Are you running the SoG 1 at a time on your 980 or 2 at a time.

Looking at your times, wanted to compare.

Are you running the SoG 1 at a time on your 980 or 2 at a time.

Looking at your times, wanted to compare.

2 at a time.

Can be seen in my stderr:

<core_client_version>7.4.42</core_client_version>
<![CDATA[
<stderr_txt>
Running on device number: 0
Number of app instances per device set to:2
Priority of worker thread raised successfully
Priority of process adjusted successfully, below normal priority class used
OpenCL platform detected: Intel(R) Corporation
OpenCL platform detected: NVIDIA Corporation
BOINC assigns device 0
Info: BOINC provided OpenCL device ID used

OK, SoG now seems to be the preferred app. I will now go to 3 at a time, and add
-sbs256 -hp to the mb_cmdline-8.07_windows_intel__opencl_nvidia_SoG.txt file.

May Dog protect me :-)

Edit: I will not add -sbs256 -hp for now. Just go for 3 at a time. Maybe I'll add that tomorrow, after having been at 3 at a time for 24 hours.

Edit: I will not add -sbs256 -hp for now. Just go for 3 at a time. Maybe I'll add that tomorrow, after having been at 3 at a time for 24 hours.

If GPU already loaded ~100% hardly 3 at time will do speedup. -sbs tuning and oclFFT tuning can improve speed further cause they control different areas of app.

---

News about SETI opt app releases: https://twitter.com/Raistmer

Let me know what you find

Edit: I will not add -sbs256 -hp for now. Just go for 3 at a time. Maybe I'll add that tomorrow, after having been at 3 at a time for 24 hours.

If GPU already loaded ~100% hardly 3 at time will do speedup. -sbs tuning and oclFFT tuning can improve speed further cause they control different areas of app.

3 at a time did speedup.

1 on 3 minutes 40 seconds
2 on 6 minutes 50-55 seconds
3 on 10 minutes 10-20 seconds

oclFFT tuning will have to wait until the appropriate readme file for this app is available.

oclFFT tuning will have to wait until the appropriate readme file for this app is available.

https://docs.google.com/spreadsheets/d/1ZkvwecPl21Ig6H66CUInDyEQnV_HJE1mKIqY6FwwikY/edit?usp=sharing
This class of options tunes oclFFT performance
-oclfft_tune_gr N : Global radix
-oclfft_tune_lr N : Local radix
-oclfft_tune_wg N : Workgroup size
-oclfft_tune_ls N : Max size of local memory FFT
-oclfft_tune_bn N : Number of local memory banks
-oclfft_tune_cw N : Memory coalesce width

That's all available for this options set.

---

News about SETI opt app releases: https://twitter.com/Raistmer

oclFFT tuning will have to wait until the appropriate readme file for this app is available.

https://docs.google.com/spreadsheets/d/1ZkvwecPl21Ig6H66CUInDyEQnV_HJE1mKIqY6FwwikY/edit?usp=sharing
This class of options tunes oclFFT performance
-oclfft_tune_gr N : Global radix
-oclfft_tune_lr N : Local radix
-oclfft_tune_wg N : Workgroup size
-oclfft_tune_ls N : Max size of local memory FFT
-oclfft_tune_bn N : Number of local memory banks
-oclfft_tune_cw N : Memory coalesce width

That's all available for this options set.

Thanks, that's all good and well, but it's Chinese to me. I haven't got a clue which settings should be appropriate for my GTX980.

oclFFT tuning will have to wait until the appropriate readme file for this app is available.

https://docs.google.com/spreadsheets/d/1ZkvwecPl21Ig6H66CUInDyEQnV_HJE1mKIqY6FwwikY/edit?usp=sharing
This class of options tunes oclFFT performance

Spreadsheet only addresses SSE2 on ATI cards??

---

If I can help out by testing something, please let me know.
Available hardware and software is listed in my profile here.

Thanks, that's all good and well, but it's Chinese to me. I haven't got a clue which settings should be appropriate for my GTX980.

I think no one have until tested. But cause quite a lot combos can return invalid data it's more suitable for offline testing. As first approximation same values that Mike recommends for high-end ATi GPUs can be tried.

---

News about SETI opt app releases: https://twitter.com/Raistmer

Spreadsheet only addresses SSE2 on ATI cards??

There is no "SSE2 on ATi card" Test was done with listed app on C-60 APU. SSE2 is irrelevant. No available systematic data for NV (at least no one cared to share).

---

News about SETI opt app releases: https://twitter.com/Raistmer

Thanks, that's all good and well, but it's Chinese to me. I haven't got a clue which settings should be appropriate for my GTX980.

I think no one have until tested. But cause quite a lot combos can return invalid data it's more suitable for offline testing. As first approximation same values that Mike recommends for high-end ATi GPUs can be tried.

