I have 6 computers with i7 processors (4+4 CPUs) and 2 other computers with dual processors running SETI and other projects. Unfortunately, in Tools > Computing preferences > Computing allowed section, it is an "all or none" for using the CPUs while the computer is in use. Ditto for the GPUs.

I wish for the ability to specify the number of CPUs/GPUs that may be used while the computer is in use, since I only rarely use more than a single CPU and GPU when I am doing work on any of my computers. Thus, the other CPUs/GPUs could continue automatically.

I know I can free a CPU by adjusting the number using the "... use at most __% of the processors ...", but I don't want to have to adjust that down every time I want to check my email and other things, and then back up when I am finished. The alternative is to always leave 1 not in use. I am aware of NO easy alternative for GPU usage adjustment.

I am sure many, many other people are running multi-processor and -GPU computers now, so this might give a bit more CPU/GPU power to science.

Why should you have to adjust the CPUs BOINC uses while you check your email? By default, the science applications run at the lowest thread priority supported by the OS, and things like email or other user-initiated tasks all run at normal priority. There is no competition for CPU resources between these two things; the user resources will all win out.

For GPU apps, use the <exclusive_gpu_app> param in cc_config.xml so that it suspends GPU processing whenever a specific executable is in memory.

Why should you have to adjust the CPUs BOINC uses while you check your email? By default, the science applications run at the lowest thread priority supported by the OS, and things like email or other user-initiated tasks all run at normal priority. There is no competition for CPU resources between these two things; the user resources will all win out.

For GPU apps, use the <exclusive_gpu_app> param in cc_config.xml so that it suspends GPU processing whenever a specific executable is in memory.

It is not the email that uses much CPU or GPU, it is the "other" things that often get performed during the time when I check my email. Some tasks use small amounts of CPU many, many times. Each little sip of CPU requires an admittedly small wait for one of the CPUs to come free. My understanding of Windows is that it is a polling that results in task switches, not an interrupt. Thus the elapsed time is extended and responsiveness suffers.

Even "normal" uses are impacted by the "background" uses of projects. It may just be that the projects are not offering to give up the CPU often enough. Whatever it is, I notice it as lag between typing and action of the program. When I have BOINC leave 1 or more CPUs for my use, I don't get the lags.

Another "thing" is when a video game is being played. One of my machines has 4 GPUs, all of which must stop if I want a single GPU for my use.

Another "thing" is when a video game is being played. One of my machines has 4 GPUs, all of which must stop if I want a single GPU for my use.

For this problem you can try GPUSwitch 2.0
"Script for disabling/enabling GPUs on the fly"
http://setiathome.berkeley.edu/forum_thread.php?id=62127

The original site (http://www.hellhorse.net/) is down (no DNS record) so I Uploaded my copy:
http://www2.zippyshare.com/v/78377495/file.html

The GPU Switch script makes changes to cc_config.xml


Before usage you need to change some strings in GPUSwitch.js

var strBOINCDir = "D:\\Sensitive data programs\\BOINC";

And all strings (read the above link for cc_config.xml to know why):
ignore_cuda_dev

... to:
ignore_nvidia_dev

 

---

 


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

Why should you have to adjust the CPUs BOINC uses while you check your email? By default, the science applications run at the lowest thread priority supported by the OS, and things like email or other user-initiated tasks all run at normal priority. There is no competition for CPU resources between these two things; the user resources will all win out.

For GPU apps, use the <exclusive_gpu_app> param in cc_config.xml so that it suspends GPU processing whenever a specific executable is in memory.

It is not the email that uses much CPU or GPU, it is the "other" things that often get performed during the time when I check my email. Some tasks use small amounts of CPU many, many times. Each little sip of CPU requires an admittedly small wait for one of the CPUs to come free. My understanding of Windows is that it is a polling that results in task switches, not an interrupt. Thus the elapsed time is extended and responsiveness suffers.

Even "normal" uses are impacted by the "background" uses of projects. It may just be that the projects are not offering to give up the CPU often enough. Whatever it is, I notice it as lag between typing and action of the program. When I have BOINC leave 1 or more CPUs for my use, I don't get the lags.

Another "thing" is when a video game is being played. One of my machines has 4 GPUs, all of which must stop if I want a single GPU for my use.

It is up to the OS to do the switching. The programs only control their priority. There is a checkbox someplace in the control panel (System maybe) that allows you to optimize for background tasks or foreground tasks. Optimize for foreground tasks means that the process switcher will respond faster to high priority tasks requesting CPU time.

that said, Most users do not notice that all of their CPUs are in use by BOINC on most machines. The reverse is true for GPUs. Most users will notice a difference if the GPUs are all in use when you are active.

---

BOINC WIKI

... There is a checkbox someplace in the control panel (System maybe) that allows you to optimize for background tasks or foreground tasks. ...

I think you mean the Computer > Properties > Advanced > Performance > Settings > Advanced tab. I can select "Adjust for best performance of: 1) Programs, or 2) Background services". Since I am running programs for my usage, I chose "Programs". If it had been the other way, your suggestion might have been just the ticket to success.

Most users do not notice that all of their CPUs are in use by BOINC on most machines

I agree with you 100%, but then I am probably NOT the average user. I have six systems with i7 CPUs, two with dual cores, and another bare bones I only use for certain standalone disk maintenance activities. I am fairly sure most users don't run that many systems. I also worked in IT for 30 years as an operator, system programmer, manager, configuration manager, and have a BSEE degree. I am quite aware of when my computers are not acting quite right, and try to adjust things so they are optimal.

