Seti on a dualie, the best way?

[Confused]

Seems strange me asking here, i'm usually the one answering the questions!


My answer to this would be "Seti Driver with max processes at 2"

This the best way? can i track both WUs with SetiWatch over the network?

CBW

 

[Poof]

That would probably be the easiest way to do it. 🙂 And I think Setiwatch (along with Setidriver run that way) will only track the oldest WU being processed.

 

[SoulAssassin]

I would also set the cpu affinity so it runs one process/wu per specific cpu.

 

[Poof]

I would also set the cpu affinity so it runs one process/wu per specific cpu.

The problem with this is that the affinity setting on Setidriver will always set the process to CPU0. It won't separate them out when running 2 processes (ie., one process solely per processor - something I found out from Mike Ober way back). 🙁

[EDIT - And if you try to change it to CPU1 during the run, when the WU is finished, it'll start back on CPU0 again]

 

[ober]

Poof,

You misunderstood me. The context of that conversation was if you are running multiple instances of SETI Driver. Each instance of SETI Driver starts its CPU affinity with 0. A single instance will indeed lock a client onto each CPU until it runs out of CPUs, at which point, any remaining processing clients will float across CPUs. A transmitting client will always float across CPUs because it spends most of its time waiting for transmission.

Also, I think Mark Loupo updated SETIWatch to deal with monitoring a SETI Driver cache.

Mike.

 

[conjur]

Hmmm...did I read that right about SetiDriver? If I have two processors, I should leave the processor affinity checkbox UNmarked???

 

[ober]

On most systems, there is no noticable difference when using the "Set Affinity" checkbox. This checkbox only really helps when you have a 2Mb L2 processor cache (XEONs).

Mike.

 

[Poof]

The context of that conversation was if you are running multiple instances of SETI Driver. Each instance of SETI Driver starts its CPU affinity with 0. A single instance will indeed lock a client onto each CPU until it runs out of CPUs, at which point, any remaining processing clients will float across CPUs.

Okay cool! My situation has always been running 2 instances crediting 2 different accounts, and I was looking for a way to explictly specify a processor! 🙂

Thanks for the clarification Mike!

 

[networkman]

Honestly, I'm not using SetiDriver on any of my PCs; I just use the CLI in concert with SetiSpy(using two seperate directories, usually S1 and S2) and have each machine/client flush to the main SetiQueue that I have set up on one of my servers.

That was the easiest way for me. 🙂