New updated CUDA version has been released.

2) Always cancel Astroid Cuda55 work units - you'll notice if one is computed by suddenly extreme slow mouse reactions. Tiresome, as you'll always have to press "cancel" if a Cuda55-WU starts.

The performance on a highend card is also a bit disappointing. My GTX Titan seems to need nearly one hour for a workunit.

There's some trick to get wus?

    <no_rsc_apps>NVIDIA</no_rsc_apps>

So 29 Dez 2013 21:04:01 CET | Asteroids@home | update requested by user
So 29 Dez 2013 21:04:05 CET | Asteroids@home | Sending scheduler request: Requested by user.
So 29 Dez 2013 21:04:05 CET | Asteroids@home | Requesting new tasks for NVIDIA
So 29 Dez 2013 21:04:07 CET | Asteroids@home | Scheduler request completed: got 2 new tasks
So 29 Dez 2013 21:04:09 CET | Asteroids@home | Started download of period_search_10000_x86_64-pc-linux-gnu__cuda55
So 29 Dez 2013 21:04:09 CET | Asteroids@home | Started download of libcudart.so.5.5
So 29 Dez 2013 21:04:11 CET | Asteroids@home | Finished download of libcudart.so.5.5
So 29 Dez 2013 21:04:11 CET | Asteroids@home | Started download of input_63267_6
So 29 Dez 2013 21:04:12 CET | Asteroids@home | Finished download of input_63267_6
So 29 Dez 2013 21:04:12 CET | Asteroids@home | Started download of input_63268_7
So 29 Dez 2013 21:04:13 CET | Asteroids@home | Finished download of input_63268_7
So 29 Dez 2013 21:04:14 CET | Asteroids@home | Finished download of period_search_10000_x86_64-pc-linux-gnu__cuda55
So 29 Dez 2013 21:04:14 CET | Asteroids@home | Starting task ps_131225_63267_6_0 using period_search version 10000 (cuda55) in slot 6

So 29 Dez 2013 20:20:00 CET |  | Starting BOINC client version 7.0.65 for x86_64-pc-linux-gnu
So 29 Dez 2013 20:20:00 CET |  | Processor: 8 GenuineIntel Intel(R) Core(TM) i7-3632QM CPU @ 2.20GHz [Family 6 Model 58 Stepping 9]
So 29 Dez 2013 20:20:00 CET |  | 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 ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx est tm2 ssse3 cx16 xtpr pdcm pcid sse4_1 sse4_2 x2apic popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm ida arat epb xsaveopt pln pts dtherm tpr_shadow vnmi flexpriority ept vpid fsgsbase smep erms
So 29 Dez 2013 20:20:00 CET |  | OS: Linux: 3.12.5-i7
So 29 Dez 2013 20:20:00 CET |  | Memory: 7.72 GB physical, 9.77 GB virtual
So 29 Dez 2013 20:20:00 CET |  | CUDA: NVIDIA GPU 0: GeForce GT 635M (driver version unknown, CUDA version 6.0, compute capability 2.1, 2048MB, 2032MB available, 182 GFLOPS peak)
So 29 Dez 2013 20:20:00 CET |  | OpenCL: NVIDIA GPU 0: GeForce GT 635M (driver version 331.20, device version OpenCL 1.1 CUDA, 2048MB, 2032MB available, 182 GFLOPS peak)
[...]
So 29 Dez 2013 20:30:04 CET | Asteroids@home | Resetting project
So 29 Dez 2013 20:30:17 CET | Asteroids@home | update requested by user
So 29 Dez 2013 20:30:20 CET | Asteroids@home | Master file download succeeded
So 29 Dez 2013 20:30:25 CET | Asteroids@home | Sending scheduler request: Requested by user.
So 29 Dez 2013 20:30:25 CET | Asteroids@home | Requesting new tasks for CPU
So 29 Dez 2013 20:30:32 CET | Asteroids@home | Scheduler request completed: got 26 new tasks
So 29 Dez 2013 20:30:34 CET | Asteroids@home | Started download of period_search_10210_x86_64-pc-linux-gnu__avx
[...]

@Cirrussc,

Did you edit your website preferences and select "Use GPU"?

So 29 Dez 2013 19:19:45 CET | Asteroids@home | Requesting new tasks for CPU

Don't fetch tasks for NVIDIA GPU: Project has no apps for NVIDIA GPU

My raspberry is now at 40% and 120 hours. The Work Unit will be finished in appoximate 180 hours, but it should be finished in 110 hour.
Such a misery - we will be too late.

Rooooobert

My raspi did longer task in 447,546.86 s. It's overclocked to 1GHz.

Thanks for pointing on 780ti dp speed. I thought it's like titan and it is not.

We need cuda 5.0 and better because of linker. I think cuda version do not affect speed on older cards. It's more related to code than to cuda version.

Ok. So let's be honest... If we're porting asteroids to gpu, the only way to go is opencl! That will of course take into account that all nvidia gpus except the titan suck at DP...

So no need to waste energy in porting the app to cuda when we already know that asteroids will be dominated by ATI/AMD gpus because those are less crippled regarding DP performance.

To face the fact, Asteroids will be a "second" Milkyway@home regarding GPU dominance.

AMD/ATI GPUs will dominate and nvidia will be left biting the dust.

Machine 'L' P4 XP32

v100.00 43956
v101.00 (sse2) 6054
v102.10 12853
v102.10 (sse2) 13650 SLOWER


Machine 'B' P4HT XP32

v100.00 48963
v101.00 (sse2) 9341
v102.10 18592
v102.10 (sse2) 18779 SLOWER


Basically, the new versions are at best no faster, and in 2 cases much slower.

I too noticed a big increase in estimated completion times shown by BOINC.
It seems that the reason is an increased <rsc_fpops_est> value in the init_data.xml file for the new WUs.

<rsc_fpops_est>315010000000000.000000</rsc_fpops_est>

That results in 315010 GFLOPS. AFAIK the old value was something like 10000(?) GFLOPS.

Kyong is testing SSE2 now. I'm working on standard app now (some backports from sse3 version) as preparation step for nVidia CUDA development.

Any idea why the linux64 app isn't 2x faster? There's an improvement, but not a 2 fold decrease in crunch time like the windows app.