optimized sources

[corsair]

Hi corsair, distrrtgen is downloading programs automatic onto your machine, nothing todo.

Thanks a lot _heinz already notice that but seen somewhere that there is people compiling it's own builds ??

[_heinz]

Titan, first results on seti forum
not so impressed as we thought  , the card does not use its full potential.
surprize....

_heinz

[Mike]

I`m not surprised at all.
It was similar with Tesla back then.

[_heinz]

I`m not surprised at all.
It was similar with Tesla back then.

Looks like a complete redesign of the app is necessary to use Tesla`s and Titan`s properties optimal.

_heinz

[_heinz]

i7-3630QM_primegrid_1Mio cpu-work
time goes on, V8-Xeon is nearly 4 years old, time to give it a hardware upgrade.

_heinz

installed 2 GTX Titan EVGA SC SLI and GTX570
Boinc shows:
08.03.2013 18:43:27 |  | NVIDIA GPU 0: GeForce GTX TITAN (driver version 314.14, CUDA version 5.0, compute capability 3.5, 4096MB, 4096MB available, 4989 GFLOPS peak)
08.03.2013 18:43:27 |  | NVIDIA GPU 1: GeForce GTX TITAN (driver version 314.14, CUDA version 5.0, compute capability 3.5, 4096MB, 4096MB available, 4989 GFLOPS peak)
08.03.2013 18:43:27 |  | NVIDIA GPU 2: GeForce GTX 570 (driver version 314.14, CUDA version 5.0, compute capability 2.0, 1280MB, 1178MB available, 1405 GFLOPS peak)
08.03.2013 18:43:27 |  | OpenCL: NVIDIA GPU 0: GeForce GTX TITAN (driver version 314.14, device version OpenCL 1.1 CUDA, 6144MB, 4096MB available)
08.03.2013 18:43:27 |  | OpenCL: NVIDIA GPU 1: GeForce GTX TITAN (driver version 314.14, device version OpenCL 1.1 CUDA, 6144MB, 4096MB available)
08.03.2013 18:43:27 |  | OpenCL: NVIDIA GPU 2: GeForce GTX 570 (driver version 314.14, device version OpenCL 1.1 CUDA, 1280MB, 1178MB available)
GTX_Titan_SLI_ready
GTX_Titan_SLI_working
GPUZ_GTX_Titan
GPUZ_Sensors_GTX_Titan
Happy crunching, a lot todo now.
_heinz

[William]

I`m not surprised at all.
It was similar with Tesla back then.

Looks like a complete redesign of the app is necessary to use Tesla`s and Titan`s properties optimal.

_heinz

Everyone knows the answer is 42.

To paraphrase Jason, He's pretty pleased that his code scales so well on new architecture. There's still a lot of optimisation potential left.
Currently the Titans underperform (when you compare to e.g. a 690) by about 30% - IOW from the specs you would expect some 30% more speed.
According to Jason, drivers for new cards take a year or so to mature - in that time there's quite a bit of speed improvement (from our POV), so I'd expect when the drivers are good, the Titans scale according to their specs. Still a lot of potential left and a lot of stuff to explore.

To me, looks like the wrong bandwagon to jump on.

[Mike]

I totally agree William.

[_heinz]

I`m not surprised at all.
It was similar with Tesla back then.

Looks like a complete redesign of the app is necessary to use Tesla`s and Titan`s properties optimal.

_heinz

Everyone knows the answer is 42.

 Still a lot of potential left and a lot of stuff to explore.

I know-->42 
greetings

_heinz
edit:
passed 500Mio_total today
last month credit go's up to 200Mio_distrrtgen_total

[_heinz]

driver 314.21:
17.03.2013 15:42:01 |  | CUDA: NVIDIA GPU 0: GeForce GTX TITAN (driver version 314.21, CUDA version 5.0, compute capability 3.5, 4096MB, 4096MB available, 4989 GFLOPS peak)
17.03.2013 15:42:01 |  | CUDA: NVIDIA GPU 1: GeForce GTX TITAN (driver version 314.21, CUDA version 5.0, compute capability 3.5, 4096MB, 4096MB available, 4989 GFLOPS peak)
17.03.2013 15:42:01 |  | CUDA: NVIDIA GPU 2: GeForce GTX 570 (driver version 314.21, CUDA version 5.0, compute capability 2.0, 1280MB, 1178MB available, 1405 GFLOPS peak)
17.03.2013 15:42:01 |  | OpenCL: NVIDIA GPU 0: GeForce GTX TITAN (driver version 314.21, device version OpenCL 1.1 CUDA, 6144MB, 4096MB available, 4989 GFLOPS peak)
17.03.2013 15:42:01 |  | OpenCL: NVIDIA GPU 1: GeForce GTX TITAN (driver version 314.21, device version OpenCL 1.1 CUDA, 6144MB, 4096MB available, 4989 GFLOPS peak)
17.03.2013 15:42:01 |  | OpenCL: NVIDIA GPU 2: GeForce GTX 570 (driver version 314.21, device version OpenCL 1.1 CUDA, 1280MB, 1178MB available, 1405 GFLOPS peak)
17.03.2013 15:42:01 | DistrRTgen | Found app_info.xml; using anonymous platform
17.03.2013 15:42:01 |  | Config: simulate 8 CPUs
17.03.2013 15:42:01 |  | Config: use all coprocessors
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
no news, machine running standard frequency produced continous ~ 2,1 Mio/day as before

[_heinz]

After testing and running Zdenek's compiled app CC2.0 for GTX570 and Titan daily output increased up to 4,3 Mio/day on 2013-03-28 
first step is done.
_heinz

[_heinz]

