Is there an advantage to using more than 2 wu's for Einstein?
Sie sagen "oder", sie mussen ein Deutsche sein?
Can I run seti on my cpu's while I run Einstein on the gpu?
If the image has the appropriate power, more than 2 Wu may well be expected in parallel. The most effective for your gpu (ati 7970?) You must herrausfinden itself. (4-6 Wu's approximately)
: 2.
You are correct
You just have to click on the nick
:3
Auf der Cpu dürfen alle Projekte laufen.
The Google Translator is also easy to recognize. :D
Is there an advantage to using more than 2 wu's for Einstein?
Sie sagen "oder", sie mussen ein Deutsche sein?
If the image has the appropriate power, more than 2 Wu may well be expected in parallel. The most effective for your gpu (ati 7970?) You must herrausfinden itself. (4-6 Wu's approximately)
: 2.
You are correct
You just have to click on the nick
The Google Translator is also easy to recognize. :D
Google Translator -- lol
Vielen Dank Pollux
Ich werde dann vier wu's benutzen.
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
I just ordered a radeon hd 270X to add to my machine. I have the 280X set to 5 wu's and it seems to work well. Will I have to adjust my app_config downward to 3 or 4 wu's so that both graphic cards will work properly? Thanks
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
Probably but not necessarily because of the new card but because you'll be running 2 cards in the machine. Einstein tasks tend to transfer a lot of data over the PCIe bus and running upwards to 10 tasks will probably overtax the available bandwidth. I would recommend that you try different settings and pay attention to GPU load and run times and then settle on the one that works best for you.
You might (probably should) also experiment with how much of the CPU you assign to support each GPU task as previous discussion on these message boards have indicated that OpenCL GPU tasks might require a full CPU core per task as support for optimal performance.
As you may be aware, SETI has run out of work so my machines are now looking for work from Einstein. Based on preferences, they have loaded up correctly but with only 1 wu per GPU. I've reviewed the dialog on this thread and have installed the app_config.xml file for concurrent (2 wu's for each GPU in this case) execution. Re-initiated Boinc and see that it it found the .xml file however, from the dialog on this tread, seems I need to bet rid of the "High Priority" mode. Have quite a bit of experience with use of app_config.xml (on SETI) but do not know how to eliminate use of "High Priority", i.e., setting buffer to 0 (??) for Einstein.
I have 2 multiple GPU machines, one with dual 480's and one with dual 780ti's. So there is power going to waste presently. Appreciate any advice on establishing the proper configuration.
Appreciate any advice on establishing the proper configuration.
Welcome to Einstein, Bill.
Here at Einstein, the nice thing is that there is a configuration control provided on the web site for running multiple jobs per GPU, so you don't have to use the configuration file approach unless you want a more specialized result.
I suggest you discard your app_config.xml, go to your einstein web site account page, click on the Einstein preferences link, and adjust the entry reading "GPU utilization factor of BRP apps"
There are two slightly inconvenient things about using this method.
1. the coding is inverse, so to get 2 jobs, you set the parameter to 0.5, to get three set 0.33, and so on.
2. a revision you enter on the web preferences does NOT take effect on a particular computer of yours until that computer finishes downloading a new WU of the application of interest.
Oh, yes, and as with all use of web site specified preferences you must take care to make changes to the Location (aka venue) for your computer of interest (default|home|work|school).
As to "high priority", the best way to avoid it is to go to your general Computing preferences and greatly shorten your days of requested work queue. SETI erratic work delivery may have gotten you in the habit of setting the sum of
Computer is connected to the Internet about every
Maintain enough work for an additional
to several days (or even many days).
I suggest something like half a day will serve you better here. But I doubt the high priority matter is causing you harm at the moment.
Your informative and well organized reply much appreciated. I had taken a look at all 3 categories of preferences and had assumed that adjustments to the GPU utilization factor as a possible solution. However, 2 things came to mind, one being the "Dangerous!" notation adjacent to each of the 3 types of wu's and the other having to do with the effects of a web based adjustments on GPU's of different power and capacities. As you may know, the nVidia 480 has about half the capability of the 780ti.
I would appreciate your thoughts on these points but a description or reference for the meanings of BRP, FGRP, and GW also appreciated. After this, I'll try to keep my questions down.
And thanks for heads-up about the inverse relationship of values to wu's assigned for GPU utilization.
merle van osdol wrote:Is
)
If the image has the appropriate power, more than 2 Wu may well be expected in parallel. The most effective for your gpu (ati 7970?) You must herrausfinden itself. (4-6 Wu's approximately)
: 2.
