DC Vault 2

CaptainUnlikely said:
Got a little bit of reorganisation done today and pointed one machine over at NFS to get a handle on how it goes. Most of my machines are crunching over at Leiden right now to give Gilthanis a hand, see if we might be able to pull one or two places in that before I refocus for April.
Click to expand...

Had a bit of trouble with one system running NFS but it seems stable now, for some reason the 6860 and 5770 weren't playing nicely even though they were OK in my old setup. The 5770 got stuck on a Collatz unit for 5 hours, aborted it and it then froze the system every time about 5 seconds after getting to the desktop, so pretty much as soon as it started crunching. I only mention it along with NFS as my hex doesn't like NFS at all, system totally locks up within 2-5 minutes of starting work. If I suspended NFS, it was fine. It's been crunching Leiden for a while with no issues and passes stress tests so I'm not sure if I just have a funky config or there were a bunch of bad units, but I've pulled it back off for now.
 
jondi said:
Hey guys, long time lurker, first time poster :)

For some reason Nvidia 5** series cards out perform many of the newer Nvidia cards, they are comparitable in performance to AMD 79** cards.
Click to expand...

Am I crazy thinking it's not a bad idea to acquire some used GTX 560 Ti 448-core cards and put them to work? There are some fairly inexpensive ones as well as GTX 580's on eBay (around $100-$125each). Looking at the GHz-days/day figures from http://www.mersenne.ca/mfaktc.php?sort=ghdpd, I want to make sure I'm not misreading it.

EDIT: It appears the GTX 560 Ti 448, GTX 570 and GTX 580 are really good at GIMPS, but not so good at Milkyway or other BOINC projects. I believe buying these for science would be a very specialized effort (i.e. GIMPS and PrimeGrid-only).

When looking at PrimeGrid's GPU list, it appears the GTX 580 gets half the performance of a 780 TI for 1/7th the cost. :) http://www.primegrid.com/gpu_list.php
 
Last edited:
I'm posting this here since it impacts us at the Vault and probably wont make much of a splash in the BOINC thread. I got this in my email today.

Dear

Thank you for your continued support of theSkyNet and our project SourceFinder. You are receiving this email because you've processed data for theSkyNet SourceFinder in the past 7 days. This announcement will soon be uploaded to theSkyNet website, but we wanted to contact you first to let you know personally.

Due to continuing issues with Java and the great feedback we’ve been getting for theSkyNet POGS (on BOINC) we've been planning on shifting SourceFinder over to BOINC some time later this year. We had planned for a smooth transition but unfortunately a technical issue has brought SourceFinder processing to a sudden and unexpected halt.

After much discussion here between theSkyNet team we've decided to close down SourceFinder today instead of fixing this issue. We think it's not a good use of resources to concentrate our development work on fixing SourceFinder given the ongoing issues with Java. Instead, we're going to spend that time and effort bringing SourceFinder to BOINC faster than originally planned.

This obviously means there will be a gap in availability for SourceFinder processing but theSkyNet POGS will continue processing (as it has been) with no change. So for those members not already processing data for theSkyNet POGS, take a look and get involved if you’d like.

These technical issues have not affected the results or data gathered so far through SourceFinder, and we will be able to pick up where we left off with BOINC sometime later this year. We'll let you know timeframes as soon as we have them. These issues have also not affected credits or your stats on theSkyNet - if that doesn't appear to be the case for you, please contact us ASAP and we will help you as fast as we can.

Thank you so much for your support of theSkyNet SourceFinder. As our original project it holds a special place in our hearts and we're looking forward to reigniting it as soon as we can. You'll soon find a special trophy in your trophy cabinet to thank you for your support.

For your interest, theSkyNet SourceFinder has been running for a total of 925 days, and in that time 6,387 volunteers using 355,967 clients have donated 39,618,920 computing hours to process a grand total of 319,948,440 MB of data. That was all you – thank you for your contributions!

