Hi,

Does 6.6.20 handle MB-only on CPU and CUDA? There are quite a few combos arund, and one of my computers has ample GPU power but not quite up to par when it comes to CPU (AP). I therefore only want the CPUs and GPU to process S@H MB.

My test so far only downloads 603 app and thus the GPU is idle. Boinc Manager does not log any errors or missing files, so it appears this is out of app_info control...?

Excerpt from log:

18.04.2009 14:17:26 Starting BOINC client version 6.6.20 for windows_intelx86
18.04.2009 14:17:26 log flags: task, file_xfer, sched_ops
18.04.2009 14:17:26 Libraries: libcurl/7.19.4 OpenSSL/0.9.8j zlib/1.2.3
18.04.2009 14:17:26 Data directory: C:\ProgramData\BOINC
18.04.2009 14:17:26 Running under account exchTestBruker
18.04.2009 14:17:26 SETI@home Found app_info.xml; using anonymous platform
18.04.2009 14:17:26 Processor: 8 GenuineIntel Intel(R) Xeon(R) CPU E5320 @ 1.86GHz [x86 Family 6 Model 15 Stepping 7]
18.04.2009 14:17:26 Processor features: fpu tsc pae nx sse sse2 pni mmx
18.04.2009 14:17:26 OS: Microsoft Windows 7: x86 Editon, (06.01.7000.00)
18.04.2009 14:17:26 Memory: 3.00 GB physical, 6.29 GB virtual
18.04.2009 14:17:26 Disk: 135.89 GB total, 13.96 GB free
18.04.2009 14:17:26 Local time is UTC +2 hours
18.04.2009 14:17:26 CUDA device: GeForce GTX 285 (driver version 18171, CUDA version 1.3, 1024MB, est. 127GFLOPS)
18.04.2009 14:17:26 Not using a proxy
18.04.2009 14:17:26 SETI@home URL: http://setiathome.berkeley.edu/; Computer ID: 3447503; location: home; project prefs: default
18.04.2009 14:17:26 SETI@home General prefs: from SETI@home (last modified 10-Apr-2009 19:59:40)
18.04.2009 14:17:26 SETI@home Computer location: home
18.04.2009 14:17:26 General prefs: using separate prefs for home
18.04.2009 14:17:26 Reading preferences override file
18.04.2009 14:17:26 Preferences limit memory usage when active to 1534.37MB
18.04.2009 14:17:26 Preferences limit memory usage when idle to 2761.86MB
18.04.2009 14:17:26 Preferences limit disk usage to 14.03GB
18.04.2009 14:17:26 SETI@home Restarting task 09fe09ac.29909.214879.14.8.139_1 using setiathome_enhanced version 603


Morten

---

Morten Ross

Hi,

Does 6.6.20 handle MB-only on CPU and CUDA? There are quite a few combos arund, and one of my computers has ample GPU power but not quite up to par when it comes to CPU (AP). I therefore only want the CPUs and GPU to process S@H MB.

My test so far only downloads 603 app and thus the GPU is idle. Boinc Manager does not log any errors or missing files, so it appears this is out of app_info control...?

Excerpt from log:

