anyone on our team that wishes to not only run this project but also help our team in the DC-Vault should read this thread post and let me know if you have any questions. http://hardforum.com/showpost.php?p=1040794811&postcount=1947

For those having trouble running the 7xxx series AMD cards, please try using the apps located here: http://www.bc-team.org/downloads.php?view=detail&df_id=80

The thread I got the link from was in the Moo! forums here: http://moowrap.net/forum_thread.php?id=200&sort_style=6&start=75

If you need help with where to put the files, feel free to let me know. I will try to update this post later when I have more time. I am not real familiar with Linux, so one of the others may need to step in with questions in regards to it. There are a few different possibilities under Linux. ;)


Edit: I created an account called HardOCPtest so that anyone whom wanted to test or contribute anonymously could do so.

•to attach a computer to your account without using the BOINC Manager. To do so, install BOINC, create a file named account_moowrap.net.xml in the BOINC data directory, and set its contents to:
<account>
<master_url>http://moowrap.net/</master_url>
<authenticator>8032_2ebde61346773521c211273838d294fd</authenticator>
</account>

teamdaily.php
 
Last edited:
Using the apps you linked to, Gilthanis, I was finally able to get Moo! running on my HD7970 and R9 280X. However, for some reason it still will not run on my HD7950, my R9 280, or my HD7870, even when using the same app files that work just fine on my HD7970 and R9 280X. Each WU just errors out within a few seconds of starting on those other three GPU's. Very odd.
 
Yeah...I couldn't tell you on that one. I only tested them on the 7970 while it was working for the week that it did..lol
 
I don't know how I managed it, but I now have Moo! running on all of my GPU's. I should start putting up some nice points for RC5-72.
 
http://moowrap.net/forum_thread.php?id=369#6157

Finally we have a newer application version that supports OpenCL! Updated versions of both AMD/ATI Stream and OpenCL apps for Windows have been offered by our server since yesterday to participants meeting requirements.

These apps require BOINC Client v7.2.28 or newer and Windows XP or newer. The Stream app is not available to AMD Radeon HD 7870/7950/7970/R9 280X series (Tahiti) or newer cards due to compatibility issues. OpenCL app is preferred for those cards.

The old app version is still available for systems running older BOINC Client. However, if your system has multiple GPUs, please consider updating it's BOINC Client to a support version. There are known issues running the old app on systems with multiple GPUs. New apps fix these by running one cruncher per device.

For detailed app change history, please read application changelog. For latest requirements enforced by our server, please read latest requirements.

Please report any problems you might encounter with new apps and our changes. Especially if you no longer get any work from our server where you previously did. You can report them by posting at our forum.

Happy crunching!
 
I don't know how I managed it, but I now have Moo! running on all of my GPU's. I should start putting up some nice points for RC5-72.

Were you able to get the account linked? I've been working with the guy who created Moo and he was able to manually get me a password for RC5-72, but it doesn't seem like it worked. Haven't tried to follow up since then.
 
Were you able to get the account linked? I've been working with the guy who created Moo and he was able to manually get me a password for RC5-72, but it doesn't seem like it worked. Haven't tried to follow up since then.

Yes, I was able to link the accounts using the directions in the post Gilthanis linked to in the very first post in this thread (first link he provides). That process worked for me when I did it a few weeks ago.
 
I hit 100 million points today. Going to back off now and concentrate on getting to 50 million in Milkyway.
 
More apps based on v1.4 of wrapper code are now available.

First there are Android apps with both non-PIE (for pre-4.1) and PIE (for 4.1 and later, especially 5) versions. Android apps are no longer in beta so everybody should get them but please report any problems at the new Android forum http://moowrap.net/forum_forum.php?id=27.

Secondly there's a new nVidia OpenCL app available for Windows. This brings the same level of support for nVidia GPUs that was previously available only for AMD GPUs. The Linux versions are on the list next and will be deployed soon.

Happy crunching!

