Recently I added a new video card. The status column under the TASKS tab used to say .1 CPUs + 1.00 NVIDIA GPU (prior to addition of 2nd GPU), then it said .22 CPUs (after addition of 2nd GPU), now it says .23 CPUs. Why is this changing? I guess I don't understand what this information means exactly.

---

Recently I added a new video card. The status column under the TASKS tab used to say .1 CPUs + 1.00 NVIDIA GPU (prior to addition of 2nd GPU), then it said .22 CPUs (after addition of 2nd GPU), now it says .23 CPUs. Why is this changing? I guess I don't understand what this information means exactly.

Mine say 0.04 CPUs + 1.0 nvidia GPUs when I had one and now that I have 2.

---

Official Abuser of Boinc Buttons...
And no good credit hound!

I have 3 GPU's and all of mine say 0.04 CPUs + 1.00 GPUs

---

The cpu has to load tasks for the GPU so I'm assuming that's what the fractional number represents...The amount of CPU that's used for CUDA.

I would have to make another guess about the changing numbers. Different GPU's use different amounts of CPU. Possibly even different versions of CUDA might use different amount of the CPU.

I would be interested in hearing from someone that knows exactly how that works. Mine is currently using .16 CPUs for two Cards.

The (0.xx CPUs + 1.00 NVIDIA GPUs) are only for to calculate how much work you have and for the new work request.
It have nothing to do (with the current available clients) with the CPU support for the GPU.

If I had all stock (no opt. apps) I had with all my OCed GTX260-216 -> 0.19 CPUs.
On my GPU cruncher (4x) also on my QX6700 (1x).

If you have 0.04, IIRC this is because you used the Lunatics Installer for to install the opt. apps and in this app_info.xml is the 0.040000 entry.

I had stock 0.19, so I inserted 0.190000 in my app_info.xml .

---

The (0.xx CPUs + 1.00 NVIDIA GPUs) are only for to calculate how much work you have and for the new work request.
It have nothing to do (with the current available clients) with the CPU support for the GPU.

If I had all stock (no opt. apps) I had with all my OCed GTX260-216 -> 0.19 CPUs.
On my GPU cruncher (4x) also on my QX6700 (1x).

If you have 0.04, IIRC this is because you used the Lunatics Installer for to install the opt. apps and in this app_info.xml is the 0.040000 entry.

I had stock 0.19, so I inserted 0.190000 in my app_info.xml .

Haha! I was way off. Not the first time...

Thanks Sutaru for explaining.

Edit: I used the Lunatics installer for opt apps and mine shows (.16). Would you know why it doesn't show .04 like you mentioned?

The (0.xx CPUs + 1.00 NVIDIA GPUs) are only for to calculate how much work you have and for the new work request.
It have nothing to do (with the current available clients) with the CPU support for the GPU.

If I had all stock (no opt. apps) I had with all my OCed GTX260-216 -> 0.19 CPUs.
On my GPU cruncher (4x) also on my QX6700 (1x).

If you have 0.04, IIRC this is because you used the Lunatics Installer for to install the opt. apps and in this app_info.xml is the 0.040000 entry.

I had stock 0.19, so I inserted 0.190000 in my app_info.xml .

Haha! I was way off. Not the first time...

Thanks Sutaru for explaining.

Edit: I used the Lunatics installer for opt apps and mine shows (.16). Would you know why it doesn't show .04 like you mentioned?

The value is the CPU time used...per Core per GPU. So it should actually be different on faster or slower computers.
It's also info for the BOINC client, when the value is 1 like Einstein, a whole core is used.
Some projects use 2 cores and the value is 2.

[...]
Edit: I used the Lunatics installer for opt apps and mine shows (.16). Would you know why it doesn't show .04 like you mentioned?

Maybe they changed the entry..

If you look in your app_info.xml, what is in the line..