18.04.2009 14:17:26 Starting BOINC client version 6.6.20 for windows_intelx86
18.04.2009 14:17:26 log flags: task, file_xfer, sched_ops
18.04.2009 14:17:26 Libraries: libcurl/7.19.4 OpenSSL/0.9.8j zlib/1.2.3
18.04.2009 14:17:26 Data directory: C:\ProgramData\BOINC
18.04.2009 14:17:26 Running under account exchTestBruker
18.04.2009 14:17:26 SETI@home Found app_info.xml; using anonymous platform
18.04.2009 14:17:26 Processor: 8 GenuineIntel Intel(R) Xeon(R) CPU E5320 @ 1.86GHz [x86 Family 6 Model 15 Stepping 7]
18.04.2009 14:17:26 Processor features: fpu tsc pae nx sse sse2 pni mmx
18.04.2009 14:17:26 OS: Microsoft Windows 7: x86 Editon, (06.01.7000.00)
18.04.2009 14:17:26 Memory: 3.00 GB physical, 6.29 GB virtual
18.04.2009 14:17:26 Disk: 135.89 GB total, 13.96 GB free
18.04.2009 14:17:26 Local time is UTC +2 hours
18.04.2009 14:17:26 CUDA device: GeForce GTX 285 (driver version 18171, CUDA version 1.3, 1024MB, est. 127GFLOPS)
18.04.2009 14:17:26 Not using a proxy
18.04.2009 14:17:26 SETI@home URL: http://setiathome.berkeley.edu/; Computer ID: 3447503; location: home; project prefs: default
18.04.2009 14:17:26 SETI@home General prefs: from SETI@home (last modified 10-Apr-2009 19:59:40)
18.04.2009 14:17:26 SETI@home Computer location: home
18.04.2009 14:17:26 General prefs: using separate prefs for home
18.04.2009 14:17:26 Reading preferences override file
18.04.2009 14:17:26 Preferences limit memory usage when active to 1534.37MB
18.04.2009 14:17:26 Preferences limit memory usage when idle to 2761.86MB
18.04.2009 14:17:26 Preferences limit disk usage to 14.03GB
18.04.2009 14:17:26 SETI@home Restarting task 09fe09ac.29909.214879.14.8.139_1 using setiathome_enhanced version 603


Morten

You could try the following, but no warranty expressed or implied...

Its a copy of the original app_info minus the astropulse bits. I've even calculated the flops figure for your GTX285, but you'll need to do the cpu flops numbers as I don't have access to your <f_pops> figure. See the 1st message in this thread for instructions on how to calculate this number.

<app_info>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>AK_v8_win_SSSE3x.exe</name>
<executable/>
</file_info>
<file_info>
<name>setiathome_6.08_windows_intelx86__cuda.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>603</version_num>
<platform>windows_intelx86</platform>
<flops>4127010920</flops>
<file_ref>
<file_name>AK_v8_win_SSSE3x.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.127970</avg_ncpus>
<max_ncpus>0.127970</max_ncpus>
<flops>25400000000</flops>
<plan_class>cuda</plan_class>
<file_ref>
<file_name>setiathome_6.08_windows_intelx86__cuda.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>
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>
</app_version>
</app_info>

---

BOINC blog

Hi,

Does 6.6.20 handle MB-only on CPU and CUDA? There are quite a few combos arund, and one of my computers has ample GPU power but not quite up to par when it comes to CPU (AP). I therefore only want the CPUs and GPU to process S@H MB.

My test so far only downloads 603 app and thus the GPU is idle. Boinc Manager does not log any errors or missing files, so it appears this is out of app_info control...?

I don't see why you feel your CPU isn't up to AP work: my dual E5320 is churning them out quite happily.

But returning to your question: yes, BOINC v6.6.20 can handle MB/CPU v6.03 and MB/CUDA v6.08 at the same time. As others have noted, the SETI scheduler can be a bit reluctant to hand out MB work to CPUs (I think all the CUDA hosts gobble it up first, and the CPUs only get left-overs if they're lucky). Provided you've set up the app_info.xml as in the opening post of this thread (minus the AP bits if you really don't want to run them), then all should be well.

Perhaps check the SETI@home 'Properties' (new command button on the projects tab), to make sure your CUDA work-fetch isn't suffering any sort of back-off.

....but you'll need to do the cpu flops numbers as I don't have access to your <f_pops> figure.

<p_fpops> (sic) for a Core2 is often very close to the speed in GHz. My E5320 is showing <p_fpops> = 1828798802.551730 - and with a Xeon, there won't be any overclocking to worry about.

....but you'll need to do the cpu flops numbers as I don't have access to your <f_pops> figure.

<p_fpops> (sic) for a Core2 is often very close to the speed in GHz. My E5320 is showing <p_fpops> = 1828798802.551730 - and with a Xeon, there won't be any overclocking to worry about.

