TPU's WCG/BOINC Team

[Nordic]

s it worth upgrading the BOINC client?

As far as I know it is not worth upgrading every iteration. Maybe if you are really behind it would be. I shuffle rigs and reinstall OS's so often that I just keep the latest iteration on my hdd to install.

 

[thebluebumblebee]

Update: Never more than a few finished tasks waiting whenever I check. Normally none.

 

[Nordic]

I am testing some boinc settings to see if they will net more ppd. I am posting this here to let you guys know what I am doing, and maybe to see if some of you will test also.

You can edit such configs here. C:\ProgramData\BOINC\cc_config.xml

<ncpus>N</ncpus>
Act as if there were N CPUs; e.g. to simulate 2 CPUs on a machine that has only 1. To use the number of available CPUs, set the value to -1 (was 0 which in newer clients really means zero).

I am testing this one right now on my i7 4820k. It has 4 cores 8 threads. I set N to 14 which runs 12 wu's on the cpu. The estimated time to completion went up by 5 minutes. This setting reminds me of when we would run 10+ Gpu wu's on one cpu.

<no_alt_platform>0|1</no_alt_platform>
If enabled, the client will run applications only for its primary platform. For example, a Win64 machine will run only Win64 apps, and not Win32. List-add.pngNew in 5.9.10

This one I don't know if it would help. Ion has always said that x64 will get a 10% increase in ppd. Is that just the boinc client or does that also include the individual apps. I don't know what this setting will do as I have not tested it yet. One might set to only do x64 apps and then not get any work sent to them, or have a decrease in ppd. Such is the nature of testing.

 

[librin.so.1]

<no_alt_platform>0|1</no_alt_platform>
If enabled, the client will run applications only for its primary platform. For example, a Win64 machine will run only Win64 apps, and not Win32. List-add.pngNew in 5.9.10

This one I don't know if it would help. Ion has always said that x64 will get a 10% increase in ppd. Is that just the boinc client or does that also include the individual apps. I don't know what this setting will do as I have not tested it yet. One might set to only do x64 apps and then not get any work sent to them, or have a decrease in ppd. Such is the nature of testing.

Not all WUs are x86_64 capable. IIRC, CEP2 WUs are x86 only, so, by setting that You'd effectively disable it.

 

[Nordic]

Not all WUs are x86_64 capable. IIRC, CEP2 WUs are x86 only, so, by setting that You'd effectively disable it.

Good things to know. I don't like cep2 anyways. Never agreed with any of my systems. Like milk to a lactose intolerant person.

 

[Peter1986C]

That is because you run your crunchers heavily overclocked. They may pass Prime95 but that is child's play compared to CEP2. Or worse, climateprediction.net. By the Seven, what is that stuff sensitive.

 

[Nordic]

That is because you run your crunchers heavily overclocked. They may pass Prime95 but that is child's play compared to CEP2. Or worse, climateprediction.net. By the Seven, what is that stuff sensitive.

Not everything I have is overclocked. My a10-5800k actually runs under clocked, and it no likey the cep2. I also had my laptop back when I tried cep2 and it had an i3-2330m that refused cep2 work too.

 

[librin.so.1]

as listed here, CEP2 and HFCC lack x86_64 versions.

 

[Arjai]

I can't find the cc_config.xml!!

Nevermind, I found it on the BOINC partition! duh?!?

OK, so, I have 4 logical cores. I decided it would be better to use a multiple of 4, thus evening the workload between cores. I set it to <ncpus>12<\ncpus> and it is now running 12 WU's. I will let it run tonight to see how it goes. I hope this little thing will burn through them!!

 

[Norton]

I ended up doing a full dump on 3 of my rigs (remove/add project- not just a reset) and the newer FAAH wu's seem to be running much better!

My user average also jumped up by nearly 10%

 

[Arjai]

I changed the <ncpus> from 12, not running well, to 8, running quite a bit better!! I will keep an eye on it again tonight and in the morning. So far, seems good. I ran my computer today at work and it finished 10 jobs in 9 hours, one CEP that had been running all night and a couple MCM's that were running this morning, along with a couple FAH's and some more MCM's.

It seems to be ok, Hopefully it is a sweet spot that will up my numbers!

 

[Nordic]

So my 4820k got 9k ppd today. Either I got real lucky or that 12 threads is helping

 

[Norton]

So my 4820k got 9k ppd today. Either I got real lucky or that 12 threads is helping

Trying my FX-8350/Win7 rig on 12 cpus- just set it now so we'll see how a Vishera chip likes this

 

[Nordic]

Trying my FX-8350/Win7 rig on 12 cpus- just set it now so we'll see how a Vishera chip likes this

I have a suspicion that this will work best with cpu's with high single threaded performance. For each cpu would have to find the optimum number of wu's to run, before diminishing returns. I remember you guys thought I was crazy for running 20 gpu wu's but that is where I got the maximum output with my highly overclocked processor.

