Hi all ;)

Thank you very much for helping me, but you are just to quick to me :o

Maybe I should put it this way:
Please show a sample of a app_info.xml which will help me crunching applications on seti@home and asteroids@home.

Which applications, you will ask me of course...

Those apps which will be the smartest to run on a 64bit system ;) As I said, I'm new to this - selecting which applications to run is a way to make my systems GPU's crunch jobs for seti@home as well. It's already running doing this for asteroids@home. No need for any configuration files there.

Will you help me accomplish this? And if files are needed, where to download these.
My system is a Linux system, Ubuntu Server 12.10 64bit with multiple Nvidia GPU's. CUDA 5.5 installed. Headless server and no GUI at all.

Linux version 3.5.0-17-generic
Boinc version 7.0.27

Why this old version when the newer 7.2.42 has been released? Because, when trying to install it from PPA/repository, I can't! When installing from ubuntu default repository, I get v.7.0.27 ;(
http://setiathome.berkeley.edu/forum_thread.php?id=74347#1490818

How many units at a time?!?! Well, normally, the system decides this. The results are showing that they are not done as the setting is setup. 50% for seti and 50% for asteroid! But, I would like the system which will have 4 cores/CPU and 4 GPUS's to run it the same way. 2GPU's and 2CPU(cores) for seti@ and 2GPU's and 2CPU(cores) for asteroid@. Basically I just need SETI to use the GPU's as well. But I would like to configure my system to run 64bit applications only, if something is to gain this way.

Sorry for the newbie questions guys! I am trying :o


.

---

Project Headless CLI Linux Multiple GPU Boinc Servers
Ubuntu Server 14.04.1 64bit
Kernel 3.13.0-32-generic
CPU's i5-4690K
GPU's GT640/GTX750TI
Nvidia v.340.29
BOINC v.7.2.42

Hi all ;)

Thank you very much for helping me, but you are just to quick to me :o

How many units at a time?!?! Well, normally, the system decides this. The results are showing that they are not done as the setting is setup. 50% for seti and 50% for asteroid! But, I would like the system which will have 4 cores/CPU and 4 GPUS's to run it the same way. 2GPU's and 2CPU(cores) for seti@ and 2GPU's and 2CPU(cores) for asteroid@. Basically I just need SETI to use the GPU's as well. But I would like to configure my system to run 64bit applications only, if something is to gain this way.

Sorry for the newbie questions guys! I am trying :o .

This part is NOT possible right now, or perhaps ever. It is a Boinc software programming thing and the Developers are not interested in going down this path.

As for the app_info.xml and cc_config.xml files they would need to be created on your own using a notepad type program, NOT a word processing program as it often adds things Boinc doesn't understand and it then cancels the whole file.

As for the Boinc files you can always get them from here:
http://boinc.berkeley.edu/dl/?C=M;O=D

The problem with Linux is ther repositories are not always updated to the latest stuff but instead sometimes stop at what works very well for almost everyone.

Please show a sample of a app_info.xml which will help me crunching applications on seti@home and asteroids@home

Not possible (to use one app_info.xml that contains 2 projects)

app_info.xml is per project and you put it in the relevant project directory.
Every app_info.xml affects only one project. You may have one project with app_info.xml and another without app_info.xml (the 'normal' way)

---

- ALF - "Find out what you don't do well ..... then don't do it!" :)

Please show a sample of a app_info.xml which will help me crunching applications on seti@home and asteroids@home

Not possible (to use one app_info.xml that contains 2 projects)

app_info.xml is per project and you put it in the relevant project directory.
Every app_info.xml affects only one project. You may have one project with app_info.xml and another without app_info.xml (the 'normal' way)

I think I have done it where I have 2 running at the same time but each in their own project folders, is that what you are saying.

Hi all ;)
But, I would like the system which will have 4 cores/CPU and 4 GPUS's to run it the same way. 2GPU's and 2CPU(cores) for seti@ and 2GPU's and 2CPU(cores) for asteroid@. Basically I just need SETI to use the GPU's as well. .

The other problem with your idea is you need to TRY running 2 gpu units at a time from your project before you actually settle in for the long term doing it. Some projects scale their gpu work based on your gpu, some don't. Some use ALOT of the onboard gpu resources for each unit, some don't. For those that use alot of gpu resources on one unit running multiple units will actually slow down the crunching as each unit fights for those limited resources. Your gpu's only have 1gb of ram on each, that may limit the total number of units you can run at once.