<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.190000</avg_ncpus>
<flops>xxxxxxxxxxxxxx*</flops>
<plan_class>cuda</plan_class>


(I deleted the <max_.. entry in my app_info.xml)


[* my calculated GPU FLOPS]

---

The value is the CPU time used...per Core per GPU. So it should actually be different on faster or slower computers.
It's also info for the BOINC client, when the value is 1 like Einstein, a whole core is used.
Some projects use 2 cores and the value is 2.

Thanks! It's something I've wondered about several times now but never asked anyone.

[...]
Edit: I used the Lunatics installer for opt apps and mine shows (.16). Would you know why it doesn't show .04 like you mentioned?

Maybe they changed the entry..

If you look in your app_info.xml, what is in the line..

<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.190000</avg_ncpus>
<flops>xxxxxxxxxxxxxx*</flops>
<plan_class>cuda</plan_class>


(I deleted the <max_.. entry in my app_info.xml)


[* my calculated GPU FLOPS]

This is the whole app_info.xml

<app_info>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>AK_v8b_win_x64_SSE3.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>528</version_num>
<file_ref>
<file_name>AK_v8b_win_x64_SSE3.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>603</version_num>
<file_ref>
<file_name>AK_v8b_win_x64_SSE3.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>MB_6.08_CUDA_V12_VLARKill_FPLim2048.exe</name>
<executable/>
</file_info>
<file_info>
<name>cudart.dll</name>
<executable/>
</file_info>
<file_info>
<name>cufft.dll</name>
<executable/>
</file_info>
<file_info>
<name>libfftw3f-3-1-1a_upx.dll</name>
<executable/>
</file_info>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<plan_class>cuda</plan_class>
<avg_ncpus>0.040000</avg_ncpus>
<max_ncpus>0.040000</max_ncpus>
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>
<file_ref>
<file_name>MB_6.08_CUDA_V12_VLARKill_FPLim2048.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>cudart.dll</file_name>
</file_ref>
<file_ref>
<file_name>cufft.dll</file_name>
</file_ref>
<file_ref>
<file_name>libfftw3f-3-1-1a_upx.dll</file_name>
</file_ref>
</app_version>
</app_info>

So in my app_info.xml it shows the .04 but in my BOINC manager it shows .16



Edit: Those two CUDA running in the pic are Collatz Conjecture work units. Not sure if that matters.

Hi everybody

Stop me if I'm wrong : the more avg_ncpus you write in the app_info, the less you give to the CPU's core. In other words, it's probably good for the GPU but it slow down the core used to load the GPU.

That's why you have to find the best "share" (sorry, i don't find the word) between CPU and GPU depending of your hardware, right ?

NB : I'm using lunatics opt.

Bests regards,

Lionel.

Hi everybody

Stop me if I'm wrong : the more avg_ncpus you write in the app_info, the less you give to the CPU's core. In other words, it's probably good for the GPU but it slow down the core used to load the GPU.

That's why you have to find the best "share" (sorry, i don't find the word) between CPU and GPU depending of your hardware, right ?

NB : I'm using lunatics opt.

Bests regards,

Lionel.

The value has nothing to do with how fast everything goes.
A lower or higher value changes ... nothing in CPU and GPU speed.
I expect (hope) the scheduler does something with this value to know how much work to request.

Well, speed it is not exactly my purpose. What I am trying to understand is this parameter is reserving CPU time to load the GPU.
The higher value you give to this parameter, the more the CPU will working to load the GPU, is that correct ?

i. e. : you give 1 to this parameter means one core will work 100 % to load the GPU...

The value has nothing to do with how fast everything goes.
A lower or higher value changes ... nothing in CPU and GPU speed.
I expect (hope) the scheduler does something with this value to know how much work to request.

Makes sense to me.