1.04 steals a whole CPU. I now have a paused CPU project because of it. 1.03 did not do this. Can that please be fixed. I know that OpenCL usually is better about CPU need than CUDA, but the CUDA app doesn't steal a CPU. I also leave cores open for GPU stuff, so I have free cores already.

http://moowrap.net/forum_thread.php?id=410
 
Follow up reply

Right, it was reported to me previously that the OpenCL (especially nVidia one) benefits from having one free CPU for it to feed the GPU and thus is specified as requiring one CPU. However, this does change a lot between Windows and Linux as well different driver versions so it's not that easy to find "correct value".

I can surely change it, that's not a problem. I'll make it same as AMD OpenCL app (it requires 0.25 CPU) for now to see how things go. The change should be live within an hour and subsequent work sent should have the changed CPU requirement.

Thanks for the report!

-w
 
News update
Project was down today between 18:00 and 21:30 EEST (that's from 15:00 UTC/7:00 PST to 18:30 UTC/11:30 PST) for about 3 and half hours while the previously failing disk was swapped with a backup disk and data was copied over. Now the project is running from a disk that's not showing signs of collapsing due to read errors. The backup disk is as old as the failed disk but hasn't had that much use so it should last until the project server is migrated to a new server with SSD disks later this year.

This happened a bit unannounced as I took advantage of the D.net keymaster problems that seems to slow down our work generation for some reason. That's also the why we run out of work before the maintenance and why we still don't have full work buffers. Hopefully that will fix itself once the keymaster is back in action. In any case, our local proxy will eventually run out of work unless the keymaster will be resurrected.

For more information about the keymaster failure, please read http://blogs.distributed.net/2015/07/10/04/28/bovine/. Thanks and happy crunching once the dust settles!

http://moowrap.net/forum_thread.php?id=418#6642
 
I hit my 250 million point goal today, so I am stopping this project for a while. Moving on to my 50 million point goal at Poem now. :cool:
 
I'm experiencing an issue with Moo! whereby the server is not testing any OpenCL tasks on any my gpus which, I'd estimate, is causing them to run at about half their potential speed. It feels a little like I'm wasting electricity whilst they plod through cuda tasks.

I've started a discussion with the project admin/developer who has asked me stick with it for a few days whilst they attempt to resolve.
 
phoenicis, I didn't get any OpenCL other than Cuda tasks through BOINC manager. It seems that if I download the OpenCL client from the distributed.net website, and run it locally without BOINC mgr, I'm able to run much faster on my GTX 960. I'm able to crunch like 1.9B keys/s compare to Cuda of about 1.1B keys/s of similar clock. I don't think the points obtained there will count towards Moo Wrapper but towards DC-Vault. It seems like some folks are able to run OpenCL tasks. Not sure how they do it. Anyone has any idea?
 
pututu, I think, similar to SETI, at some point the server used to go through the cuda/OpenCL selection process correctly. Computers that had OpenCL selected under that process have continued to receive those subsequently.

The server is clearly not working as it should anymore and so, now that they have been made aware, all we can do is see whether the project can come up with a fix over the coming days.
 
Could you force that application by using an app_info.xml file? It has been a super long time since I made one so would have to re-learn it, but that might be one way to get them.
 
Yeah I did that using a suggestion from the Moo's forum here. Event log says something about invalid application task name, even though I exited and launched the BOINC mgr, just to be sure. Need to find valid application task name that works.
 
Do you have all of these files?

Moo!.PNG



My systems keep pulling the CUDA apps too but I don't have time to tinker right now.

Oh...and are you running Windows 10 by chance?
 
Last edited:
I do not have version 1.4 (dnetc520) as I believe it will auto pull these files if you have that specific task/WU downloaded.
 
Windows 7.

I noticed that one of phoenicis TR (linux) just got version 1.04 (cuda60) WUs and it runs faster than his other TR (linux) with v1.03 (cuda31) with similar GPU. Not sure about both GPU settings.


upload_2017-10-15_16-14-16.png




upload_2017-10-15_16-15-55.png
 
The app_info.xml requires you to download those files in advance as the app_info.xml works differently then the app_config.xml files. It was intended to force BOINC to run "alternate platform". Basically, it was so that people could tell the server to send a specific application and run it on that hardware anyway. Typically needed when a project wasn't on the ball with supporting hardware. NativeBOINC does this to run Milkyway work units on Android devices. Milkyway doesn't support them officially.