I found this in the readme file "ReadMe_MultiBeam_OpenCL_ATI.txt" in the latest installer. I'll try the high-end settings. Again, may Dog protect me :-)

============= NV specific info ================

Known issues:

With NV drivers past 267.xx GPU usage can be low if CPU fully used with another loads.
App performance can be increased by using -cpu_lock switch in this case, and CPU time
savings are possible with -use_sleep switch.

Suggested command line switches:

Entry Level cards NV x20 x30 x40
-sbs 128 -spike_fft_thresh 2048 -tune 1 2 1 16
(*requires testing)

Mid range cards x50 x60 x70
-sbs 192 -spike_fft_thresh 2048 -tune 1 64 1 4
(*requires testing)

High end cards x8x 780TI Titan / Titan Z
-sbs 256 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64
(*requires testing)

===============================================

Geeze: That setting shaved more than a minute from the approx 10 minutes 20-30 seconds, running 3 at a time.

CPU usage went up a bit to approx 14-17%, from 9-10%, but that doesn't really bother me. Edit: Strike that. It was only the first tasks, while the .bin files was being recompiled.

So, for my GTX980 Strix, this setting seems to work well:

-sbs 256 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64

I also have -instances_per_device 3 in the settings, just to make it visible in the stderr, the number of tasks I run simultaneously.

Edit: Dropping -sbs 256 from the settings, (going to default 128), seems to have made it even faster.

Is this something that will translate over to ATI cards or will it be Nvidia specific?

Thanks,

Chris

Thanks, that's all good and well, but it's Chinese to me. I haven't got a clue which settings should be appropriate for my GTX980.

I think no one have until tested. But cause quite a lot combos can return invalid data it's more suitable for offline testing. As first approximation same values that Mike recommends for high-end ATi GPUs can be tried.

I found this in the readme file "ReadMe_MultiBeam_OpenCL_ATI.txt" in the latest installer. I'll try the high-end settings. Again, may Dog protect me :-)

============= NV specific info ================

Known issues:

With NV drivers past 267.xx GPU usage can be low if CPU fully used with another loads.
App performance can be increased by using -cpu_lock switch in this case, and CPU time
savings are possible with -use_sleep switch.

Suggested command line switches:

Entry Level cards NV x20 x30 x40
-sbs 128 -spike_fft_thresh 2048 -tune 1 2 1 16
(*requires testing)

Mid range cards x50 x60 x70
-sbs 192 -spike_fft_thresh 2048 -tune 1 64 1 4
(*requires testing)

High end cards x8x 780TI Titan / Titan Z
-sbs 256 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64
(*requires testing)

===============================================

I have tested it on a 980 also.
Otherwise there were a 980 section as well.

-sbs value is host dependend.
On my GPU -sbs 384 is fastest.
Thats why i added "requires testing" to the read me.

---

With each crime and every kindness we birth our future.

I have tested it on a 980 also.
Otherwise there were a 980 section as well.

-sbs value is host dependend.
On my GPU -sbs 384 is fastest.
Thats why i added "requires testing" to the read me.

Thanks Mike. You're doing a good job with all these testings of various "extra" settings.

BETA is out of work for the moment, so I couldn't continue testing. Let's hope somone kicks the splitters, so I can continue with my SoG crunching.

Eric,

You're supplied an app_info.xml with the SETI@home v8 v8.00 (armv6l)arm-unknown-linux-gnueabihf app, I found it in the slot directory as well as the project directory, and referenced in the client_state.xml

Was that intentional?, Now if anyone restarts Boinc (or the host) on a host that has had that app then they switch to anonymous platform, and will remain only using that app.

Claggy

Hi all!

I received this message today -

Mon Feb 8 09:28:23 2016 | SETI@home Beta Test | [coproc] ATI instance 0: confirming 1.000000 instance for 23ap11ag.31103.495.3.37.178_0
Mon Feb 8 09:29:19 2016 | SETI@home Beta Test | Task postponed: Suspicious spike results, host needs reboot or maintenance


Any idea what this means?

My configuration:
GenuineIntel Intel(R) Core(TM) i7 CPU 860 @ 2.80GHz [x86 Family 6 Model 30 Stepping 5] (8 processors)
AMD Radeon HD 4850 (512MB) OpenCL: 1.0
Darwin 15.3.0
BOIC 7.6.22

Eric,

You're supplied an app_info.xml with the SETI@home v8 v8.00 (armv6l)arm-unknown-linux-gnueabihf app, I found it in the slot directory as well as the project directory, and referenced in the client_state.xml

Unintentional. :( Will be releasing new version today.

---

And what about limited work supply? Intentional or not? ;)
Cause my host can't get work for awhile already...

---

News about SETI opt app releases: https://twitter.com/Raistmer