optimized sources

[_heinz]

VoIP test statistics
--------------------
Jitter: you --> server: 0.8 ms
Jitter: server --> you: 3.9 ms
Packet loss: you --> server: 0.0 %
Packet loss: server --> you: 0.0 %
Packet discards: 0.0 %
Packets out of order: 0.0 %
Estimated MOS score: 4.0

Speed test statistics
---------------------
Download speed: 1016312 bps
Upload speed: 274936 bps
Download quality of service: 98 %
Upload quality of service: 95 %
Download test type: socket
Upload test type: socket
Maximum TCP delay: 49 ms
Average download pause: 28 ms
Minimum round trip time to server: 213 ms
Average round trip time to server: 217 ms
Estimated download bandwidth: 1840000bps
Route concurrency: 1.8104676
Download TCP forced idle: 0 %
Maximum route speed: 2461408bps
~~~~~~~~~~~~~~~~~~~~~~~~
done by modifying some ADSL parameters of DG834PN ADSL2+ router
PPoA to PPoE
use service G992_3_A
Multiplexmethod: LLC-BASED
VPI: 8
VCI: 35
~~~~~~~~~~~~~
it is now near the max I can get. 
heinz

[_heinz]

By the way, the new GTX580 is already available in Germany (479,99 Euro)
Run my GPU CUDA ? Most questions are answered there:
List of CUDA GPU's
And CUDA Toolkit 3.2 RC2 is available now.
~~~~~~~~~~~~~~~~~~~~~~~~~~
Have todo some updates now.

heinz

[RottenMutt]

...
By the way, the new GTX580 is already available in Germany (479,99 Euro)
...

heinz

no need to get one with seti being down...  yes nvidia, some of your bussness is from crunchers..

[_heinz]

I installed CUDA 3.2 RC2 Release64 on V8-SK01
If you want to compile all samples from the SDK you must install latest DXSDK (June 2010)
compile all samples from CUDASDK
========== Alles neu erstellen: 92 erfolgreich, Fehler bei 1, 0 übersprungen ==========
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 error in project simpleD3D11Texture, library "D3D11Effects_vc9.lib" not found.
If we lookup in the project dir under ./d3d11effect there are two lib-files:
D3DX11Effects_vc9_x64.lib
D3DX11Effects_vc9_x64D.lib

lib "D3D11Effects_vc9.lib" is not there.
This is a bug in the sample project of  CUDASDK 64Bit package.

Solution:
we goto linker-->Additional Dependencies take off the entry D3D11Effects_vc9.lib and add
the two libs D3DX11Effects_vc9_x64.lib D3DX11Effects_vc9_x64D.lib

Then the project simpleD3D11Texture compiles successful. (and exe is executable also)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3>simpleD3D11Texture - 0 Fehler, 6 Warnung(en)

All 93 samples are successful compiled.
We can assume installation and all necessary setting are correct. 

[_heinz]

Speedup test (ap cpu-version)
IPP SSE2 without BLANKIT compiled on V8-VM3 shows following results on my 3 machines

WU's   V8-VM3   P4,2.6   R3600(1.6)
----------------------------------------------------------------
1LC25   45,48   25,60   33,72
    1,83    1,34    1,51
----------------------------------------------------------------
Rtiny   43,95   18,25   31,27
    1,78    1,22    1,45
----------------------------------------------------------------
sigi_v5   50,26   24,33   35,11
    2,01    1,32    1,54
----------------------------------------------------------------

A special ATOM build is in our public beta test area for download.
http://lunatics.kwsn.net/18-astropulse-testing/beta-testing-astropulse-on-atom-cpu.0.html

heinz

[_heinz]

ap (cpu-version)
with the data from Claggy's ATOM N450 (from our beta test) we can confirm the speedup
Quick timetable
 
WU : ap_18se08aa_B6_P1_00046_1LC25.wu
ap_5.05r168_SSE3.exe :
  Elapsed 2203.207 secs
      CPU 2192.469 secs
ap_5.05r409_SSE.exe :
  Elapsed 1995.041 secs, speedup: 9.45%  ratio: 1.10
      CPU 1985.284 secs, speedup: 9.45%  ratio: 1.10
ap_5.05r460_SSE3_ATOM_ICC_MKLS_O3.exe :
  Elapsed 1429.883 secs, speedup: 35.10%  ratio: 1.54
      CPU 1424.461 secs, speedup: 35.03%  ratio: 1.54
ap_5.05r468_SSE3_ATOM_IXE_MKLS_O3.exe :
  Elapsed 1443.502 secs, speedup: 34.48%  ratio: 1.53
      CPU 1434.445 secs, speedup: 34.57%  ratio: 1.53
 
