Application configuration

Nobody316
Nobody316
Joined: 14 Jan 13
Posts: 141
Credit: 2008126
RAC: 0
Topic 196749

Quote:

Application configuration

This mechanism allows you to specify scheduling parameters associated with specific applications. It is available with 7.0.40+ client versions.
To do this, create a file app_config.xml in the project's directory, with the following format:


uppercase
1

.5
.4

Include one element for each application you want to configure.
name
short name of the application.
max_concurrent
the maximum number of tasks of this application to run at a given time.
gpu_usage
the number of GPU instances (possibly fractional) used by GPU versions of this app
cpu_usage
the number of CPU instances (possible fractional) used by GPU versions of this app

Ok help... :( what is the short name of the application? I have tried everything I can think of "task p2030.20120214.G194.30-03.81.N.b6s0g0.00000_3568_0 using einsteinbinary_BRP4 version 132 (BRP4cuda32nv301)"

Ok so what I am trying to do without much success changing the 0.2 of cpu usage to 0.5 but I really don't know what I am doing... lol :( Can anyone help I would be most grateful.

also there is not a stable version out 7.0.40+ only Development version do I need to download that?

Edit: This is windows 7 side don't run linux side too much at the moment.

PC setup MSI-970A-G46 AMD FX-8350 8 core OC'd 4.45GHz 16GB ram PC3-10700 Geforce GTX 650Ti Windows 7 x64 Einstein@Home

Bikeman (Heinz-Bernd Eggenstein)
Bikeman (Heinz-...
Moderator
Joined: 28 Aug 06
Posts: 3522
Credit: 686043163
RAC: 586162

Application configuration

Hi!

In general, for this kind of configuration you can look into the client_state.xml file in the BOINC base directory (see startup messages in Boinc manager) to get the required "template" info. search for BRP4 or app_version etc.

But before investing more time into this, maybe it's worthwhile discussing what you are expecting to gain by the modification?
And yes, to use this new feature (if it turns out you really want to use it) you'll have to download a developer version, or wait until it turns into a stable version and upgrade to that one.

Cheers
HB

Nobody316
Nobody316
Joined: 14 Jan 13
Posts: 141
Credit: 2008126
RAC: 0

RE: Hi! In general, for

Quote:

Hi!

In general, for this kind of configuration you can look into the client_state.xml file in the BOINC base directory (see startup messages in Boinc manager) to get the required "template" info. search for BRP4 or app_version etc.

But before investing more time into this, maybe it's worthwhile discussing what you are expecting to gain by the modification?
And yes, to use this new feature (if it turns out you really want to use it) you'll have to download a developer version, or wait until it turns into a stable version and upgrade to that one.

Cheers
HB

Thanks Bikeman I will be looking in to it. As far as why I want the GPU to have a bit more access to CPU power if it wants or needs it simple. I changed from 8 cores running down to 6 so I can do some work on here with out slowing down boinc "per say" I have 2 free cores to light game or search web lol nothing heavy but did notice a drop of speed while working/playing. lol

PC setup MSI-970A-G46 AMD FX-8350 8 core OC'd 4.45GHz 16GB ram PC3-10700 Geforce GTX 650Ti Windows 7 x64 Einstein@Home

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109387650130
RAC: 35916591

RE: ... I want the GPU to

Quote:
... I want the GPU to have a bit more access to CPU power if it wants or needs it simple.


This is the bit you're misunderstanding. The 'specification' that comes with the task (eg 0.2 CPUs + 1.0 NVIDIA GPUs) is a (fairly generous) estimate and not a limit. You wont really get better performance by changing the 0.2 to 0.5 for example. The GPU app will access the CPU as much as it needs despite whatever larger value you set.

Quote:
I changed from 8 cores running down to 6 so I can do some work on here with out slowing down boinc "per say" I have 2 free cores to light game or search web lol nothing heavy but did notice a drop of speed while working/playing. lol


This is quite a wise move if you are running multiple tasks on the GPU. If you have free cores anyway, you don't need to change anything else.