We also have a few prize packs to send out, and we’ll do a random drawing of everyone that’s helped process on theSkyNet SourceFinder since the start and let you know the winners on theSkyNet website soon.

We're sorry for any inconvenience this might cause, and as always we're so appreciative of everyone that is part of theSkyNet and processes data for any of our projects.

Please feel free to contact us with questions.

TheSkyNet Team
Click to expand...
 
brilong, if I'm remembering my history correctly, nVidia crippled crunching performance when they went to the 600 series. It improved gaming but performed worse/less efficient/or not much better than the 500 series cards. However, that has been a while and my memory is cloudy. But if I also recall correctly, it was mostly dual precision that got hammered.

Milkyway and PrimeGrids GFN work units are dual precision. If you look at GFN work units, the 580 is almost just as good or better....
 
Last edited:
Gilthanis said:
brilong, you could always just pick one of these up. http://www.anandtech.com/show/7897/nvidia-announces-geforce-gtx-titan-z-dualgpu-gk110-for-3000
Click to expand...

Yeah, no thanks. I have been looking at PrimeGrid a little closer. Not sure about the true usefulness of the Math projects compared to the Bio/Medical ones, but need to increase helping our Vault scores. :)

Need to figure out how PrimeGrid scoring works and if it's better to concentrate on one or two subprojects or run something non-BOINC like GIMPS.
 
Usefulness is only valuable in hind sight. I like the idea of Bio/Medical, but Prime numbers certainly benefit society a great deal in the real world and don't have to wait for some greedy corporation to take notice of the research to use them. Most of the practical uses of the Prime numbers is encryption. Other than that, it is nothing more than a math geeks obsession. Then you look at the Bio/Medical projects and have to wonder how much of that research actually makes a difference and whether the efforts/resources were really worth the small bump. That part is extremely hard to tell because most of what we do are the foundation to the skyscrapers yet to be built. I will say that it was refreshing to get feedback from WCG in regards to the 7 drug candidates that came from the HFCC project. It isn't very often that you get real usable results and then get the feedback in a timely manner. I also tend to look at some of the other projects like CEP2, C4CW and C4SW. Cleaner energy and clean water will help more people than a cure for cancer. So, if I wanted to do the most "good", I would focus all of my efforts into projects like that and not worry about cancer. Cancer is small compared to all the problems that filthy water brings in this world.... So, it really just comes down to what makes you feel good. Practicality or the warm fuzzy feeling you get patting yourself on the back. That is why I crunch a little of everything. You never know what will actually have real results.
 
CaptainUnlikely said:
Had a bit of trouble with one system running NFS but it seems stable now, for some reason the 6860 and 5770 weren't playing nicely even though they were OK in my old setup. The 5770 got stuck on a Collatz unit for 5 hours, aborted it and it then froze the system every time about 5 seconds after getting to the desktop, so pretty much as soon as it started crunching. I only mention it along with NFS as my hex doesn't like NFS at all, system totally locks up within 2-5 minutes of starting work. If I suspended NFS, it was fine. It's been crunching Leiden for a while with no issues and passes stress tests so I'm not sure if I just have a funky config or there were a bunch of bad units, but I've pulled it back off for now.
Click to expand...

It could be your machine running out of memory, the NFS WUs that give the most points use just over a gig of ram each, another type half a gig, you can specify what WUs you want to receive, also the WUs run about a third quicker in Linux.


brilong said:
Am I crazy thinking it's not a bad idea to acquire some used GTX 560 Ti 448-core cards and put them to work? There are some fairly inexpensive ones as well as GTX 580's on eBay (around $100-$125each). Looking at the GHz-days/day figures from http://www.mersenne.ca/mfaktc.php?sort=ghdpd, I want to make sure I'm not misreading it.

EDIT: It appears the GTX 560 Ti 448, GTX 570 and GTX 580 are really good at GIMPS, but not so good at Milkyway or other BOINC projects. I believe buying these for science would be a very specialized effort (i.e. GIMPS and PrimeGrid-only).

