[electrician1972]

I've been trying to get my system setup with Seti optimized for Altivec and my numbers are now higher in terms of compute time. I started out with the non optimized boinc and seti programs (from the setiathome site) and was getting times of 22000 to 27000. Then I put on the optimized boinc and seti programs (from the MacNN site) and the times are now 27000 to 28000! Shouldn't it be the other way around? Hardware is as follows:

1 GHz G4 7447A, 1MB L2 Cache @ 1GHz
768 MB RAM
Boinc version - 4.19 off Team website
Seti version - 4.30 off Team website

Could I have a bad Altivec subprocessor? Zap the PRAM?

[Shaktai]

Good to hear those links are working again. They were down for a very long time.

Did you include the app_info.xml in the project folder along with the optimized client? It must be in the Seti project folder along with the application.

If you did, you should see a line in the terminal that reads:
"Found app_info.xml; using anonymous platform.

If you didn't get that line, then it didn't recognize the optimized app and is using the old SETI app.

If everything above is okay then: You indicate you have a 1 GHz G4 7447A. If you downloaded the G4 7400/7410 version, try using the G4 7450 and see if it works better. (or vice versa). Remember to use the correct app_info.xml file, because that is necessary to point to the optimized client.

Also, how many work units did you run? Seti work units vary in time dramatically sometimes. The longer work units are the better ones, because that indicates the application is finding more interesting information. You will usually need to run several in order to get a feel for what it is doing.

If that doesn't fix it, let us know.

Oh, you might also want to try the "beta" 4.27 client version from our beta page. It is stable and works well.

[electrician1972]

D-OH! I forgot to mention I was using the 7450 Altivec optimized. I was not able to determine from the forums if the 4.27 beta was good to go, so I decided to stick with the safe bet. Thanks for the information. Looking at my results today, there are now a few that are under 27000. Guess I could've waited a little longer before hitting the panic button.

Brings to mind a new question. I have a work unit in progress on my system and would like to finish it, stop the client, switch to the 4.27 beta worker, then startup again and not lose a work unit. Any ideas how this could be accomplished?

Thanks!

[Shaktai]

Originally Posted by electrician1972

D-OH! I forgot to mention I was using the 7450 Altivec optimized. I was not able to determine from the forums if the 4.27 beta was good to go, so I decided to stick with the safe bet. Thanks for the information. Looking at my results today, there are now a few that are under 27000. Guess I could've waited a little longer before hitting the panic button.

Brings to mind a new question. I have a work unit in progress on my system and would like to finish it, stop the client, switch to the 4.27 beta worker, then startup again and not lose a work unit. Any ideas how this could be accomplished?

Thanks!

Actually, you shouldn't lose any work if just switching the BOINC client. It is when you switch the SETI app that work is lost. Just in case though, Backup your project folder before switching.

Otherwise, you can set your cache to zero days. It will complete the current work, but not download any more. Once it is empty make the switch. With the GUI, you can actually specify "no more work" and it will complete what's in cache without downloading.

[electrician1972]

OK, now I have the 4.27 BOINC client and the 4.30 7450 Altivec worker bee. The benchmarks are not any different from the 4.19 BOINC client.
729 Whetstone
1725 Dhrystone

BUT they are considerably higher than the non-optimized stuff:
489 W
1522 D

I realize this doesn't get the work unit done any faster, but does it get you more credit from the project?

[Shaktai]

Originally Posted by electrician1972

I realize this doesn't get the work unit done any faster, but does it get you more credit from the project?

It might. It will result in higher claimed credit. However the actual award credit is a compromise of the claimed credits. Once in a while you will get claimed credit, but instead of being on the low end, you will be on the high end. This may actually increase awarded credit for everyone for the given work unit.