App_config files were meant to kinda replace the app_info files but they don't fully overlap. App_config are easier to write but what can you do...
 
I noticed that one of phoenicis TR (linux) just got version 1.04 (cuda60) WUs and it runs faster than his other TR (linux) with v1.03 (cuda31) with similar GPU. Not sure about both GPU settings

Both Linux rigs have the same settings. I'd like to say that the cudu60 tasks were the result of something intelligent I did but it was pure blind luck. That rig downloaded mostly cuda60 tasks when it first attached and has continued to do so ever since.
 
The project has been silent for the last few days so I've followed the app_info approach, as you chaps suggested, and am now running opencl tasks in both Windows and Linux. pututu. the link you gave does work with the right windows files in the project folder. You can download them from:

http://moowrap.net/download/dnetc_wrapper_1.4_windows_intelx86__opencl_nvidia_101.exe
http://moowrap.net/download/dnetc520-win32-x86-opencl.exe
http://moowrap.net/download/dnetc-gpu-1.4.ini
http://moowrap.net/download/job-win32-opencl-520.1.xml

The tweaked windows app_info I used is:

<app_info>
<app>
<name>dnetc</name>
<user_friendly_name>Distributed.net Client</user_friendly_name>
</app>
<file_info>
<name>dnetc_wrapper_1.4_windows_intelx86__opencl_nvidia_101.exe</name>
<executable/>
</file_info>
<file_info>
<name>dnetc520-win32-x86-opencl.exe</name>
<executable/>
</file_info>
<file_info>
<name>dnetc-gpu-1.4.ini</name>
</file_info>
<file_info>
<name>job-win32-opencl-520.1.xml</name>
</file_info>
<app_version>
<app_name>dnetc</app_name>
<version_num>104</version_num>
<api_version>7.5.0</api_version>
<file_ref>
<file_name>dnetc_wrapper_1.4_windows_intelx86__opencl_nvidia_101.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>dnetc520-win32-x86-opencl.exe</file_name>
<copy_file/>
</file_ref>
<file_ref>
<file_name>dnetc-gpu-1.4.ini</file_name>
<open_name>dnetc.ini</open_name>
<copy_file/>
</file_ref>
<file_ref>
<file_name>job-win32-opencl-520.1.xml</file_name>
<open_name>job.xml</open_name>
<copy_file/>
</file_ref>
<is_wrapper/>
<plan_class>opencl_nvidia_101</plan_class>
<coproc>
<type>NVIDIA</type>
<count>1</count>
</coproc>
</app_version>
</app_info>

Don't be surprised if you receive smaller tasks to start with the usual size and credits appearing after half an hour or so. So far at least, the results are looking pretty darn good :)
 
Thanks phoenicis. I'll try this when I back home tonite. If it works, I might add 970 onto the same rig. BTW, are you running at 70% power limit? I'm running at 55%. See how much power draw from my rig with two cards.
 
Looks like we might get that top spot a few days earlier than planned.

pututu, I'm still running at 100% I'm afraid. I was going to move a couple of gpus to 55% but then, following a temperature drop in the UK, I decided to re-position the computers so as to avoid grief from the wife about turning the central heating on. A TR/1080Ti rig currently crunching Numberfields and Moo! is pulling 455w from the wall.
 
phoenicis, I have one 1080 and one 970 running now in the same rig after successfully installing the dnetc520 version. Thought I might share the PPD/W data that I just got. No CPU project running on this rig during the study. My indoor temp varies around 68F-70F or 20-21°C. No voltage tweaking on 1080 & both cards are with default settings other than the power limit setting change.

upload_2017-10-21_15-54-21.png



Power limit = 55% for 1080 & 53% for 970. PPD/system_watt = 4193. (If I put the third card in, no room for air flow between cards. I've asus X79 mobo and 650W PSU)
upload_2017-10-21_15-43-42.png


Power limit = 100% for 1080 & 970. PPD/system_watt = 3305
upload_2017-10-21_15-44-4.png
 
That's excellent pututu. I didn't know how to limit the power on the Linux boxes until you posted the command line screenshots of gpu power which led me to the nvidia-smi interface. I’m still learning new things every day.