Joyeuses Pâques
Frohe Ostern
Happy Easter
Thank you to all readers not lost interest looking up here.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Try the new astropulse OCL apps from Raistmer and Urs Echternacht

_heinz

[_heinz]

Because I had have trouble to run BOINC multiple clients with mixed configuration GTXTitan /GTX570 to give every card its optimal client, I decided to remove GTX570 and run now triple SLI GTX Titan
31.03.2013 19:33:12 |  | Processor: 8 GenuineIntel Intel(R) Xeon(R) CPU           E5405  @ 2.00GHz [Family 6 Model 23 Stepping 6]
31.03.2013 19:33:12 |  | Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 cx16 sse4_1 syscall nx lm vmx tm2 dca pbe
31.03.2013 19:33:12 |  | OS: Microsoft Windows 7: Ultimate x64 Edition, Service Pack 1, (06.01.7601.00)
31.03.2013 19:33:12 |  | Memory: 16.00 GB physical, 31.99 GB virtual
31.03.2013 19:33:12 |  | Disk: 931.51 GB total, 840.36 GB free
31.03.2013 19:33:12 |  | Local time is UTC +2 hours
31.03.2013 19:33:12 |  | CUDA: NVIDIA GPU 0: GeForce GTX TITAN (driver version 314.22, CUDA version 5.0, compute capability 3.5, 4096MB, 4096MB available, 4989 GFLOPS peak)
31.03.2013 19:33:12 |  | CUDA: NVIDIA GPU 1: GeForce GTX TITAN (driver version 314.22, CUDA version 5.0, compute capability 3.5, 4096MB, 4096MB available, 4989 GFLOPS peak)
31.03.2013 19:33:12 |  | CUDA: NVIDIA GPU 2: GeForce GTX TITAN (driver version 314.22, CUDA version 5.0, compute capability 3.5, 4096MB, 4096MB available, 4989 GFLOPS peak)
31.03.2013 19:33:12 |  | OpenCL: NVIDIA GPU 0: GeForce GTX TITAN (driver version 314.22, device version OpenCL 1.1 CUDA, 6144MB, 4096MB available, 4989 GFLOPS peak)
31.03.2013 19:33:12 |  | OpenCL: NVIDIA GPU 1: GeForce GTX TITAN (driver version 314.22, device version OpenCL 1.1 CUDA, 6144MB, 4096MB available, 4989 GFLOPS peak)
31.03.2013 19:33:12 |  | OpenCL: NVIDIA GPU 2: GeForce GTX TITAN (driver version 314.22, device version OpenCL 1.1 CUDA, 6144MB, 4096MB available, 4989 GFLOPS peak)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
GTX_Titan_3SLI_working
measured power 960Watt
Production output ~4,3Mio/day continous

_heinz

[_heinz]

27th march 2013, excluded my old P4 2,66Mhz AGP HD4670 from crunching after 5Mio on primegrid
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
today 2013-04-12 I'm at the same boinc_place_291 as two years ago on 2011-04-10.
RAC is still climbing..

[_heinz]

Although distrrtgen reduced its credit by longer wu's, production output is now ~5,2Mio/day and RAC is still climbing.
V8-Xeon calculate now a distrrtgen wu in 929 - 1060 sec, that is as fast as HD7970 and a little bit faster.
The differences in runtime comes from different clock speeds of the cards. The upper two cards of dev 0 and 1 are hotter and run with slower clocks than dev 2 the undermost card.
modify:
With boinc_avg_number_29 I'm now back in the first 30 worldwide.
Back to the first 10 tophosts distrrtgen_tophosts_number_9
One of the first 20 top user distrrtgen_topuser_number_19

[_heinz]

For all who are interested in technical datails of GK110, here there are:
aida64_grafikprozessor_GTX_Titan
aida64_GPGPU_CUDA_GTX_Titan_properties_1
aida64_GPGPU_CUDA_GTX_Titan_properties_2
aida64_GPGPU_Direct3D_GTX_Titan_properties_3
aida64_GPGPU_OpenCL_GTX_Titan_properties_4
aida64_GPGPU_OpenCL_GTX_Titan_properties_5
aida64_GPGPU_OpenCL_GTX_Titan_properties_6
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
My answer to: http://lunatics.kwsn.net/2-windows/optimized-sources.msg49420.html#msg49420
As soon as I have a new PSU, I will test v8-Xeon with distrrtgen to get real data.
here there are:

The question is:How many KWh need my machines to get 1 Mio Cobblestone?
Calculated for distrrtgen
1 WU = 19825 Points
1000000 / 19825 = 50,441 ~ 51WU's needed.

Machine 4: V8-Xeon, 2,4GHz  3xNVIDIA GTX Titan
GTX Titan need ~1006 sec/WU, 51WU*1006sec/3600= 14,25h/Mio
While we have 3 Titan we must divide by 3
14,25h/3=4,75h per Mio
Measured crunch power = 860W
860W*4,75h=4085Wh ~4,085KWh
1Mio = 4,085KWh
~~~~~~~~~~~~~~~~~
Machine 5: Laptop Acer Aspire V3-771G, i7-3630QM 2,4GHz NVIDIA GT650M
Duration per WU = 10246 sec
51WU*10246sec/3600= 145,15h/Mio
crunch power ~60W
60W*145,15h=8709Wh ~8,709KWh
1Mio = 8,709KWh
~~~~~~~~~~~~~~~~~
Machine 3: ATOM 1,6GHz, ION GPU
Duration per WU = 111283sec
51wu * 111283/3600 = 1576,50h/Mio
Measured crunch-power = 28W
1576,5h*28W=44142Wh ~44,14KWh
1Mio = 44,14KWh
~~~~~~~~~~~~~~~~