WU : JasonShort_v5.wu
ap_5.05r168_SSE3.exe :
  Elapsed 3439.962 secs
      CPU 3425.423 secs
ap_5.05r409_SSE.exe :
  Elapsed 3251.639 secs, speedup: 5.47%  ratio: 1.06
      CPU 3236.646 secs, speedup: 5.51%  ratio: 1.06
ap_5.05r460_SSE3_ATOM_ICC_MKLS_O3.exe :
  Elapsed 2353.124 secs, speedup: 31.59%  ratio: 1.46
      CPU 2342.168 secs, speedup: 31.62%  ratio: 1.46
ap_5.05r468_SSE3_ATOM_IXE_MKLS_O3.exe :
  Elapsed 2348.225 secs, speedup: 31.74%  ratio: 1.46
      CPU 2336.895 secs, speedup: 31.78%  ratio: 1.47
 
WU : Raistmer_tinyrr.wu
ap_5.05r168_SSE3.exe :
  Elapsed 877.392 secs
      CPU 870.907 secs
ap_5.05r409_SSE.exe :
  Elapsed 877.424 secs, speedup: -0.00%  ratio: 1.00
      CPU 868.130 secs, speedup: 0.32%  ratio: 1.00
ap_5.05r460_SSE3_ATOM_ICC_MKLS_O3.exe :
  Elapsed 609.477 secs, speedup: 30.54%  ratio: 1.44
      CPU 604.395 secs, speedup: 30.60%  ratio: 1.44
ap_5.05r468_SSE3_ATOM_IXE_MKLS_O3.exe :
  Elapsed 602.364 secs, speedup: 31.35%  ratio: 1.46
      CPU 593.069 secs, speedup: 31.90%  ratio: 1.47
 
WU : short_ap_21oc08ab_B2_P0_00081_20081130_08605_v5.wu
ap_5.05r168_SSE3.exe :
  Elapsed 1819.665 secs
      CPU 1809.362 secs
ap_5.05r409_SSE.exe :
  Elapsed 1724.053 secs, speedup: 5.25%  ratio: 1.06
      CPU 1716.885 secs, speedup: 5.11%  ratio: 1.05
ap_5.05r460_SSE3_ATOM_ICC_MKLS_O3.exe :
  Elapsed 1232.933 secs, speedup: 32.24%  ratio: 1.48
      CPU 1226.059 secs, speedup: 32.24%  ratio: 1.48
ap_5.05r468_SSE3_ATOM_IXE_MKLS_O3.exe :
  Elapsed 1223.495 secs, speedup: 32.76%  ratio: 1.49
      CPU 1217.635 secs, speedup: 32.70%  ratio: 1.49
 
WU : sigind_v5.wu
ap_5.05r168_SSE3.exe :
  Elapsed 4775.746 secs
      CPU 4747.922 secs
ap_5.05r409_SSE.exe :
  Elapsed 4278.386 secs, speedup: 10.41%  ratio: 1.12
      CPU 4241.121 secs, speedup: 10.67%  ratio: 1.12
ap_5.05r460_SSE3_ATOM_ICC_MKLS_O3.exe :
  Elapsed 3097.042 secs, speedup: 35.15%  ratio: 1.54
      CPU 3084.374 secs, speedup: 35.04%  ratio: 1.54
ap_5.05r468_SSE3_ATOM_IXE_MKLS_O3.exe :
  Elapsed 3108.524 secs, speedup: 34.91%  ratio: 1.54
      CPU 3090.879 secs, speedup: 34.90%  ratio: 1.54
------------------------------------------------------------------------------
ap cpu-version for ATOM processor will be published together with next unified installer version
heinz

[Raistmer]

I'd say ICC and IXE builds have same speed in error range.

[_heinz]

Milestones

today 25 Mio total

tomorrow 1 Mio Docking  (all cpu work)

most done by V8-Xeon

happy crunching 

[_heinz]

Some construction ideas for your next monster cruncher.
but then liquid cooled.
CUDA@MIT
Found on MIT

have fun

heinz 

[Mod:] corrected unfortunate typo, that might have resulted in too many google search hits

[_heinz]

Hi Jason,
do you know Australias greenest Supercomputer   

[_heinz]

Surprize,
Today I'm one of the "Top Contributors" in Intels Software Network forums

 

[_heinz]

I had have some trouble with latest CUDA3.2 package, it has some issues using different compilers, some error reports are still open. (Sometime, in different projects, ptxas hung up)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
the 2 GTX470 runs fine under full load.
gtx470_geraeteeigenschaften
gtx470_auslastung
gtx470_gpuz
gpuload
cpu runs on all 8 cores docking, no problem to feed 2 GPU's
cpu_auslastung
temps

heinz
modify: some links added

[_heinz]

2 x GTX470 Colorful_stable light oc'ed, prepared for some good numbers.
GPU_load_stable
To confirm the values I run two GTX470 3 days under full load. (took part at primegrids "Winter Solstice Challenge" place 71 )
Awesome!

Sure there are some other cards with some higher core clocks...but Colorful surprized me.

heinz

[_heinz]

Sylvester 2010,
we are not far away from the magic click of 100 000

Time to say "thank you" to all readers of this epic thread
Kindest regards 

Happy New Year 2011

heinz

[_heinz]

Awaiting the 50 Mio credit today
...waiting
Total  50,689,071.39
got it