The AP 7 apps are Lunatics technology.
Brilliant work by our group i might add.

So does that mean the credit granted for AP will drop like a stone as happened with MBv7 thanks to Credit New?

I am seeing an average of 519.43 credits for AP on Beta, but they are running faster. So it might work out to be the same average per day. I have not done that calculation as of yet.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

The AP 7 apps are Lunatics technology.
Brilliant work by our group i might add.

So does that mean the credit granted for AP will drop like a stone as happened with MBv7 thanks to Credit New?

I am seeing an average of 519.43 credits for AP on Beta, but they are running faster. So it might work out to be the same average per day. I have not done that calculation as of yet.

When Lunatics code was used for MBv7, the end result was a halving of RAC for those that had been using optimised applications previously.
If Credit New continues to work the same way, that's what will happen to those who have been using optimised AP applications.

Of course they could fix Credit New, and pay MB crunchers appropriately, then current AP optimised crunchers won't suffer deflation in their credit.

---

Grant
Darwin NT

The AP 7 apps are Lunatics technology.
Brilliant work by our group i might add.

So does that mean the credit granted for AP will drop like a stone as happened with MBv7 thanks to Credit New?

I am seeing an average of 519.43 credits for AP on Beta, but they are running faster. So it might work out to be the same average per day. I have not done that calculation as of yet.

When Lunatics code was used for MBv7, the end result was a halving of RAC for those that had been using optimised applications previously.
If Credit New continues to work the same way, that's what will happen to those who have been using optimised AP applications.

Of course they could fix Credit New, and pay MB crunchers appropriately, then current AP optimised crunchers won't suffer deflation in their credit.

Or they could just let AP credit fall to the same ratio of payout that used to be before the roll out on MBV7. Used to be nobody wanted AP work untill the new MB version payed less. Then everyone started likeing AP work. I have a feeling it will be back to par for both types of work.

---

[/quote]

Old James

Or they could just let AP credit fall to the same ratio of payout that used to be before the roll out on MBV7. Used to be nobody wanted AP work untill the new MB version payed less. Then everyone started likeing AP work. I have a feeling it will be back to par for both types of work.

We can hope so anyway. ;-)

Cheers.

Considering at least my AMD R9 290 cards have surprisingly similar runtimes for MB tasks as they do for AP, the credits resulting from that are simply *mind blown*

AMD R9 290
~ 21min for 2 parallel average MB tasks, roughly 6 per hour
~ 31min for 2 parallel average AP tasks, roughly 4 per hour

That's about a 1.5:1 runtime ratio for V6 AP tasks vs. V7 MB tasks under OpenCL (Lun 0.42).
Of course, the credit ratio per hour of work for that is like ~2800Cr AP vs. 510Cr MB or a 5.5 : 1 .

1.5:1 for a 5.5:1 crediting I'll take anyday of course ;)
Even with pretty well-blanked tasks (to a limit), the ratio is still clearly in favor of AP) with the current status quo.

I've generally observed that the NVidia cards comparably scream through Cuda MB tasks like hell while taking their time on AP (although still good) - while AMD cards eat AP tasks for breakfast but really take their time chugging through MB tasks.
The differences seem to decrease only when increasing the number of parallel tasks further towards its optimum.
( I got GTX 780 vs. R9 290 for that comparison, as well as GTX 750Ti vs. HD7850 , and that's what I experienced)

I didn't know that before, hence I was surprised to see things moving that way.

---

I've generally observed that the NVidia cards comparably scream through Cuda MB tasks like hell while taking their time on AP (although still good) - while AMD cards eat AP tasks for breakfast but really take their time chugging through MB tasks.

That's what Raistmer say from a long time:
- the most effective use of GPUs here (not 'by credit' but 'by performance'/'by work done') is to:
use NVIDIA for CUDA MultiBeam
use ATI AMD for OpenCL AstroPulse

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

I've generally observed that the NVidia cards comparably scream through Cuda MB tasks like hell while taking their time on AP (although still good) - while AMD cards eat AP tasks for breakfast but really take their time chugging through MB tasks.

That's what Raistmer say from a long time:
- the most effective use of GPUs here (not 'by credit' but 'by performance'/'by work done') is to:
use NVIDIA for CUDA MultiBeam
use ATI AMD for OpenCL AstroPulse

So is the "problem" that there's no CUDA AP app?