If tomorrow I have a repeat of today then I will test this on all my cpu's. My 4820 has 8 threads. I bumped it up to 12 wu's running, up from 8. That is a 50% bump. That cpu usually gets ~6k, so 9k would be 50% if it was linear. It would be really cool if it was. I doubt it will be linear though.

 

[Arjai]

I have a suspicion that this will work best with cpu's with high single threaded performance. For each cpu would have to find the optimum number of wu's to run, before diminishing returns. I remember you guys thought I was crazy for running 20 gpu wu's but that is where I got the maximum output with my highly overclocked processor.

If tomorrow I have a repeat of today then I will test this on all my cpu's. My 4820 has 8 threads. I bumped it up to 12 wu's running, up from 8. That is a 50% bump. That cpu usually gets ~6k, so 9k would be 50% if it was linear. It would be really cool if it was.

So, do you think running 8 WU's on an i5 3317U, 100% increase, is a good thing? You think 6, 50% would be better for me? I know that running12 was a bust, so I just dropped it this morning to 8, just because. Since you seem to have a little more experience doing this, is 50% the sweet spot? I can deal with that. I am wondering because I would rather not waste my time during a challenge testing every possible multiplication for a day...

 

[Nordic]

So, do you think running 8 WU's on an i5 3317U, 100% increase, is a good thing? You think 6, 50% would be better for me? I know that running12 was a bust, so I just dropped it this morning to 8, just because. Since you seem to have a little more experience doing this, is 50% the sweet spot? I can deal with that. I am wondering because I would rather not waste my time during a challenge testing every possible multiplication for a day...

I don't know yet. I am suspecting that the more single threaded performance the better this will work. I am starting with a 50% as it seems like a goldilocks number in my head that I can increase or decrease easily from if this proves to actually improve ppd. So far 1 day in it looks good. A week from now we will really have an answer.

My celeron g1620 would be an interesting test case being on the slow spectrum, it would be a good test for the lowest common denominator. If it can do 100%, 4 wu's then I would suspect any processor could do 100%.

 

[ThE_MaD_ShOt]

Trying my FX-8350/Win7 rig on 12 cpus- just set it now so we'll see how a Vishera chip likes this

Let me know how this does bud.

 

[Nordic]

Trying my FX-8350/Win7 rig on 12 cpus- just set it now so we'll see how a Vishera chip likes this

The more people who test this, even just 1 cpu, the more we can we if it works or not.

 

[Arjai]

I just forced an update, looked at my User Average, on stats.

I switched to 8 WU's 100% this morning....

 

[Norton]

The more people who test this, even just 1 cpu, the more we can we if it works or not.

Going to give it 24hrs to see what happens- if it's positive then the other 8350 and possibly the Xeon are going next....

 

[Arjai]

Just got taller.

 

[librin.so.1]

I am going to try running 1.5 WU per core for a total of 12 WUs per CPU on my both of my FX-8320 and see how it rolls.

EDIT: At this point I am very worried about Kai / @[Ion] / Aperture_Science_Innovators
He's just playing Europa Universalis IV non-stop, almost doesn't crunch any more and hasn't visited TPU since 2014-01-10
This makes my spidey sense go through the roof. I hope Kai's OK

 

[t_ski]

EDIT: At this point I am very worried about Kai / @[Ion] / Aperture_Science_Innovators
He's just playing Europa Universalis IV non-stop, almost doesn't crunch any more and hasn't visited TPU since 2014-01-10
This makes my spidey sense go through the roof. I hope Kai's OK

He probably got burnt out on crunching, perhaps getting in trouble for his power usage/noise/heat from a new roommate. Seeing as he's still in school, perhaps school pressures are an issue, too.

 

[Norton]

I am going to try running 1.5 WU per core for a total of 12 WUs per CPU on my both of my FX-8320 and see how it rolls.

EDIT: At this point I am very worried about Kai / @[Ion] / Aperture_Science_Innovators
He's just playing Europa Universalis IV non-stop, almost doesn't crunch any more and hasn't visited TPU since 2014-01-10
This makes my spidey sense go through the roof. I hope Kai's OK

Weird seeing my 8350 reported by FreeDC as a 12 core

I hope Ion works through whatever is going on with him and rejoins us here soon. Would also like to see Chicken Patty, KieX, Daimus, and a few others stop by from time to time.....

 

[thebluebumblebee]

Weird seeing my 8350 reported by FreeDC as a 12 core

How do you see your hosts?
Edit: Do I have to check "Show hosts" @ WCG?