As I’ve already lost out on the heating discussion following a comment this morning of "It’s bloody freezing, turn the effing heating on" I may as well save a bit on the electricity bill. Oh well, I got away with it for a couple of weeks.

Feedback on my experience on a TR/1080Ti Linux box using the same parameters you used (eg no cpu project running):

At 100% power
Watts 100%.png Wall 100%.png

At 55% power
Watts 55%.png Wall 55%.png

Summary
Moo! Power Comparison.png


The ppd was estimated by taking the last 20 tasks completed from each respective power level.

I have a suspicion that, with high end gpus, the opencl app can be bottlenecked by the cpu. Regardless of gpu power settings the tasks seem to take up an entire cpu thread whereas the Linux cuda app used to only use about 25% of a thread.

Thanks to you and Gilthanis for the all the opencl and power optimisation pointers.
 
Thanks for sharing your finding. If you run two 1080Ti in a single rig, the system efficiency will increase more than 20%. Those 1080Ti are super-efficient, alas don't have the fund to get one yet....
 
Does anyone know who is Sand_Dune?. He/she has being contributing to our team in Moo and NFS.

Sand_Dune, if you read this forum (I think you do), pls drop us a line here.
 
Sand_Dune is a bit too young to come to the forums and so I've been feeding him tips/suggestions.

He is one of my sprogs who I thought was old enough to have more freedom with his PC and, to his credit, he's decided to carry on crunching with it ..... he's fascinated by the project goals, stats and all things technical.

Saying that, I'm not sure how long it will last as never has so much gaming been done.
 
At the rate that we are producing, we should take the top spot in two days.


View attachment 41236

Y'all aren't doing bad - but you're going to have to EARN the top spot in Moo, and between me and that grccoin-3 pool you got a lot of work to do.

9-)


Could be worse - I could be pool "mining" GRC instead of solo "mining" it.

phoenicis is kinda irritating - pushed one of my rigs out of the top 10 into the top 20, and one more out of the top 20 entirely.
Seems to be getting kinda serious about this stuff.

8-O

For reference - the main issue for NVIdia is that the dnet520 client doesn't seem to work AT ALL when you try to run OpenCL on a NVidia card and reverts to running CUDA mode instead - needs the 521 client to get it working in OpenCL with NVidia in my testing.
The performance is a major change - ballpark 6-7 GKeys/sec as I recall to a hair over 10 at the SAME settings - on the Aorus GTC 1080ti card I was doing my testing of that on.
It MIGHT have been a driver-version-specific issue, I only tested in on one driver version (378,something IIRC) on a machine I was usually folding on at that time.

For comparison - my single Vega 56 machine is pushing right around 9 GKeys/sec and while still running under 65C - dunno if I could push it to get 10 reliably or not - which indcates that a Vega 64 air-cooled, with 1/7 more cores and normally boosts HIGHER than the 56, could easily hit 10-11 and might get to 12-13 range.
Even 9 is nice for a card that RIGHT NOW is in the same price range as teh GTX 1070 (that's a 4ish GKey performer) and last week was in the same price range as the GTX 1070ti (which pushes ballpark 6).
 
phoenicis is kinda irritating - pushed one of my rigs out of the top 10 into the top 20
He is our man. We are running this project to secure the top spot in our FB league. I believe both of our rigs are optimized for PPD/W efficiency (like MH/s/W for mining) and we could push higher if we wanted to:whistle:

Anyway, you will see your rigs in the top 20 soon again since we will be moving on to other projects...

BTW, we are running openCL on our Nvidia cards. See post #28 from phoenics. I did run dnetc521 in my earlier post#17 and my 960 did gain about 64% in PPD but the points are not counted towards BOINC.
 
Should I move my farm over to Moo! for a little while. ;)

Besides, I think phoenicis has moved his GPUs over to Einstein at this time.
 
Back
Top