---

Overclock with the MSI G31M3-L and Intel E8600 3.33Ghz
Intel D865GLC Socket 478 Motherboard
~How To Overclock The Eee ASUS 1005HA Netbook To 1.9Ghz~

Well, speed it is not exactly my purpose. What I am trying to understand is this parameter is reserving CPU time to load the GPU.
The higher value you give to this parameter, the more the CPU will working to load the GPU, is that correct ?

i. e. : you give 1 to this parameter means one core will work 100 % to load the GPU...

A lot of us would wish..... But the CPU fills as fast as it can. Normally around 5-10% If it is not busy with something else of course.

I got it.
Thanks a lot for helping me to understand.
Lionel.

The (0.xx CPUs + 1.00 NVIDIA GPUs) are only for to calculate how much work you have and for the new work request.
It have nothing to do (with the current available clients) with the CPU support for the GPU.

If I had all stock (no opt. apps) I had with all my OCed GTX260-216 -> 0.19 CPUs.
On my GPU cruncher (4x) also on my QX6700 (1x).

If you have 0.04, IIRC this is because you used the Lunatics Installer for to install the opt. apps and in this app_info.xml is the 0.040000 entry.

I had stock 0.19, so I inserted 0.190000 in my app_info.xml .

Haha! I was way off. Not the first time...

Thanks Sutaru for explaining.

Edit: I used the Lunatics installer for opt apps and mine shows (.16). Would you know why it doesn't show .04 like you mentioned?

The value is the CPU time used...per Core per GPU. So it should actually be different on faster or slower computers.
It's also info for the BOINC client, when the value is 1 like Einstein, a whole core is used.
Some projects use 2 cores and the value is 2.

This is what I meant, but didn't used enough words to explain..
English isn't my mother language, so not easy to explain what I have in my head..

The value you see in BOINC (and/or the entry in your app_info.xml) is for to calculate how much work you have on HDD and for to calculate how much new work you need for to hold your x day WU cache settings.

The current BOINC client don't limit the CPU support with this entry.
The GPU get this support from the CPU which is needed.

---

[...]
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<plan_class>cuda</plan_class>
<avg_ncpus>0.040000</avg_ncpus>
<max_ncpus>0.040000</max_ncpus>
<coproc>
<type>CUDA</type>
<count>1</count>
[...]

So in my app_info.xml it shows the .04 but in my BOINC manager it shows .16



Edit: Those two CUDA running in the pic are Collatz Conjecture work units. Not sure if that matters.

That it's..

This are Collatz WUs. I have no knowledge about this project. You let run stock or opt. app there? If opt., then you would find this entry in the Collatz app_info.xml .

The 0.040000 entry of the SETI@home app_info.xml you will see if SETI@home WUs are calculated.

---

[...]
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<plan_class>cuda</plan_class>
<avg_ncpus>0.040000</avg_ncpus>
<max_ncpus>0.040000</max_ncpus>
<coproc>
<type>CUDA</type>
<count>1</count>
[...]

So in my app_info.xml it shows the .04 but in my BOINC manager it shows .16



Edit: Those two CUDA running in the pic are Collatz Conjecture work units. Not sure if that matters.

That it's..

This are Collatz WUs. I have no knowledge about this project. You let run stock or opt. app there? If opt., then you would find this entry in the Collatz app_info.xml .

The 0.040000 entry of the SETI@home app_info.xml you will see if SETI@home WUs are calculated.

I temporarily suspended Collatz to make SETI Cuda run. Sure enough the line in BOINC manager for the running SETI CUDA WU's say .04 LOL

I understand now.. Thanks for your patients. :)

I'm not sure I understand. Nothing's changed on my system in the last week or so, but STATUS now says 0.34 CPUs + 1.00 NVIDIA GPU. This is on each of 2 different lines. My apps haven't changed, but it seems like overhead is climbing. The only GPU crunching occurs on SETI WU's, and the only other type of WU's I do crunch are Einstein's. Are their WU's just getting that more complicated? If so, you'd think the numbering scheme for each WU would be significantly different, but they're not.

---