I think I was clear but let make an example:
You may have 5 projects added/attached to one computer and you want to run 3 of them by using app_info.xml
In this case you need to create 3 different app_info.xml files and put every (one) of them in the relevant project directory.
And the 2 remaining projects will run without app_info.xml (the 'normal' way)

When you write app_info.xml for one of the projects you don't need to think for other projects
(with the only exception of 'count' parameter if you want to run more then 1 task per GPU - you may want to think how to set this, e.g. 0.51 for one app and 0.48 for other app (from the same or different project))

The original desire was to "show a sample of a app_info.xml which will help me crunching applications on seti@home and asteroids@home"
- he thinks that one and only app_info.xml ("a app_info.xml") can or must be used for 2 projects which is not the way to do it.

---

- ALF - "Find out what you don't do well ..... then don't do it!" :)

I think I was clear but let make an example:
You may have 5 projects added/attached to one computer and you want to run 3 of them by using app_info.xml
In this case you need to create 3 different app_info.xml files and put every (one) of them in the relevant project directory.
And the 2 remaining projects will run without app_info.xml (the 'normal' way)

When you write app_info.xml for one of the projects you don't need to think for other projects
(with the only exception of 'count' parameter if you want to run more then 1 task per GPU - you may want to think how to set this, e.g. 0.51 for one app and 0.48 for other app (from the same or different project))

The original desire was to "show a sample of a app_info.xml which will help me crunching applications on seti@home and asteroids@home"
- he thinks that one and only app_info.xml ("a app_info.xml") can or must be used for 2 projects which is not the way to do it.

That's what I thought you were saying and it's easy now, thanks!

Hi Mikey,

The other problem with your idea is you need to TRY running 2 gpu units at a time from your project before you actually settle in for the long term doing it. Some projects scale their gpu work based on your gpu, some don't. Some use ALOT of the onboard gpu resources for each unit, some don't. For those that use alot of gpu resources on one unit running multiple units will actually slow down the crunching as each unit fights for those limited resources. Your gpu's only have 1gb of ram on each, that may limit the total number of units you can run at once.

Thanks you for that ;) I changed my next test system. Lets see how it goes.. Thanks again..

---

Project Headless CLI Linux Multiple GPU Boinc Servers
Ubuntu Server 14.04.1 64bit
Kernel 3.13.0-32-generic
CPU's i5-4690K
GPU's GT640/GTX750TI
Nvidia v.340.29
BOINC v.7.2.42

Hi Mikey,

The other problem with your idea is you need to TRY running 2 gpu units at a time from your project before you actually settle in for the long term doing it. Some projects scale their gpu work based on your gpu, some don't. Some use ALOT of the onboard gpu resources for each unit, some don't. For those that use alot of gpu resources on one unit running multiple units will actually slow down the crunching as each unit fights for those limited resources. Your gpu's only have 1gb of ram on each, that may limit the total number of units you can run at once.

Thanks you for that ;) I changed my next test system. Lets see how it goes.. Thanks again..

No problem, do you have a copy of gpu-z yet? If not grab one here:
http://www.guru3d.com/files_tags/gpu_z_download.html

It will give you info on your gpu and on the second tab about half way down will tell you the gpu load, that is the key thing to look at when trying to see if you can run multiple workunits at once. You are looking to keep that number down below 90% or so. If you use too much of the gpu it must swap to the system which will slow things down dramatically. But if one unit is using 80% of the gpu then trying to run 2 units won't work either. The other thing to look at is the memory in use part, if you have a 1gb gpu and are using 700mb for a single unit then trying to run multiple units won't be efficient as you are back to swapping out to the system memory again.

For gpu cards think of a gpu as having alot of teeny tiny cpu's on it, the more of those the card has the faster it is. So if you look at gpu-z again it will tell you how many 'shaders' your card has, those are the teeny tiny cpu's I am talking about. An AMD 5770 has 800 shaders, while a 5870 has 1600 and a 7970 has 2048 shaders. The more shaders the faster you can crunch!

Your Nvidia 640 has 340 cuda cores while your 610 has 48. You CANNOT directly compare AMD and Nvidia in that way. But you can compare the number of cuda cores/shaders within each brand. My 560Ti for instance has 384 cuda cores while my 760 has 1152. That is why your 640 is faster then your 610, and why my 760 is faster then my 560.

Now when you go to buy a new gpu that is NOT the only thing to consider though, the speed of those shaders/cuda cores is also important as is the total onboard memory. If you can't fit the whole workunit into the onboard gpu memory you are then swapping out to the system memory and your crunching speed will suffer horribly.

He uses Linux (and mostly on CLI)

For Linux GUI there are some 'Alternatives to GPU-Z for Linux':
https://www.google.bg/#q=GPU-Z+Linux
http://alternativeto.net/software/gpu-z/?platform=linux

Clicking on programs names (on the alternativeto.net pages) you get even more Alternatives

---

- ALF - "Find out what you don't do well ..... then don't do it!" :)