With regard to short names of apps, just browse your state file (client_state.xml) and immediately after the .... block you should find some .... blocks, one for each app you are running. Within that block will be .... which is what you are looking for, I believe. I'm doing this from memory at the moment but I think that's correct.

Cheers,
Gary.

Nobody316
Nobody316
Joined: 14 Jan 13
Posts: 141
Credit: 2008126
RAC: 0

RE: This is the bit you're

Quote:
This is the bit you're misunderstanding. The 'specification' that comes with the task (eg 0.2 CPUs + 1.0 NVIDIA GPUs) is a (fairly generous) estimate and not a limit.

Ah ok if that's not a limit then I don't need to change it... gggggggrrrrrrreeeeeaaaaatttttttt!! :) Oh and it's 0.2 cpu's + .5 GPU what a surprise I found when changing the setting the time didn't change much went up a bit but with the geforce 9400 gt it's doing both units but the last few units seems to be faster now. Hopefully soon I can get another 650 TI or maybe 2 newer cards... Thanks Gary I will not bother the settings since I don't need to :)

PC setup MSI-970A-G46 AMD FX-8350 8 core OC'd 4.45GHz 16GB ram PC3-10700 Geforce GTX 650Ti Windows 7 x64 Einstein@Home

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5842
Credit: 109387650130
RAC: 35916591

RE: ... Oh and it's 0.2

Quote:
... Oh and it's 0.2 cpu's + .5 GPU ...


Only if you've changed the GPU utilisation factor preference setting. I was just quoting the default values.

Quote:
... what a surprise I found when changing the setting the time didn't change much went up a bit ...


Are you saying that the time to do two tasks in parallel was the same as the time to do one or are you really saying (as I think you must be) that the time to do two was double the time to do one so you got no benefit from running 2x. I'm not really surprised if that's the case.

Quote:
... but with the geforce 9400 gt it's doing both units but the last few units seems to be faster now. Hopefully soon I can get another 650 TI or maybe 2 newer cards... Thanks Gary I will not bother the settings since I don't need to :)


I like the GTX650 for lower power use, lower price and performance not that much less than the 650Ti. It's very good as an upgrade card for CPU only crunching systems. The main problem with upgrading older CPU crunchers is that the PSU probably needs an upgrade as well. The 650 helps in that regard.

Within the last couple of days, I put one in this host which has crunched CPU only for the last 3 years or so. The CPU is a celeron dual core (E3200) and the PCIe slot is NOT PCIe V2 so I tried it purely as an experiment. The motherboard is a socket 775 with a G41 chipset and it handles DDR3 RAM so I had a hunch it might not be as bad as I was expecting. I run all tasks on all GPUs 2x and the best GPU performance I've seen on hosts with Sandy Bridge CPUs and mobos with PCIe V2 x16 slots is 2 tasks in about 58-59 minutes. As you can see if you browse the tasks list, the linked host is doing the GPU tasks in just over the hour - around 63 minutes! It's doing a lot better than I'd even hoped! I guess I wont be retiring that host anytime soon :-). I also should point out that a similar generation (but G31 chipset) mobo with DDR2 only, performs much worse than the above. Maybe the key is DDR3 RAM.

It doesn't seem to make a noticeable difference if I free up a CPU core so both cores are crunching FGRP2 tasks as well. If you go back through its task list you will see when it was crunching CPU only. At that point, FGRP2 task times were about 5700/5600 (Run/CPU) and after the GPU was fired up they have become 6750/6000 approximately. So the CPU time has only increased a little and the run time has further blown out by pretty much exactly the CPU time required to service the GPU tasks.

As a CPU cruncher this machine had a theoretical RAC of just over 2K. Now it is around 24K (CPU component <2K). Not a bad performance gain for a $100 GPU!!

Cheers,
Gary.