This forum is for "wishes", and this is one of mine. Do the developers read these wishes and put them into some road map for incorporation? It would be good to know which are planned and which are still just wishes.

To BilBg: Thanks for the references. I will take a look to see if it/they will satisfy my wish in another way.

... There is a checkbox someplace in the control panel (System maybe) that allows you to optimize for background tasks or foreground tasks. ...

I think you mean the Computer > Properties > Advanced > Performance > Settings > Advanced tab. I can select "Adjust for best performance of: 1) Programs, or 2) Background services". Since I am running programs for my usage, I chose "Programs". If it had been the other way, your suggestion might have been just the ticket to success.

Most users do not notice that all of their CPUs are in use by BOINC on most machines

I agree with you 100%, but then I am probably NOT the average user. I have six systems with i7 CPUs, two with dual cores, and another bare bones I only use for certain standalone disk maintenance activities. I am fairly sure most users don't run that many systems. I also worked in IT for 30 years as an operator, system programmer, manager, configuration manager, and have a BSEE degree. I am quite aware of when my computers are not acting quite right, and try to adjust things so they are optimal.

This forum is for "wishes", and this is one of mine. Do the developers read these wishes and put them into some road map for incorporation? It would be good to know which are planned and which are still just wishes.

If there is a possible configuration problem, we try to fix that first before bringing the wish to the attention of the developers. There have been quite a few wishes that were solved by fixing a configuration issue,

What is the state of using GPU when you are active?

---

BOINC WIKI

This forum may be for wishes, but what you wish for is something that needs to be added to the Boinc framework, not to Seti. And no, the Boinc developers certainly don't read these forums.

That said, what you wish for is one thing that gets wished for about 7 times a year on the Boinc forums and which I no longer forward to the developers anymore, as it's in some form on their "future to do list" already: http://boinc.berkeley.edu/trac/wiki/PrefsRemodel.

---

What is the state of using GPU when you are active?

I'm not quite sure what you are asking here. What I see when I notice lags is one task running a heavy CPU workload and displaying a progress bar while a second task, also heavy CPU, is crunching away on what it is doing. The progress bar normally has a steady progression from left to right for each chunk of work, except when my second task is also running. Then the progress bar will move, pause a little, move some more from where it stopped, pause, move, ... If I use Tools - > Computing preferences > processor usage > "... use at most 90% ..." to have BOINC stop using a CPU, the progress bar returns to its normal, steady progressing. Neither of these tasks use the GPU for anything other than normal displays.

Did this help any? Could this be the result of some O/S resource locking which isn't getting enough CPU to release its lock in a timely manner? Perhaps I will try giving CPU preference to the background processes to see if it changes.

Any other questions or suggestions to try?

That said, what you wish for is one thing that gets wished for about 7 times a year on the Boinc forums and which I no longer forward to the developers anymore, as it's in some form on their "future to do list" already: http://boinc.berkeley.edu/trac/wiki/PrefsRemodel.

Excellent! You are acting as the "front end processor" for the devs to filter duplicate wishes. And the link you gave is a great place to look before making any more wishes here. Thanks!

On which computer ID is this:
http://setiathome.berkeley.edu/hosts_user.php?userid=61201

How do you know "Neither of these tasks use the GPU"?
Are they SETI@home v7 or AstroPulse?
What is the process name (.exe)?

What you describe (lag and % progress in 'chunks') seems like OpenCL GPU AstroPulse task

---

 


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

On which computer ID is this:
http://setiathome.berkeley.edu/hosts_user.php?userid=61201

The one I use most frequently for my other work is ID: 5111135

How do you know "Neither of these tasks use the GPU"? Are they SETI@home v7 or AstroPulse? What is the process name (.exe)?

Neither one is a BOINC project - they are "other" programs I use. The program that shows the progress bar is a batch-type conversion program working on (usually) a large number of files, and each file gets its progress of conversion being reported individually. Thus I can see the progress file by file, and during each file conversion. The other program is a preprocessor for the conversion program and works on a large number of files, but has no progress bar. While the preprocessor works one batch of files, the conversion program works on the previous batch. There is no overlap between the batches.

Since computer ID: 5111135 only has an iGPU, I don't run games on it, but it is the one where I see lags from the CPU usage. The lag I see is NOT for the display - it is in the progress of the conversions. If the progress bar "jumped", then it would be a display lag. The progress bar is smooth, but with intermittent pauses in processing.

Neither one is a BOINC project - they are "other" programs I use.

Then why we (you) discuss this here??

"other" programs run in normal (or higher) priority
CPU tasks of any project run by BOINC use low/idle priority so do not introduce lag (even when they show as using 100% CPU)
Even on 500 MHz CPU I do not feel lag (i.e. no difference if BOINC runs or not CPU SETI@home task on the single core of this CPU)

---

 


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

Then why we (you) discuss this here??

I posted my wish for a BOINC capability and you started asking questions.

My experience with all CPUs running BOINC projects (not just SETI), is not the same as yours. I DO notice lags when I don't make BOINC use fewer than all CPUs when I run my heavy tasks.

In the link posted earlier in this thread, I see that others have also wished for the same capability, or at least one that is similar, so I am content to wait for the developers to add it, someday. Meanwhile, I will either throttle BOINC manually or just live with the lags.

So far as I am concerned, this thread is ended.