AK V8 + CUDA MB team work mod

[k6xt]

Raistmer says:
No-no-no, for AMD will have_AMD in name. V8 and V8a differs in will it kill VLAR or will it process VLAR. Due current Berkeley's network troubles I would recommend V8.
(and AMD can't support SSSE3 btw, only SSE3)
V8 it is, thank you. Especially since I don't know what VLAR is. :-)

[k6xt]

It appears I have things running but can't tell if the GPU is processing. Task Manager shows five instances of AK_v8b_win_SSSE3x_GPU_CPU_team, and one instance of
MB_6.08_mod_CPU_team_CUDA. All 4 processors are at 100%

I have ncpu=5 in cc_config which shows up in BOINC messages and also Boinc shows 5 running tasks.

Have I got it right?

Thanks
Art

[Raistmer]

V8 it is, thank you. Especially since I don't know what VLAR is. :-)

Sorry, problems with acronims was pointed out on SETI main already, maybe it's worth to have some dictionary thread indeed. VLAR=VeryLowAngleRange - in short, very small value of one of task parameters. It depends from how big angle motion of telescope was at recording this particular WU and will be different for different WUs.

And about your installation, if MB_6.08_mod_CPU_team_CUDA takes few % of CPU time to time, 4 copies of AK_v8b_win_SSSE3x_GPU_CPU_team take ~25% of CPU each and 5th copy of AK_v8b_win_SSSE3x_GPU_CPU_team takes no CPU - yes, it works as should. You can check it via  Windows Task Manager (Ctr-Shift-Esc or Ctr-Alt-Del )

[k6xt]

My WinTask Manager looks like you say it should. MB_*_CUDA.exe takes 1% occasionally, there are four AK tasks taking 25% each.

But I have two AK tasks running at zero % not one. I tried "end process" but #6 keeps coming back. Well I guess all is OK. Boinc tasks shows 5 tasks running.

The Q6600 is overclocked at 3.1GHz. I can squeeze 3.2GHz out of it but around that speed Prime95 will occasionally stop on an error. Not worth the trouble or lost WU I think. I set CPU fan target in BIOS to 60 deg C which the system maintains easily.

Now to see if host average rises!

Thanks
Art

[Grey Shadow]

Hi Raistmer,

Would you mind updating your mod to switch between CPU/GPU for VLAR units (as you mentioned on the first page - if task is VLAR - stay with CPU, else - try pass task to GPU)? VLAR Autokill works perfectly, at last I have stable SETI setup using GPU and not requiring constant attention, but sometimes it is very pity to see that 20 WUs downloaded between constant "out of work" messages go to trash bin...

Thanks in advance

[Raistmer]

Hi Raistmer,

Would you mind updating your mod to switch between CPU/GPU for VLAR units (as you mentioned on the first page - if task is VLAR - stay with CPU, else - try pass task to GPU)? VLAR Autokill works perfectly, at last I have stable SETI setup using GPU and not requiring constant attention, but sometimes it is very pity to see that 20 WUs downloaded between constant "out of work" messages go to trash bin...

Thanks in advance

Well, still can't figure out how accomplish that w/o going to idle GPU. as you said VLARs will come in bunches (20 tasks in your example) if GPU will reject to do them passing them to CPU - well, GPU will be just idle. No gains. So, much better to use V8 (not V8a) for now - as I remember it should process VLARs on GPU instead of abort it. BTW, CUDA app from all V8 packages are interchengeable so you could just overwrite single EXE (with stopped BOINC of course) to switch between VLAR kill and VLAR process.

[Grey Shadow]

I understand your concern, but GPU-crunching of VLAR workunit requires at least the same time as CPU-crunching. So there will be almost no gain from allowing GPU-crunching of VLARs. Also when I was using stock apps or V8 I was very worried by PC overload during VLAR crunching by GPU - PC starts working very-very slow and unstable, it is almost impossible to use it for any other purposes during that time... really I think that CPU efficace in such conditions also decrease so we loose any gain from GPU involvement.

