I would have posted this in number crunching, but I don't have enough credit to post there.

I downloaded the optimized file from lunatics and installed. I have an AMD Phenom II x6 1075T processor that is 6 core and runs SSE, SSE2, SSE3 and SSE4A. No matter what I select in the installation screen, it always slows down my processing. Right now when I click properties on a WU I'm running my estimated app speed is 8.95 GFLOPS/S. With the optimized software, it's closer to 3 GFLOPs. Am I doing something wrong?

The diffinitive answer would be in the time it takes to complete the work unit. Let the optimized app run and complete the work. Then decide if it faster or not.

I have not seen anyone put a lot of faith in the flops reading you are looking at. Mine at the moment reads 0.08 GLOPS. So take it with a grain of salt and look at the completion times.

Oh.......welcome back to SETI!!!

---

Boinc....Boinc....Boinc....Boinc....

Thanks I will have it crunch a few WUs and let you know how it ends up. I started to investigate when my time remaining column jumped significantly.

Everything looks good. The duration for future WUs is trending down as BOINC is learning how long it will take me to crunch WUs. If it definitely faster than the stock app. Now, if I could figure out how the that validated on of my WUs crunched it in 3 minute.

1) For your CPU you have to choose SSE3 (AMD) (Not SSE3 (Intel), SSE2, SSE (p3))

2) The speed of the Lunatics' optimized CPU apps is 150-200% faster than standard app, but of course there are shorter and longer running tasks.
(those marked with "vlar" at the end of the name may run longer (on any app) but the credit is bigger (at least the "NewCredit" system tries to give more credit for longer running tasks on the same hardware/app))

3) The Lunatics' optimized ATI apps are at the moment the only way you can use your GPU for SETI

---

 


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

Those "crunched in 3 minutes" are tasks with "Error while computing":
http://setiathome.berkeley.edu/results.php?hostid=6136002&offset=0&show_names=1&state=5&appid=

No credit is given for them.


P.S.
To check everything is correct now:

1) Do you see 6 processes in Windows Task Manager with the name "AK_v8b_win_SSE3_AMD.exe" running?

2) Do you see 1 process in Windows Task Manager with the name "MB_6.10_win_SSE3_ATI_HD5_r177.exe" running?

3) When Astropulse tasks are running (may take ~15-20 hours on CPU or 2-3 hours (?) on GPU)
you will see one or several processes running (one per Astropulse task) with names:
CPU (up to 6 processes):
ap_5.05r409_SSE.exe

GPU (one process (if you don't modify your app_info.xml manually))
ap_5.06_win_x86_SSE2_OpenCL_ATI_r521.exe


All of the above .exe files you can see in (by default on Win 7):
C:\ProgramData\BOINC\projects\setiathome.berkeley.edu\

---

 


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

I do see all the processes as you listed them. Things have sped up considerably. I understand the WU that came up as errors for me. The 3 minute crunch I was referring to is found here: http://setiathome.berkeley.edu/workunit.php?wuid=802983649

??? But it's not "3 minute crunch"

It has "CPU time 2,442.01 sec" and "Run time 2,509.53 sec" which is ~42 minutes


If you refer to the other computer using nVidia CUDA GPU (SETI@home Enhanced v6.09 (cuda23) app)
the 155 sec CPU time is not telling much as the computation was done on GPU.
For GPU tasks look at the "Run time", not "CPU time"

955 sec (16 min) is normal Run time on nVidia CUDA GPU (GeForce 9800 GT)
with this kind of short tasks (VHAR, WU true angle range is : 2.570742)


Your computer may have finished some ATI GPU tasks ("SETI@home Enhanced Anonymous platform (ATI GPU)")
but they are still not reported so I can't tell how much time they take on your GPU.
(They will be reported automatically by BOINC in the next 24 hours.
BOINC delays the report for max 24 hours to report many tasks at once for less load on the servers)

---

 


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

That makes a lot of sense since that was run on the GPU. Thanks for all your help.

For SETI@home Enhanced VLAR tasks your GPU computes at the same speed as one of the CPU cores (~8000 sec/task (2+ hours)):
http://setiathome.berkeley.edu/results.php?hostid=6136002&offset=0&show_names=1&state=3&appid=

But for Astropulse tasks you can expect significant difference (>5 times) in the favor of ATI GPU

Your computer still do not have Astropulse tasks assigned for the ATI GPU:
http://setiathome.berkeley.edu/results.php?hostid=6136002&offset=0&show_names=1&state=0&appid=5

You can also see that the VHAR tasks ("shorties") are finishing ~3 times faster (~2500 sec) than VLAR tasks (8000 sec)
And the credit given is (almost) consistent with the Run time.
I think with the time you will get higher credit per CPU task as now you are at transition period and the SETI server "don't know what it's doing".
It still thinks that you used the standard app for all the tasks marked as "SETI@home Enhanced v6.03"
So after the first two took 15000 sec and the next few 8000 sec the server gives (as it thinks) "less credit for less work done by the same app"

The optimized CPU app finishes the VLAR tasks ~2 times (200%) faster than standard app (8000 vs 15000 sec)

---

 


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