TheSkyNet news thread - 17/10/11

shadowscotland · 17 Oct 2011 at 14:17

After last weeks hardware issues at mission control I'll like to start by saying many thanks to all the tech's that when in at the weekend
and got it sorted. Cookies for you all. As we know you read/hear about this thread

Another great week for the team - Stomping up 4 places on the leaderboard.
Is the dizzy heights of page one (top 15) possable - watch this space :cool:

A new coloumn for you all this week - gain/loss of average ppd from last week
Slacker have been painted RED - Oh the shame

An cool cucumbers are green :cool:

The return of an old graph. but with rainbow colours - red is now, violet is last monday both are labled
Smaller bar, higher your global position

shadowscotland · 17 Oct 2011 at 14:19

All of us - almost on one graph

shadowscotland · 17 Oct 2011 at 14:23

Which day did the servers go down - hmmmm not sure

Lots of us now and some mightly impressive numbers

TheSkyNet.org

previous theads 10/10, 3/10, First post

Acknowledgements

All banners in news posts are courtesy of Halz

ToxicTBag · 17 Oct 2011 at 14:29

Great job with the news Mr shadowscotland sah much appreciated

shadowscotland · 17 Oct 2011 at 15:39

That took a whole lot longer that it was surposed to.
Sorry it wasn't ready at lunchtime

News 3.1
Table revised to show all new members
Missing 'Global Position' graph added

News 3.2
Revised both top graphs - colour / type

greywolf · 17 Oct 2011 at 15:45

Nice work on all the graphs, I like graphs

Nice to see a weeks work layed out.

Oh and if this only uses core and not threads, im only actually running 6 then as im on an i7 990x.

derk · 17 Oct 2011 at 16:30

Nice job mate, very informative

- 1,300 views on the last news, lets hope this is a sign of things to come.

Its a shame 'theSkynet' started right in the middle of the Team Competition but its ground that we cannot
afford to lose, on the plus side hopefully our output will increase once its complete.

I knew my stay at the number one spot would be brief, just didnt realise how brief it was going to be until I
saw those graphs ......sob

..dam you Gulftown cpu

Again, well done mate

ToxicTBag · 17 Oct 2011 at 21:37

Bit perplexed ( i've taken to reading dictionaries before i open my gob now)

As per Mr shadowscotlands advice about it being core rather than thread related i have 4 instances running on an sb setup/ 4 on a q6600 setup and 1 on a monster you wouldnt scrape muck of your shoe with.
It was showing fine x machines etc now when i log in the dashboard wants to have me believe 0 machines online and 11 today, even though i have all these windows open am i still getting credit or do i ned to close em and start again.

Would appreciate the help thanks muchly, as long as i know i can set n forget am happy to continue

PhilthyPhil · 17 Oct 2011 at 21:46

I think it is down atm, my dashboard also says zero machines online and none of them seem to have work - no CPU usage, although they still use 500mb memory

ToxicTBag · 17 Oct 2011 at 21:49

Ahh thank you for that PhilthyPhil just wanted to be sure i was doing it right, at least its relatively simple to get going .... us old fartknockers prefer things that way after a certain age H'onward and Skyward!

BaBa · 17 Oct 2011 at 23:22

I am coming to help out on this ,is anyone running it on linux?

Just wondered how the beta client was working out,will it handle 32 instances on one rig?

Also what sort of ppd are people getting from what cpu's?

Ageotas · 18 Oct 2011 at 06:49

Hi everyone, Tim from theSkyNet Team again.

Just to help clear up some questions you obviously all have:

1/ The Nereus client is not hyperthreaded, so it will only reliably use physical cores, meaning that yes, an i7 Gulftown CPU will only use 6 clients simultaneously, a Bloomfield will only use 4, etc. I have seen people get around this by running it through virtual machines, but keep in mind that if you are doing this, you are running the client outside of the parameters that we can reliably test here in the lab, and so the amount of tech support we can offer will be reduced. Saying that, you guys have a forum of very intelligent people, and I don't doubt you could figure any issues out between yourselves.

2/ If you have not already, it would be a good idea to reinstall the client with our newest version (0.1) which has a new way of handling load balance that is much better. Yes, I know it is a pain to uninstall and reinstall, but at the moment the Nereus client doesn't upgrade nicely. It is something we are working on.
http://www.theskynet.org/download/index

3/ Linux Beta is available here. The same rules apply for cores vs clients as in the Windows or Mac environment, and the same deal applies if you want to attempt running them on virtual environs.
http://theskynet.org/account/linux

Thanks for being rad, guys!

Tim
(AKA Ageotas)
(AKA Corsair)
- theSkyNet Team

derk · 18 Oct 2011 at 07:16

Ok, going to install the latest client now,thanks.

Question - if for some reason you have to kill one instance of theskynet do you lose all credit for work done so far or will it restart from where it left off.

Ageotas · 18 Oct 2011 at 07:32

halz said:
Ok, going to install the latest client now,thanks.

Question - if for some reason you have to kill one instance of theskynet do you lose all credit for work done so far or will it restart from where it left off.

You will lose the credit for that job, but not for any previous jobs that the client has processed. After a job is finished it is immediately uploaded, and that is when the credit is awarded. While you lose the job you were working on, there is enough redundancy in the system to make sure that we don't loose any data.

The credits you have accrued for being connected are STILL awarded, even if you have to force close the client; that one is measured by time connected, not be how much you have crunched.

Tim

shadowscotland · 18 Oct 2011 at 11:10

Welcome back to our news thread Tim - always good to have your input.
What you've said above make sense - and matchs what we've found in practice.

It's a shame that the partially complete data packaged isn't 'saved' for completion on restart,
But I can understand why, based on the 'background client' ethos

Can I ask how long does it take to process a single packet of data on say a SB system.
Or any tested system for that matter

My guess is it's well under a hour, but I haven't sat and watched my network traffic with a stopwatch - well not yet anyway (Halz might have

)

derk · 18 Oct 2011 at 11:50

shadowscotland said:
My guess is it's well under a hour, but I haven't sat and watched my network traffic with a stopwatch - well not yet anyway (Halz might have )

I am actually anal enough to do that

but I'm back on the night shift so dont have much time for the next few days

I am however still monitoring inter team performance and at present our ppd would ultimately get us to position number 8.

However, our output is still rising so who knows how far we will go

....and better still OcUK are pulling away from ze Germans - das ist gut ya :eek:

Colasanto · 18 Oct 2011 at 12:37

Do you reckon in the future there could be the option to have a in your face client where you have full control over everything? Could this client also have CUDA support in the future for those who wanted to donate more time resources?

Wishful thinking.

derk · 18 Oct 2011 at 13:22

ermmm......something's not quite right, just been awarded a new trophy

The Space Traveller - 384,000

I actually have less than 20,000 :confused:

.....I dont mind though

Colasanto · 18 Oct 2011 at 13:44

Checked the PC at mid-day and had 3 instances running fine. Just gone to check again (1hr 30 later) nothing running, shows as 0 MIPS and no credits being given.

greywolf · 18 Oct 2011 at 14:22

It does seem alittle up and down at the moment so keeping an eye on it.

I also seem to have collected the 100,000 and 384,000 trophies.