I keep getting that one wrong.

@ Morten, use the <p_fpops> number to calculate the <flops> value.

I just got myself a pair of GTS250's so have one more cuda-capable machine. I swapped out a 9800GT from one and put the other one in a machine that didn't have any. I like them so much I ordered another pair :-)

---

BOINC blog

Hi,

All is ok - it just took many downloads batches before 608 was downloaded. Why I don't know.

Thanks for your feedback.

Morten

---

Morten Ross

I don't see why you feel your CPU isn't up to AP work: my dual E5320 is churning them out quite happily.

Hi

This is a Dell computer and it cannot be overclocked unless I take my chances with replacing the BIOS, so the chrunch-time IMO is just too slow.

Morten

---

Morten Ross

I have a Dell 2950 with two 5320s running Astropulse at stock speed. It has an RAC >10,000 and 1.8 million total credit.

---

--
Classic 82353 WU / 400979 h

I seem to have done something somewhere as processing times on my work units have increased dramatically.

Here are the specs and my app_info file, I am using optimized clients

Found app_info.xml; using anonymous platform
Processor: 2 GenuineIntel Intel(R) Core(TM)2 CPU 6400 @ 2.13GHz [x86 Family 6 Model 15 Stepping 6]
Processor features: fpu tsc pae nx sse sse2 mmx
CUDA device: GeForce 8500 GT (driver version 18250, CUDA version 1.1, 1024MB, est. 5GFLOPS)

<app_info>
<app>
<name>astropulse</name>
</app>
<file_info>
<name>ap_5.00r103_SSE3.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse</app_name>
<version_num>500</version_num>
<flops>4643141391</flops>
<file_ref>
<file_name>ap_5.00r103_SSE3.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>astropulse_v5</name>
</app>
<file_info>
<name>ap_5.03r112_SSE3.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse_v5</app_name>
<version_num>503</version_num>
<flops>4663777574.96</flops>
<file_ref>
<file_name>ap_5.03r112_SSE3.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>MB_6.08_mod_CUDA_V11_VLARKill_refined.exe</name>
<executable/>
</file_info>
<file_info>
<name>MB_6.08_mod_CUDA_V11_VLARKill_refined.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>603</version_num>
<platform>windows_intelx86</platform>
<flops>3611332193</flops>
<file_ref>
<file_name>MB_6.08_mod_CUDA_V11_VLARKill_refined.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.127970</avg_ncpus>
<max_ncpus>0.127970</max_ncpus>
<flops>10000000000</flops>
<plan_class>cuda</plan_class>
<file_ref>
<file_name>MB_6.08_mod_CUDA_V11_VLARKill_refined.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>
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>
</app_version>
</app_info>

Any help would be appreciated. If there is an error somewhere in my app_info can someone point it out please.

Cuda tasks seem to take 2hrs (0.13 cpus, I cuda)
AP seems to be 35-40 hours.

---

Your app_info has 'MB_6.08_mod_CUDA_V11_VLARKill_refined.exe' referenced too often,
With that app_info you will be trying to run 3 Cuda WU's at once, and none on the CPU,
you need to change the 603 entry to an AK_V8 CPU app, and one of the

<file_info>
<name>MB_6.08_mod_CUDA_V11_VLARKill_refined.exe</name>
<executable/>

to the same CPU app.

Claggy

Had a feeling it might have been that but wasn't 100% sure.

Thank you for the response.

---

I have a Dell 2950 with two 5320s running Astropulse at stock speed. It has an RAC >10,000 and 1.8 million total credit.

Mine has a rac of ~6000 and that's with CPU+GPU and you're rac'ing 10k on stock CPU speed without GPU - that's really odd! A high percentage of AP crunching cannot explain the rac difference...

I haven't tried with current AP, so based on your input I'll do a test run for some weeks to see how the rac changes.

Morten

---

Morten Ross

So this will be correct then now?