When looking at PrimeGrid's GPU list, it appears the GTX 580 gets half the performance of a 780 TI for 1/7th the cost. :) http://www.primegrid.com/gpu_list.php
Click to expand...

I can vouch for those 570 speeds in GIMPs, mine overclocked do over 400 ghz/days, so do my 7950s. I also have 2 r290s that do close to 600 ghz/days each, there are also awesome at collatz but they don't currently work on milkyway.
 
jondi said:
It could be your machine running out of memory, the NFS WUs that give the most points use just over a gig of ram each, another type half a gig, you can specify what WUs you want to receive, also the WUs run about a third quicker in Linux.
Click to expand...

8GB in both machines I had trouble with, I did a little reading and saw some of the apps have large RAM requirements and these were 8 and 12 thread machines, but the usage was never over around 50%, so it doesn't look like they were grabbing the large units and choking. Maybe I'll try it again on this rig and limit the core usage to drop the requirements, see if it helps. Good to know Linux is significantly faster though, I only have one Linux box up at the minute but I'm leaning towards deploying a few more.
 
Out of the Error work units I have recently seen, most were with the 16e Lattice Sieve v1.08 (notphenomiix6) app. I will try and keep a closer eye on it. I have two different errors amongst two machines. One seems to be a file access issue..so not sure if it is malware protection or not. The other one I couldn't tell and it is a borg anyways.
 
I have noticed recently while diagnosing other project issues, that Malwarebytes seems to have been blocking some of my DIMES transmissions. That and cgminer that was being used at BitcoinUtopia was being picked up too. So, possibly time to re-evaluate Malware protection options and whether it is worth the hassle of adding exceptions to each. If these were only used for DC'ing, I wouldn't even bother with anything but the basics, but alas my wife has problems clicking before she reads....
 
All the computation errors I had with the 2600 machine were with the 16e app, yes. The errors I got were "1 (0x1) Unknown error number" and looking on their forums, some other crunchers have had the same issue, but there's no clear cause or fix I can see.
The ones on my hex that was just locking up were a mix of 14e and 15e, I ended up aborting the lot as it just wouldn't stay stable long enough for me to do anything. I'll give it some new tasks, but slowly, when I have some time to babysit it and find out where the problem lies.

I use only MSE on my non-dedicated machines and I exclude the BOINC directory just in case. On my dedicated crunchers I don't even bother with an AV, they don't have access to anything on my network and worst case scenario I can just wipe the SSD and have the machine running again inside an hour.
 
So, I have been in touch with the admin and he says that when they move Source Finder to BOINC, that it will most likely be called Source Finder 2 and they haven't decided if it will be on the same server as POGS yet. He also said that it would have a new set of points for Source Finder 2, so any points from the original Source Finder will basically become historical data. I posted in this forum because it could be a change to the current POGS project.
 
jondi said:
I can vouch for those 570 speeds in GIMPs, mine overclocked do over 400 ghz/days, so do my 7950s. I also have 2 r290s that do close to 600 ghz/days each, there are also awesome at collatz but they don't currently work on milkyway.
Click to expand...

I couldn't resist the temptation and picked up a GTX 560 Ti 448 Cores for $122 last night. I also picked up a GTX 580 for $155. I just created my GIMPS account and joined the HardOCP team, so I'm ready to try out these two cards when they arrive. :D
 
brilong said:
I couldn't resist the temptation and picked up a GTX 560 Ti 448 Cores for $122 last night. I also picked up a GTX 580 for $155. I just created my GIMPS account and joined the HardOCP team, so I'm ready to try out these two cards when they arrive. :D
Click to expand...

That is awesome news. I haven't checked the stats yet at GIMPS, but it appears we dropped a spot again.
 