I've generally observed that the NVidia cards comparably scream through Cuda MB tasks like hell while taking their time on AP (although still good) - while AMD cards eat AP tasks for breakfast but really take their time chugging through MB tasks.

That's what Raistmer say from a long time:
- the most effective use of GPUs here (not 'by credit' but 'by performance'/'by work done') is to:
use NVIDIA for CUDA MultiBeam
use ATI AMD for OpenCL AstroPulse

So is the "problem" that there's no CUDA AP app?

Please see AP6_win_x86_SSE2_OpenCL_NV_r2399.exe from the v0.42 installer. While not CUDA it is the app that will give you the best use of NV cards for AP.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

I've generally observed that the NVidia cards comparably scream through Cuda MB tasks like hell while taking their time on AP (although still good) - while AMD cards eat AP tasks for breakfast but really take their time chugging through MB tasks.

That's what Raistmer say from a long time:
- the most effective use of GPUs here (not 'by credit' but 'by performance'/'by work done') is to:
use NVIDIA for CUDA MultiBeam
use ATI AMD for OpenCL AstroPulse

So is the "problem" that there's no CUDA AP app?

If there is a 'problem' as such, and I'm not certain there is because I'm not currently focussed on the AP side of things, it'll be because Cuda Cores are not like whatever AMD has, so require different programming approaches. Direct/straight translation of generic OpenCL to Cuda would likely yield poor results (and vice versa).

For Multibeam, I estimate that single instance currently runs, on modern Cuda cards, at about 5% of peak theoretical efficiency, edging up to 7-8% or so with 2 or 3 instances. Looking at the fpops estimates AP would be significantly lower efficiency than that at the moment.

I'm not sure the work supply or infrastructure is 'ready' for an all-tech thrown in Cuda variant based on everything that has been learned so far on multibeam, But moving toward adding AP support within x42, along with improved existing multibeam, is planned.

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

So is the "problem" that there's no CUDA AP app?

Please see AP6_win_x86_SSE2_OpenCL_NV_r2399.exe from the v0.42 installer. While not CUDA it is the app that will give you the best use of NV cards for AP.

Yes I am using that, just wondered if there was an CUDA version if that would be better.

So is the "problem" that there's no CUDA AP app?

Please see AP6_win_x86_SSE2_OpenCL_NV_r2399.exe from the v0.42 installer. While not CUDA it is the app that will give you the best use of NV cards for AP.

Yes I am using that, just wondered if there was an CUDA version if that would be better.

Perhaps. Lot of time both SETI apps do FFT. And proprietary cuFFT was significally better than open-source oclFFT on nVidia hardware when I tested them year or more ago.
I'm working to make oclFFT more flexible and tunable now, but expect cuFFT being faster still.
So, in my estimations, even direct porting of AstroPulse OpenCL code to CUDA would improve performance on nVidia cards. This porting requires time though.
Any volunteers?

@those who reported issues in OpenCL AstroPulse v6 from this Lunatic installer release:

Please test OpenCL APv7.03 on beta site and check if it works OK with your setups.
Doing so allow to release AP v7 that will work "out of box" on your own hardware, w/o additional efforts.

So is the "problem" that there's no CUDA AP app?

Please see AP6_win_x86_SSE2_OpenCL_NV_r2399.exe from the v0.42 installer. While not CUDA it is the app that will give you the best use of NV cards for AP.

Yes I am using that, just wondered if there was an CUDA version if that would be better.

If there were, it would be.

OpenCL, as the name implies, is (was?) an open standard for video hardware to do generalised computing work.
However many years ago the group responsible for OpenCL could best be described as dysfunctional.
Since OpenCL was such as mess, Nvidia decided to develop their own proprietary GPU programming language- CUDA. Once ATi (now AMD) saw that there was a big future in GPUs doing non-video work they decided to the same thing. By that time the OpenCL group had gotten their act together & instead of developing their own programming language as Nvidia did they decided to go with OpenCL.

AMD don't want to develop CUDA applications because that would legitimise CUDA & support Nvidia. Nvidia don't want to do OpenCL applications because that would support AMD. Since Intel have also chosen OpenCL for their iGPU programming that has given the standard a boost.
End result is that Nvidia does support OpenCL, but it could best be described as a half arsed attempt at best. CUDA is where they're putting their major efforts. Hence AP on Nvidia is slow, mostly due to the fact it is OpendCL based & Nvidia's support for OpenCL is rather poor.