You are correct
You just have to click on the nick
:3
Auf der Cpu dürfen alle Projekte laufen.
The Google Translator is also easy to recognize. :D
RE: merle van osdol
)
Google Translator -- lol
Vielen Dank Pollux
Ich werde dann vier wu's benutzen.
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
— Salman Rushdie
With pleasure Note to
)
With pleasure
Note to HyperThreading -> if the simulated cores are not sufficient, please adjust app_config accordingly.
0.25
0.5
Greeting Pollux
I just ordered a radeon hd
)
I just ordered a radeon hd 270X to add to my machine. I have the 280X set to 5 wu's and it seems to work well. Will I have to adjust my app_config downward to 3 or 4 wu's so that both graphic cards will work properly? Thanks
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
— Salman Rushdie
Probably but not necessarily
)
Probably but not necessarily because of the new card but because you'll be running 2 cards in the machine. Einstein tasks tend to transfer a lot of data over the PCIe bus and running upwards to 10 tasks will probably overtax the available bandwidth. I would recommend that you try different settings and pay attention to GPU load and run times and then settle on the one that works best for you.
You might (probably should) also experiment with how much of the CPU you assign to support each GPU task as previous discussion on these message boards have indicated that OpenCL GPU tasks might require a full CPU core per task as support for optimal performance.
Also watch for the GPU memory
)
Also watch for the GPU memory consumption of the tasks. If your new card has 2 GB you should be fine, at 1 GB you may be limited to 3 or 4 WUs.
MrS
Scanning for our furry friends since Jan 2002
Thanks for the info. If I
)
Thanks for the info. If I need to ask more questions after I get the card. I'll ask again.
merle
What is freedom of expression? Without the freedom to offend, it ceases to exist.
— Salman Rushdie
Hello - As you may be
)
Hello -
As you may be aware, SETI has run out of work so my machines are now looking for work from Einstein. Based on preferences, they have loaded up correctly but with only 1 wu per GPU. I've reviewed the dialog on this thread and have installed the app_config.xml file for concurrent (2 wu's for each GPU in this case) execution. Re-initiated Boinc and see that it it found the .xml file however, from the dialog on this tread, seems I need to bet rid of the "High Priority" mode. Have quite a bit of experience with use of app_config.xml (on SETI) but do not know how to eliminate use of "High Priority", i.e., setting buffer to 0 (??) for Einstein.
I have 2 multiple GPU machines, one with dual 480's and one with dual 780ti's. So there is power going to waste presently. Appreciate any advice on establishing the proper configuration.
Bill Greene wrote: Appreciate
)
Welcome to Einstein, Bill.
Here at Einstein, the nice thing is that there is a configuration control provided on the web site for running multiple jobs per GPU, so you don't have to use the configuration file approach unless you want a more specialized result.
I suggest you discard your app_config.xml, go to your einstein web site account page, click on the Einstein preferences link, and adjust the entry reading "GPU utilization factor of BRP apps"
There are two slightly inconvenient things about using this method.
1. the coding is inverse, so to get 2 jobs, you set the parameter to 0.5, to get three set 0.33, and so on.
2. a revision you enter on the web preferences does NOT take effect on a particular computer of yours until that computer finishes downloading a new WU of the application of interest.
Oh, yes, and as with all use of web site specified preferences you must take care to make changes to the Location (aka venue) for your computer of interest (default|home|work|school).
As to "high priority", the best way to avoid it is to go to your general Computing preferences and greatly shorten your days of requested work queue. SETI erratic work delivery may have gotten you in the habit of setting the sum of
to several days (or even many days).
I suggest something like half a day will serve you better here. But I doubt the high priority matter is causing you harm at the moment.
Your informative and well
)
Your informative and well organized reply much appreciated. I had taken a look at all 3 categories of preferences and had assumed that adjustments to the GPU utilization factor as a possible solution. However, 2 things came to mind, one being the "Dangerous!" notation adjacent to each of the 3 types of wu's and the other having to do with the effects of a web based adjustments on GPU's of different power and capacities. As you may know, the nVidia 480 has about half the capability of the 780ti.
I would appreciate your thoughts on these points but a description or reference for the meanings of BRP, FGRP, and GW also appreciated. After this, I'll try to keep my questions down.
And thanks for heads-up about the inverse relationship of values to wu's assigned for GPU utilization.