Of course, the best way is to postpone VLAR workunit if GPU is idle and go to next available workunit, but as I understand this is BOINC task so optimized apps can't achieve this without client support.

[Raistmer]

I understand your concern, but GPU-crunching of VLAR workunit requires at least the same time as CPU-crunching. So there will be almost no gain from allowing GPU-crunching of VLARs.

Well, you recive one additional core instead 2 or even 3 - it worse of course but not bad too

Also when I was using stock apps or V8 I was very worried by PC overload during VLAR crunching by GPU - PC starts working very-very slow and unstable, it is almost impossible to use it for any other purposes during that time...

I see, but because experiment with tuning AK_v8 on different ARs failed (there was no speed increase if I did PGO only on VLAR dyn data set instead of complete set) adding AR recognition now on low priority really.

really I think that CPU efficace in such conditions also decrease so we loose any gain from GPU involvement.

Not so sure. Don't forget, sluggish user interface - one thing and CPU app performance - completely another...

Of course, the best way is to postpone VLAR workunit if GPU is idle and go to next available workunit, but as I understand this is BOINC task so optimized apps can't achieve this without client support.

Yes, interaction with BOINC API needed...

[cyclejon]

As you can see SSE2 build performs better than SSE3 one on SSE3-capable (early) AMD. Maybe on latest Phenom SSE3-situation was improved ?
Could someone of our pre-testers or regular users try to run KWSN bench for AK_v8b_SSE3 and AK_v8b_SSE2 on new Phenom CPU to shed light on current situation with AMD SSE3 support quality ?

Ran it on a phenom 9500 with the v8b apps and again with the regular v8 op apps. SSE2 was faster in the v8b apps, but the v8 apps were mixed. My athlon x2 4200 showed the same results except when put against x64 builds. SSE3 x64 ran faster than SSE2 x86. I'll attach the benchmark files from the phenom.

[attachment deleted by admin]

[Raistmer]

WU : testWU-1.wu
AK_v8b_win_SSE3_GPU_CPU_team.exe : 199 seconds
AK_v8b_win_SSE2_GPU_CPU_team.exe : 6 seconds
Speedup: 96.98%, Ratio: 33.17 x
 
Strange result.
How did you prevent call to CUDA app?

[Jason G]

Restarted at 100.00 percent.

 

[cyclejon]

The computer that ran on doesn't have a CUDA card in it. I don't know what happened to that wu, why it ended so soon.

[Raistmer]

The computer that ran on doesn't have a CUDA card in it. I don't know what happened to that wu, why it ended so soon.

Hm... interesting what will be with team app in that case...
Because bot tested apps is not AMD specific ones, better to re-test their analogues w/o team mod. That is, "standart" AK_v8b_SSE3/2 versions. Theys should give same speeds and much better support standalone runs than "teamed" ones.

[cyclejon]

Hm... interesting what will be with team app in that case...
Because bot tested apps is not AMD specific ones, better to re-test their analogues w/o team mod. That is, "standart" AK_v8b_SSE3/2 versions. Theys should give same speeds and much better support standalone runs than "teamed" ones.

I thought both of those apps were from your AMD packages.
The second test I posted is with the only other v8 apps i could find. I hadn't seen ak_v8b without gpu support. I don't have direct access to either of the phenom machines on my account, so i don't know when i'll be able to retest.

[Raistmer]

Hm... interesting what will be with team app in that case...
Because bot tested apps is not AMD specific ones, better to re-test their analogues w/o team mod. That is, "standart" AK_v8b_SSE3/2 versions. Theys should give same speeds and much better support standalone runs than "teamed" ones.

I thought both of those apps were from your AMD packages.
The second test I posted is with the only other v8 apps i could find. I hadn't seen ak_v8b without gpu support. I don't have direct access to either of the phenom machines on my account, so i don't know when i'll be able to retest.

SSE2 one was in AMD-specific package indeed, now that version obsolete. Try to download new one (first page). There should be _AMD in app name.
AK_v8 available in "Download" section of this site.