So the current source for my SoB crunching is my Athlon 64 3200, a mighty fine machine that’s lasted me well for a couple years now. But I finally couldn’t take it anymore and yesterday I ordered some more parts, one of which is a Athlon 64 X2 4200 (AM2 Windsor), to build a new machine. (Can’t wait!)
I’ll still be running the good ole’ 3200, so the dual-core should more than double my current overall production. Does anyone know whether it’s more productive to run one instance of SoB or two instances (one for each core)?
j_freeman wrote:I'll still be running the good ole' 3200, so the dual-core should more than double my current overall production. Does anyone know whether it's more productive to run one instance of SoB or two instances (one for each core)?
The SB client will only use one logical processor per instance… so you’ll want to definitely run separate instances per core.
For my HT-enabled P4’s - even run 2 instances for a single core because the HyperThreading gives a slight bit of extra crunching when running a second instance. Each instance runs less than by themselves, but together they equal more production overall.
umccullough wrote:j_freeman wrote:I'll still be running the good ole' 3200, so the dual-core should more than double my current overall production. Does anyone know whether it's more productive to run one instance of SoB or two instances (one for each core)?
The SB client will only use one logical processor per instance… so you’ll want to definitely run separate instances per core.
For my HT-enabled P4’s - even run 2 instances for a single core because the HyperThreading gives a slight bit of extra crunching when running a second instance. Each instance runs less than by themselves, but together they equal more production overall.
Awesome, work those bad boys! 
I’ve never had a dual-core, HT, or SMP machine before so this is all new to me. 
I might be posting here if I run into some problems setting it up.
j_freeman wrote:I've never had a dual-core, HT, or SMP machine before so this is all new to me. :) I might be posting here if I run into some problems setting it up.
If you’re using Windows - I assume you’ll be using sobsvc.exe - there are good instructions on how to use that in one of the readme files that comes with the client.
If you’re using Linux, you’re on your own 
- I’m sure there are instructions on specifying which processor the client runs on.
No problems, once I got the parts working (bad memory stick), one of the SoB FAQs got straight to the point on how to install it as a service for each core… just had to add a flag to the sobsvc command.
I’m getting about 4M cEMs/sec, which is about what I was getting from all my machines combined, and I haven’t even set those back up yet. 
j_freeman wrote:No problems, once I got the parts working (bad memory stick), one of the SoB FAQs got straight to the point on how to install it as a service for each core... just had to add a flag to the sobsvc command.
I’m getting about 4M cEMs/sec, which is about what I was getting from all my machines combined, and I haven’t even set those back up yet.
That’s awesome!
I’ve been messing around with BOINC lately - and I notice that running SB clients alongside BOINC actually retains fairly good performance (why? I have no clue!)… So far, SETI and RieselSieve are pretty easy and small on BOINC (not too huge) - ClimatePrediction and BBC Climate Change Experiment are HUGE… I don’t recommend more than one BOINC project per machine if you’re also running SoB though 
I have to stop and restart the SB clients to get them all running simultaneously, since BOINC appears to defer to SB, but not the other way around
Also, installing DIMES (http://www.netdimes.com) as a service is easy - and it doesn’t use much CPU at all (mostly idle) We are "Team Haiku" on Dimes.
Hmm, never heard of either of those. I’ll have to check them out.
j_freeman wrote:Hmm, never heard of either of those. I'll have to check them out.
Check the sticky thread at the top of this forum - it has most of the projects where Team Haiku exists
More experimentation shows that not all BOINC projects behave the same. It would appear that each project can have different results when working side-by-side with SoB clients. SETI seems pretty decent - but many of the others i’ve tried so far don’t get enough processor time to be effective…
RieselSieve maybe does…
Let me correct my previous statements…
The reason I was seeing different cpu-sharing results between different BOINC projects (which i’m running on separate machines) is mainly because I had the SB client configured differently on some of my machines. Some were running at Low prio, while others were running at “Idle” (using the SB config screen)…
When running SB at Idle, SB and BOINC projects seem to share the processor time pretty equally - but when running in Low, SB seems to take most of the cycles for itself and BOINC doesn’t get very many.