<app_info>
<app>
<name>astropulse</name>
</app>
<file_info>
<name>ap_5.00r103_SSE3.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse</app_name>
<version_num>500</version_num>
<flops>4643141391</flops>
<file_ref>
<file_name>ap_5.00r103_SSE3.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>astropulse_v5</name>
</app>
<file_info>
<name>ap_5.03r112_SSE3.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse_v5</app_name>
<version_num>503</version_num>
<flops>4663777574.96</flops>
<file_ref>
<file_name>ap_5.03r112_SSE3.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app>
<name>setiathome_enhanced</name>
</app>
<file_info>
<name>MB_6.08_mod_CUDA_V11_VLARKill_refined.exe</name>
<executable/>
</file_info>
<file_info>
<name>AK_v8_win_SSE3.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>603</version_num>
<platform>windows_intelx86</platform>
<flops>3611332193</flops>
<file_ref>
<file_name>AK_v8_win_SSE3.exe</file_name>
<main_program/>
</file_ref>
</app_version>
<app_version>
<app_name>setiathome_enhanced</app_name>
<version_num>608</version_num>
<platform>windows_intelx86</platform>
<avg_ncpus>0.127970</avg_ncpus>
<max_ncpus>0.127970</max_ncpus>
<flops>10000000000</flops>
<plan_class>cuda</plan_class>
<file_ref>
<file_name>MB_6.08_mod_CUDA_V11_VLARKill_refined.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>
<coproc>
<type>CUDA</type>
<count>1</count>
</coproc>
</app_version>
</app_info>

---

I don't see why you feel your CPU isn't up to AP work: my dual E5320 is churning them out quite happily.

Hi

This is a Dell computer and it cannot be overclocked unless I take my chances with replacing the BIOS, so the chrunch-time IMO is just too slow.

Morten

Mine is also a Dell (Precision 490): I don't think even a BIOS replacement would allow overclocking! The speed of these Xeons is controlled by voltages applied to pins in the CPU holder: overclocking involves covering them with little bits of insulating tape....

But I don't mind running AP on it: they're positively racy compared with the 4-month CPDN tasks I'm running on three of the cores.

I have a Dell 2950 with two 5320s running Astropulse at stock speed. It has an RAC >10,000 and 1.8 million total credit.

Mine has a rac of ~6000 and that's with CPU+GPU and you're rac'ing 10k on stock CPU speed without GPU - that's really odd!

Are you running optimized apps?

---

--
Classic 82353 WU / 400979 h

That looks Better, But replace the AK_V8_SSE3 app with the SSSE3 app,
use CPU-Z to confirm your C2D supports SSSE3 first,

Claggy

I have a Dell 2950 with two 5320s running Astropulse at stock speed. It has an RAC >10,000 and 1.8 million total credit.

Mine has a rac of ~6000 and that's with CPU+GPU and you're rac'ing 10k on stock CPU speed without GPU - that's really odd!

Are you running optimized apps?

I sure do - Raistmers v10a, but now I've downloaded lot's of APs in order to see if that'll up the rac (+4000 that is...).

Morten

---

Morten Ross

I've downloaded lot's of APs in order to see if that'll up the rac (+4000 that is...).

Be very patient.

---

--
Classic 82353 WU / 400979 h

To FLOPS or not to FLOPS ?

I've got two questions in regards to the flops-entry:

1: Will a future BOINC version control this calculation?
2: What is the consequence of dropping the flops entry in app_info?

When you have many computers this current implementation is very time comsuming and it's also very easy to miscalculate.

Morten

---

Morten Ross

What is the consequence of dropping the flops entry in app_info?

Very little to none. Your initial estimates for time required for tasks will be off, but BOINC figures them out soon enough. They're not very accurate anyway. I quit entering values for flops and platform in app_info.xml quickly. With a mix of machines running 32 and 64 bit operating systems with varying CUDA capability, I found it to be WAY more trouble than it's worth.

---

--
Classic 82353 WU / 400979 h