Legends never die: GPUPI - Seite 9

Seite 9 von 59 - Forum: Number Crunching auf overclockers.at

URL: https://www.overclockers.at/number-crunching/legends-never-die-gpupi_240993/page_9 - zur Vollversion wechseln!

Zitat von der~erl

Ja, genau das mein ich mat.

aber ich habs jetzt mal durchlaufen lassen, hat ganz schön lang gedauert.
R9 290:

Code:

Selected device: Hawaii
OpenCL 1.2 AMD-APP (1573.4) is ready.

Compiling OpenCL kernels ... done.

Calculating 1.000.000.000nth digit of PI. 20 iterations.

Allocated device memory : 671088640 Bytes
Batch Size              : 20M
Reduction Size          : 64

 00h 00m 02.078s Batch  1 finished.
 00h 00m 21.433s Batch  2 finished.
 00h 00m 40.805s Batch  3 finished.
 00h 01m 10.053s Batch  4 finished.
 00h 02m 00.131s Batch  5 finished.
 00h 02m 52.758s Batch  6 finished.
 00h 03m 12.478s Batch  7 finished.
 00h 03m 31.514s Batch  8 finished.
 00h 03m 58.588s Batch  9 finished.
 00h 04m 45.479s Batch 10 finished.
 00h 05m 35.563s Batch 11 finished.
 00h 05m 54.525s Batch 12 finished.
 00h 06m 13.608s Batch 13 finished.
 00h 06m 42.320s Batch 14 finished.
 00h 07m 32.383s Batch 15 finished.
 00h 08m 24.647s Batch 16 finished.
 00h 08m 44.152s Batch 17 finished.
 00h 09m 03.024s Batch 18 finished.
 00h 09m 30.641s Batch 19 finished.
 00h 10m 17.459s Batch 20 finished.
 00h 11m 06.049s PI value output -> 5895585A0

 GPU time for   pi calculation: 632.716 s
 GPU time for memory reduction:  33.333 s

mal ein erster durchlauf

R9 290:



PS: Sollte ich alles ausmachen was GPU Last erzeugt evtl. ? Kommr mir lahm vor iwie verglichen zu der~erl seiner R9 290.

Bitte unbedingt die Version 1.0.6 runterladen, bevor ihr AMD bencht! Die Versionen davor nehmen wegen einem Bug die CPU-Implementierung von AMD für OpenCL, nicht die GPU!

besorg dir mal die neue version, du rechnest gerade mit deiner cpu.


echte R9 290 Werte:

Code:

Selected device: Hawaii
OpenCL 1.2 AMD-APP (1573.4) is ready.

Compiling OpenCL kernels ... done.

Calculating 1.000.000.000nth digit of PI. 20 iterations.

Allocated device memory : 671088640 Bytes
Batch Size              : 20M
Reduction Size          : 64

 00h 00m 00.496s Batch  1 finished.
 00h 00m 04.913s Batch  2 finished.
 00h 00m 09.330s Batch  3 finished.
 00h 00m 13.748s Batch  4 finished.
 00h 00m 18.165s Batch  5 finished.
 00h 00m 22.581s Batch  6 finished.
 00h 00m 26.985s Batch  7 finished.
 00h 00m 31.394s Batch  8 finished.
 00h 00m 35.803s Batch  9 finished.
 00h 00m 40.222s Batch 10 finished.
 00h 00m 44.640s Batch 11 finished.
 00h 00m 49.056s Batch 12 finished.
 00h 00m 53.470s Batch 13 finished.
 00h 00m 57.887s Batch 14 finished.
 00h 01m 02.305s Batch 15 finished.
 00h 01m 06.720s Batch 16 finished.
 00h 01m 11.139s Batch 17 finished.
 00h 01m 15.554s Batch 18 finished.
 00h 01m 19.974s Batch 19 finished.
 00h 01m 24.394s Batch 20 finished.
 00h 01m 28.371s PI value output -> 5895585A0

 GPU time for   pi calculation:  74.772 s
 GPU time for memory reduction:  13.599 s

also das funktioniert jetzt schonmal richtig

Code:

Devices found that don't support double precision:
 => GPU: Cedar

Selected device: Tahiti
OpenCL 1.2 AMD-APP (1573.4) is ready.

Code:

 GPU time for   pi calculation:  93.485 s
 GPU time for memory reduction:  13.822 s

Also wenn 1.05 noch CPU war, dann gibt es mit 1.06 kaum einen Unterschied, wenn jetzt wirklich die GPU berechnet.

würd die cpu rechnen hätts 10x so lang gedauert

Hmm ...

Zitat von s!LeNt_tR!ggEr

Hmm ...

Bitte



Sorry nutze immer das Sys aus der Sig.

@lalaker: nö nö, nix da

Turbo aus (CPU), eh nicht relevant und Karte ist @ stock. Weiß aber die geht gut, kurz getestet Aber no need for oc ... noch nicht zumindest

Zitat von s!LeNt_tR!ggEr

Hmm ...

Zitat

ERROR: OpenCL error: CL_MEM_OBJECT_ALLOCATION_FAILURE happened in file pi.cpp at line 624.

Hab aus offensichtlichen Gründen abgebrochen Aber ja, rennt jetzt anscheinend fehlerfrei mit 1.0.6 (@sk/\r: Link in der News, nicht im Thread):

overclockers.at v4.thecommunity
© all rights reserved by overclockers.at 2000-2024