1)
Message boards :
Number crunching :
World Animal Day Challenge
(Message 69578)
Posted 3641 days ago by  VictordeHollander
Shouldn't the server be loaded with much more TRP LLR work?
As of this moment I can see 187/600 WUs, not that much for a challenge?
|
2)
Message boards :
Generalized Fermat Prime Search :
Source code of Genefer for OpenCL is available.
(Message 68610)
Posted 3686 days ago by VictordeHollander
Windows 7 64bit
2500k
HD7950
3.1.2-5
Command line: geneferocl-windows3.1.2-5.exe -b
Running on platform 'AMD Accelerated Parallel Processing', device 'Tahiti', vers
ion 'OpenCL 1.2 AMD-APP (1124.2)' and driver '1124.2 (VM)'.
Generalized Fermat Number Bench
2199064^8192+1 Time: 78.1 us/mul. Err: 0.2344 51956 digits
1798620^16384+1 Time: 77.5 us/mul. Err: 0.2266 102481 digits
1471094^32768+1 Time: 79.3 us/mul. Err: 0.2344 202102 digits
1203210^65536+1 Time: 103 us/mul. Err: 0.2656 398482 digits
984108^131072+1 Time: 146 us/mul. Err: 0.2188 785521 digits
804904^262144+1 Time: 342 us/mul. Err: 0.2188 1548156 digits
658332^524288+1 Time: 645 us/mul. Err: 0.2188 3050541 digits
538452^1048576+1 Time: 1.17 ms/mul. Err: 0.2266 6009544 digits
440400^2097152+1 Time: 2.42 ms/mul. Err: 0.2266 11836006 digits
360204^4194304+1 Time: 4.84 ms/mul. Err: 0.2031 23305854 digits
294612^8388608+1 Time: 9.69 ms/mul. Err: 0.1895 45879398 digits
Haven't tried any real task with this version yet.
|
3)
Message boards :
Generalized Fermat Prime Search :
Source code of Genefer for OpenCL is available.
(Message 68566)
Posted 3687 days ago by VictordeHollander
So far my 7950 completed 6 'short' Genefer OpenCL tasks without any errors. All are validated by wingmens using CUDA Genefer versions.
GPU: ASUS HD7950 DC2T (Factory OC: 900MHz)
OS: Windows 7 64bit
AMD Catalyst: 13.4
Genefer OpenCL version 3.1.2-2
CPU load: <1%
GPU load: 97-99%
Runtimes: 26,516 - 28,080 sec
For the record:
http://www.primegrid.com/workunit.php?wuid=339797861
http://www.primegrid.com/workunit.php?wuid=342700453
http://www.primegrid.com/workunit.php?wuid=343200843
http://www.primegrid.com/workunit.php?wuid=345650390
http://www.primegrid.com/workunit.php?wuid=345651275
http://www.primegrid.com/workunit.php?wuid=345653871
Some benchmarks:
3.1.2-2
Command line: geneferocl-windows.exe -b
Running on platform 'AMD Accelerated Parallel Processing', device 'Tahiti', vers
ion 'OpenCL 1.2 AMD-APP (1124.2)' and driver '1124.2 (VM)'.
Generalized Fermat Number Bench
2199064^8192+1 Time: 83.3 us/mul. Err: 0.2344 51956 digits
1798620^16384+1 Time: 113 us/mul. Err: 0.2266 102481 digits
1471094^32768+1 Time: 122 us/mul. Err: 0.2344 202102 digits
1203210^65536+1 Time: 127 us/mul. Err: 0.2656 398482 digits
984108^131072+1 Time: 155 us/mul. Err: 0.2188 785521 digits
804904^262144+1 Time: 364 us/mul. Err: 0.2188 1548156 digits
658332^524288+1 Time: 768 us/mul. Err: 0.2188 3050541 digits
538452^1048576+1 Time: 1.48 ms/mul. Err: 0.2266 6009544 digits
440400^2097152+1 Time: 2.81 ms/mul. Err: 0.2266 11836006 digits
360204^4194304+1 Time: 5.55 ms/mul. Err: 0.2031 23305854 digits
294612^8388608+1 Time: 11 ms/mul. Err: 0.1895 45879398 digits
3.1.2-4
Command line: geneferocl-windows3.1.2-4.exe -b
Running on platform 'AMD Accelerated Parallel Processing', device 'Tahiti', vers
ion 'OpenCL 1.2 AMD-APP (1124.2)' and driver '1124.2 (VM)'.