Jord
Joined: 26 Jan 05
Posts: 2952
Credit: 5779100
RAC: 0

Don't get too used to using

Don't get too used to using the app_config.xml file. Apparently it's a temporary thing, until all projects use a preference setting such as the Einstein project does, for setting amounts of tasks per GPU.

Nobody316
Nobody316
Joined: 14 Jan 13
Posts: 141
Credit: 2008126
RAC: 0

RE: Are you saying that the

Quote:

Are you saying that the time to do two tasks in parallel was the same as the time to do one or are you really saying (as I think you must be) that the time to do two was double the time to do one so you got no benefit from running 2x. I'm not really surprised if that's the case.

I like the GTX650 for lower power use, lower price and performance not that much less than the 650Ti. It's very good as an upgrade card for CPU only crunching systems. The main problem with upgrading older CPU crunchers is that the PSU probably needs an upgrade as well. The 650 helps in that regard.

Within the last couple of days, I put one in this host which has crunched CPU only for the last 3 years or so. I run all tasks on all GPUs 2x and the best GPU performance I've seen on hosts with Sandy Bridge CPUs and mobos with PCIe V2 x16 slots is 2 tasks in about 58-59 minutes. As you can see if you browse the tasks list, the linked host is doing the GPU tasks in just over the hour - around 63 minutes! It's doing a lot better than I'd even hoped! I guess I wont be retiring that host anytime soon :-). I also should point out that a similar generation (but G31 chipset) mobo with DDR2 only, performs much worse than the above. Maybe the key is DDR3 RAM.

It doesn't seem to make a noticeable difference if I free up a CPU core so both cores are crunching FGRP2 tasks as well.As a CPU cruncher this machine had a theoretical RAC of just over 2K. Now it is around 24K (CPU component <2K). Not a bad performance gain for a $100 GPU!!

Nice little setup gary. I do plan on taking the Geforce 9400 GT off line it's speed doing 2 task is too much "for the price". Meaning some units "Only talking about the 9400 right now" takes 4 hours and some last night was taking up to about 14 hours. Do Note this is under windows I have at this moment 1 unit left running before switching over to linux side to run some benchs and get boinc setup and running. This desktop was built without doing this type of work in mind only for gaming at that time. Since then I learned of boinc and started using it. I only play games once in awhile and at most on the given game day 12 hours at most but mostly 4 to 6 hours. As to what day is game day... when ever I feel like it but that is not really often due to back problems and it takes alot out of me to sit at the computer. I use the laptop mostly for on line stuff as it's easier on the back.

Now as far as the DDR's go. DDR2 will work fairly fine if overclocked but not pushed too far as too far is not stable enough for doing this type of work. DDR3 is by far the best for memory "at this moment in time" and the GDDR5 for the GPU. I am still researching some stuff. There is alot of difference between same GPU's make "ie" MSI 650Ti vs the Gigabyte 650Ti. Drivers is about the biggest difference with some working for some while others have bad luck. I use to program in the early 80's but have not done much lately. Things have changed.

Now to answer about the split. The 9400 listed above now the 650Ti. It was around an half hour not sure with out looking it up with 1 unit. After split to 2 time was like an hour. Not really a gain but not really a lost either. I will be trying different settings in linux. The last unit now is rather the longer running 1's but it's at 80% "BRP4SSE" so soon I get to switch and will see the difference. Not going to be running the 9400 anymore. It's old and I am sure it's tired as I have worked it close to death over the past 3 or 4 years now. LOL I will let the younger gen. of GPU get it's work on. Will let ya know about the benchs and the times.

Oh almost forgot... This desktop was about $800 just before X-mas.. Now it's higher. Got my card for just under $80 at that time but now is over $150 I think or maybe alittle lower/higher. Having a look around and timing is everything in the computer world. Just before X-mas is really a great time to shop. :)

PC setup MSI-970A-G46 AMD FX-8350 8 core OC'd 4.45GHz 16GB ram PC3-10700 Geforce GTX 650Ti Windows 7 x64 Einstein@Home

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.