Well, I tried the Muon1 client out, it seems to only be Windows or wine... but config is easy just create a user.txt with [H]username in it and edit config.txt for number of theads/core you want... I pointed two cpus are it for a dayish and got 43K points.... and looking at historic posts is looks like we can combine our results.dat files together with higher positive seed and get better points overall...
 
Well, it was just unzip the download, add a user.txt and run the command line batch file :) It took me longer to figure out our team name abbr (team names are anywhere in the username wrapped in square brackets so [H]NickOfTime or NickOfTime[H] etc...), I'll try allocating all 24 cores to it later today after current pogs wu complete, for a more defined time period and get a more specific point value...
 
Last edited:
Gilthanis said:
That is awesome news. I haven't checked the stats yet at GIMPS, but it appears we dropped a spot again.
Click to expand...

I've been reading about mfaktc (TF) and CUDAlucas (LL) but I don't quite understand which might be better to increase our points and efforts. I'm not concerned about reward money (LL).

Which ones have you guys run for GIMPS?
 
I've not ran GPU yet. My one test laptop is running CPU Prime95 client. I was saving the GPU testing for a later date. ;)
 
Well, it looks like the best way to run muon would be thru the yoyo client, since it runs most efficiently as one thread per cpu anyways. We can get credit under yoyo/wuprop and muon if you edit your nickname for yoyo to include an [H] . It only runs under windows with a 32bit executable.

I was getting about est 17,500 mpts points per cpu/day from muon1 and it looks like roughly 1/2 point per cpu minute from yoyo, so 700 points per cpu/day there also...
 
Last edited:
NickOfTime said:
Well, it looks like the best way to run muon would be thru the yoyo client, since it runs most efficiently as one thread per cpu anyways. We can get credit under yoyo/wuprop and muon if you edit your nickname for yoyo to include an [H] . It only runs under windows with a 32bit executable.

I was getting about est 17,500 mpts points per cpu/day from muon1 and it looks like roughly 1/2 point per cpu minute from yoyo, so 700 points per cpu/day there also...
Click to expand...

Is that as simple as logging into your YoYo account and clicking on other account info and then changing that name to include [H] or do you change your nickname somewhere else? And why have our veterans not point this out by now? I have almost half a million points for Muon at YoYo...that could have went towards the team. I hate double dipping as an option in the Vault, but if it is allowed, then no reason not to be competitive like the other teams...
 
Last edited:
Holy shit balls, our whole team needs to adjust their nick names then on YoYo....
 
Well, unless someone finds a distinct advantage to running their climate, I see no reason to use anything but the BOINC client and attach to YoYo.
 
Gilthanis said:
For those that have already ran Muon work through YoYo/BOINC, after changing your nickname at YoYo you will need to wait until the new nick shows up under our team on Muon1's stat pages. http://stephenbrooks.org/muon1/stats/?optim=*&team=[H]&sortby=Mpts&user=*&oldness=168 Then you can email them to merge your names so that your stats on Muon1's page are all under our team.
Contact them at this address:
[email protected]
They were quick to respond to me.
Click to expand...

Does this mean that past credit on muon1 via yoyo on boinc can be credited to the muon1 (DPAD) non boinc stats? .. backdated so to speak?
 
jondi said:
Does this mean that past credit on muon1 via yoyo on boinc can be credited to the muon1 (DPAD) non boinc stats? .. backdated so to speak?
Click to expand...

That is what I'm about to find out. I didn't do any DPAD but was listed under Coleslaw because of YoYo. Now that I have changed my nick at YoYo, I am now showing up under our team stats except my points under the team just started. They just list that it was from the BOINC wrapper. Stay posted on whether they merge my pre YoYo name change to the team points or not..... :cool: (he gets back to work on Monday and that is when he said he would merge my stats) But yes your YoYo stats do show up on the DPAD stats under the nick you use at YoYo...so if you have crunched both, you may be listed on their stats twice....
 
Last edited:
Good, that should give us 7 Tpts more , some space in muon since were only ahead of 10th postion by 7 Tpts...