Generalized Fermat Number Bench
2199064^8192+1 Time: 77.8 us/mul. Err: 0.2188 51956 digits
1798620^16384+1 Time: 78.1 us/mul. Err: 0.2344 102481 digits
1471094^32768+1 Time: 80.2 us/mul. Err: 0.2344 202102 digits
1203210^65536+1 Time: 103 us/mul. Err: 0.2813 398482 digits
984108^131072+1 Time: 148 us/mul. Err: 0.2295 785521 digits
804904^262144+1 Time: 354 us/mul. Err: 0.2188 1548156 digits
658332^524288+1 Time: 635 us/mul. Err: 0.2266 3050541 digits
538452^1048576+1 Time: 1.17 ms/mul. Err: 0.2188 6009544 digits
440400^2097152+1 Time: 2.37 ms/mul. Err: 0.2188 11836006 digits
360204^4194304+1 Time: 4.84 ms/mul. Err: 0.1953 23305854 digits
294612^8388608+1 Time: 9.78 ms/mul. Err: 0.2070 45879398 digits
Edit:
So 3.1.2-4 is definitely faster
|
4)
Message boards :
Generalized Fermat Prime Search :
Source code of Genefer for OpenCL is available.
(Message 68489)
Posted 3688 days ago by VictordeHollander
My first WU with Genefer OpenCL on my 7950 is complete and validated:
http://www.primegrid.com/workunit.php?wuid=343200843
It took 27,721 sec.
|
5)
Message boards :
Generalized Fermat Prime Search :
Source code of Genefer for OpenCL is available.
(Message 68469)
Posted 3689 days ago by VictordeHollander
For reference, it seems that OpenCL version is faster than CUDA on GTX Titan.
CUDA:
538452^1048576+1 Time: 1.75 ms/mul. Err: 0.2031 6009544 digits
OpenCL:
538452^1048576+1 Time: 859 us/mul. Err: 0.2266 6009544 digits
Wow, OpenCL is almost twice as fast on your Titan! That is totally unexpected.
|
6)
Message boards :
Generalized Fermat Prime Search :
Source code of Genefer for OpenCL is available.
(Message 68465)
Posted 3689 days ago by VictordeHollander
Testing on a HD7950 with 1 CPU core reserved, estimated time for a WU ~28,000 sec.
|
7)
Message boards :
Number crunching :
Low Country Festivities Challenge
(Message 65366)
Posted 3783 days ago by VictordeHollander
Are you sure these figures are correct?
That fits to about the average run time per unit, so on the 7th-9th a bunch were sent back into the wild and they finished. I bet the numbers are just fine.
I meant the 980 results returned for the team positions, while only 392 results have been returned in total?
|
8)
Message boards :
Number crunching :
Low Country Festivities Challenge
(Message 65364)
Posted 3783 days ago by VictordeHollander
CLEANUP UPDATE
May 13th: 4039 tasks outstanding; 3048 in individual (199) scoring positions; 2909 in team (94) scoring positions.
May 14th: 3647 tasks outstanding; 2747 in individual (199) scoring positions; 1929 in team (87) scoring positions.
Are you sure these figures are correct?
From 13 to 14th of May: -392 tasks outstanding; -301 tasks in individual; -980 tasks in team??
|
9)
Message boards :
Number crunching :
New badge proposal - discuss!
(Message 64920)
Posted 3793 days ago by VictordeHollander
I like the first 'new' badge system with multiples of 10x and the distinctive colours. I see no reason to add 10+ badges, as only a handful of people will accomplish them, but who am I with only 4 ruby badges.
edit:
Just keep it simple like this:
10M
100M
1.000M = 1B
10.000M =10B
100.000M = 100B
|
10)
Message boards :
Number crunching :
Low Country Festivities Challenge
(Message 64695)
Posted 3798 days ago by VictordeHollander
Although I'm not really fond of royalty, it's always nice to have a day off and party.
So, happy Queens night/day from the Netherlands!
|
