Greatful for a sanity check from the great & the good :)

I switched to Astropulse from CUDA WUs a couple of weeks ago as I wanted to support a particular Project on CUDA for many reasons. End result of switching around which machine does what etc to make room for the Project is that I ended up with two cores of an AMD Phenom2 running Astropluse.

Seems to be going fine, just want to check I am in fact using the right Opt App, I am getting either side of 12hrs per WU (7x24) depending on wind direction and whether there is an R in the month :)

The Phenom2 reports via CPU-Z SSE, SSE2, SSE3, SSE4A. I went for SSE3, and use ap_5.03r112_SSE3 app. Good to go?


Also as part of the switch round I will be dedicating either a Core2 Duo T7250 2Ghz Laptop (A VAIO) or an AMD FX60 dual core beastie to Astropulse. I have one core of each currently running ap_5.03r112_SSE3 app as a test, again both appear ok, laptop crunches in about 20hrs, the FX60 in about 24-48hrs (gets lots of other use so a bit slow on expected times propably, and always varies due to the other activities) - both 7x24 - good to go?

CPU-Z/Laptop reports SSE, SSE2, SSE3, SSSE3
CPU-Z/FX60 reports SSE, SSE2, SSE3

Edit:
Sorry, one more thing - is Astropulse ok running on "all cores" or will it benefit from having one or more cores running other project WUs, in order to maximise cpu architecture? I run other cpu based WUs, so have the option to "mix and match" as appropriate.

Regards
Zy

---

The Phenom2 reports via CPU-Z SSE, SSE2, SSE3, SSE4A. I went for SSE3, and use ap_5.03r112_SSE3 app. Good to go?

Yes, provided you're specifically referring to Astropulse_v5 (the current run). Don't use r112 for the older version.

Sorry, one more thing - is Astropulse ok running on "all cores" or will it benefit from having one or more cores running other project WUs, in order to maximise cpu architecture? I run other cpu based WUs, so have the option to "mix and match" as appropriate.

I've not noticed any problems with running AP on many cores - and I run a dual-quad Xeon which *does* choke on, e.g., 8 x VHAR MB running at the same time.

Greatful for a sanity check from the great & the good :)

I switched to Astropulse from CUDA WUs a couple of weeks ago as I wanted to support a particular Project on CUDA for many reasons. End result of switching around which machine does what etc to make room for the Project is that I ended up with two cores of an AMD Phenom2 running Astropluse.

Seems to be going fine, just want to check I am in fact using the right Opt App, I am getting either side of 12hrs per WU (7x24) depending on wind direction and whether there is an R in the month :)

The Phenom2 reports via CPU-Z SSE, SSE2, SSE3, SSE4A. I went for SSE3, and use ap_5.03r112_SSE3 app. Good to go?


Also as part of the switch round I will be dedicating either a Core2 Duo T7250 2Ghz Laptop (A VAIO) or an AMD FX60 dual core beastie to Astropulse. I have one core of each currently running ap_5.03r112_SSE3 app as a test, again both appear ok, laptop crunches in about 20hrs, the FX60 in about 24-48hrs (gets lots of other use so a bit slow on expected times propably, and always varies due to the other activities) - both 7x24 - good to go?

CPU-Z/Laptop reports SSE, SSE2, SSE3, SSSE3
CPU-Z/FX60 reports SSE, SSE2, SSE3

Edit:
Sorry, one more thing - is Astropulse ok running on "all cores" or will it benefit from having one or more cores running other project WUs, in order to maximise cpu architecture? I run other cpu based WUs, so have the option to "mix and match" as appropriate.

Regards
Zy

The best you can get for the Astropulse app is the SSE3 version. Its only the Multibeam app that has SSSE3 and SSE4.1 versions. The AMD's as you have noted only go up to SSE3 support.

We tend to find if you have a hyperthreading CPU (older P4's or newer i7's) that they benefit from a mix of tasks, but for "normal" cpu's it doesn't really make a difference.

There was a suggestion made to the boinc developers if they could incorporate the logic to have a mix or work, but don't hold your breath waiting for it, they have enough issues with the current task scheduling.

---

BOINC blog

Many Thanks for the check Richard, Most Grateful.

Its V5, not the older one.

The "all cores" question was prompted by a CPDN issue where no more than two hadam3p models should be run on a quad - three or four will run, but they end up fighting each other for space and choking a bit. On the Phenom2 I now have two 3p's and 2 Astropulse and they seem to live "happily ever after" with each other.

Regards
Zy

---

Thanks Mark Most Grateful

Regards
Zy

---