---

Grant
Darwin NT

Quick question: is there a difference in AP CPU apps between win32 and win64 installers? I have a mix of both operating systems and there were no problems installing AVX app on 64bit Win 7, but when I tried to run win32 installer on my 32bit Win 7 it gave me SSE3 app and instead of AVX there is SSE3(AMD) sitting on it's spot. BTW, rigs I'm talking about are identical, except for the 32 vs 64 win7 type.

---

Who the hell is General Failure and why is he reading my harddisk?¿

Quick question: is there a difference in AP CPU apps between win32 and win64 installers? I have a mix of both operating systems and there were no problems installing AVX app on 64bit Win 7, but when I tried to run win32 installer on my 32bit Win 7 it gave me SSE3 app and instead of AVX there is SSE3(AMD) sitting on it's spot. BTW, rigs I'm talking about are identical, except for the 32 vs 64 win7 type.

Some situation for me, miss the AVX for 32bit. I installed the app manuel, but it isn't there.

You can check by yourself:
Open the Lunatics_Win??_v0.42_setup.exe files with 7-Zip and see what files are inside:
http://www.7-zip.org/

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Note in particular that the only available Astropulse AVX build is itself a 64-bit executable, and depends on the 64-bit version of the FFTW library DLL: AP6_win_x64_AVX_CPU_r2163.exe

I wasn't supplied with a 32-bit AVX AP executable to install - that's why there isn't one in the 32-bit installer. If a significant number of people are running the 32-bit version of Windows 7 SP1 on AVX hardware (minimum OS level for AVX to be used), then I can go back to the developer and ask for one. But it does seem an unbalanced OS/hardware combination.

I finally installed 0.42 on my MB/GPU-only box. Hard to say if it's running any faster or better than before, but I don't see any problems with it.

---

David
Sitting on my butt while others boldly go,
Waiting for a message from a small furry creature from Alpha Centauri.

I finally installed 0.42 on my MB/GPU-only box. Hard to say if it's running any faster or better than before, but I don't see any problems with it.

The Nvidia MB app, Lunatics_x41zc_win32_cudaNN.exe, is the same as in v0.41 installer. So I would expect the performance will be exactly the same as before for that machine. Unless you switched to a differing CUDA version app as there are 4 to choose.
Lunatics_x41zc_win32_cuda23.exe
Lunatics_x41zc_win32_cuda32.exe
Lunatics_x41zc_win32_cuda42.exe
Lunatics_x41zc_win32_cuda50.exe

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

Noticed I had an invalid CPU task 1567544761 on my host 7324426 with the new AP app . I saved the full Stderr output from all 3 results for when this vanishes, but can't figure out the fanout for AP tasks to save the work itself. Looks like the way to calculate that is different from MB?

Looks like my system found a repetitive pulse but the NV & ATI GPUs did not. It could just be a fluke or such. Just the first invalid CPU task I have seen on a long while and figured it might be noteworthy.

Here is what I think is pertinent & links to the output I saved for analysis.

Task ID: 3681602891
AstroPulse v6 Anonymous platform (NVIDIA GPU)
AstroPulse v6 Windows x86 rev 2399

    single pulses: 2
repetitive pulses: 0
  percent blanked: 3.18
Single pulse: peak_power=38.13 dm=5946 fft_num=29147136 peak_bin=29154000 scale=2
Single pulse: peak_power=130 dm=9335 fft_num=17809408 peak_bin=17820800 scale=6

Task ID: 3681602892
AstroPulse v6 Anonymous platform (CPU)
AstroPulse v6 Windows x64 rev 2163

    single pulses: 2
repetitive pulses: 1
  percent blanked: 3.18
Single pulse: peak_power=38.13 dm=5946 fft_num=29147136 peak_bin=29154000 scale=2
Single pulse: peak_power=130 dm=9335 fft_num=17809408 peak_bin=17820800 scale=6
Rep. pulse: peak_power=9469 dm=-9776 peak_bin=1280 scale=4 ffa_scale=1 period=236.14

Task ID: 3719572871
AstroPulse v6 v6.06 (opencl_ati_100)
Windows x86 rev 1832

    single pulses: 2
repetitive pulses: 0
  percent blanked: 3.18
Single pulse: peak_power=38.13; dm=5946; scale=2
Single pulse: peak_power=130; dm=9335; scale=6

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[