Hmm, my [email protected] system will only contribute about 1 Tpt every 3 days, which is abit less than Position 8's current daily contribution...

Now we just need another 100 Tpts and 7.14M yoyo points for the next position for 133 muon and 34 yoyo = 150 vaults points..
and then 700 Tpts (Ouch) for the next position (12 yoyo positions 135+135=270)
 
I got the first GPU today and installed it inside an idle i5-750 system for GIMPS. It's the EVGA 560 Ti 448-cores and it's running at 380 GHz-d/day (stock clock). This is mfaktc 0.20 and it uses absolutely zero CPU. I requested 10 work units from www.gpu72.com "LL Trial Factoring Work". I'm also running their spider script every two hours via cron to automatically submit the results.

Info on the GPU72 project
 
I started reading through the GIMPS Wiki FAQ and saw the following:
Code: 
Does your team get credit when you manually check out work?

Test Results checked into PrimeNet using the Manual Test Results Check In form, are recorded for official GIMPS credit, but are not credited for CPU time by PrimeNet. This means the team will get the credit for them on the Gimps Top 100 page but not on the Primenet one.

I wonder how this affects DC-Vault points calculation. Any ideas?
 
brilong, I can't answer that one. It would require more research. Maybe one of the others know the answer.

Team, we never officially decided on April's push. POGS is still in the running. NFS is a pretty good choice and if we get our team mates to change their nicknames and contact the admins at Muon1 to merge the two accounts, we could also hammer away at Muon work through the BOINC client. However, for our Linux users, there is additional work they will have to do to get Muon work at YoYo to work under Linux. This thread explains how to get it going. http://www.rechenkraft.net/forum/viewtopic.php?f=57&t=10396 If anyone has other ideas, please chime in soon because we only have another day before our next push begins.
 
I am planning to run yoyo/muon on my amd cpus, and switch my gpus to gimps, and pogs on the intel laptop... Hmm, another week on pogs at our current rate should give us 4-5 positions (80-100 vault points) ...

And misfit seems to work Ok for queuing and uploading gimps results...
 
I think I'm going to finish a small push on Leiden to hopefully gain 2 more spots, then go for NFS. If POGS would be better overall I'm happy to switch over to that instead of NFS, though.
 
Stephen Brooks merged my Muon1 accounts (both were created by YoYo using BOINC). I went from 26th on the team to 14th. My BOINC Stats haven't changed by doing this either. So if anyone has ran Muon under YoYo or plans to do so, make sure to change your nick name to include [H] and email Stephen to have the names merged.
 
Dutch Power Cows will be focused on POGS for the month of April. They are currently 2 places behind us at POGS and they can put up some serious numbers once the stampede begins. Just an FYI for those looking to maximize our positioning.
 
ClimatePrediction is still having trouble with exporting their stats to some of the sites. I do not know if this has affected anyone at the DC-Vault yet. I do find it quite annoying. http://climateapps2.oerc.ox.ac.uk/cpdnboinc/forum_thread.php?id=7735&nowrap=true#48546 This is what the mod said about the issue on 3/26
The credits system is still broken and still isn't being worked on.

The new models, and others still being tested, are the priority.

Consider this project to be a credit free zone for now and some time into the future.
Click to expand...

Since then it seems to have been fixed on their site, but external stats exporting still isn't working fully.
 
Gilthanis said:
Stephen Brooks merged my Muon1 accounts (both were created by YoYo using BOINC). I went from 26th on the team to 14th. My BOINC Stats haven't changed by doing this either. So if anyone has ran Muon under YoYo or plans to do so, make sure to change your nick name to include [H] and email Stephen to have the names merged.
Click to expand...

I don't know by what stats the DC Vault is going by but I've noticed both yours and Nickoftime's Muon1 yoyo work credit is not showing up here.
 
You must log in or register to reply here.
Back
Top