• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

DC Vault 2

What setting do you have for "switch application every"... Set that to something high. Also, change the "While Processor use is less than to 0" and see if either of those do the trick. That should cover all the rest of the possibilities I think as far as your settings go. The last thing I would try is to reserve one core/thread to dedicate to each GPU in the machine. Theoretically a GPU could be stealing the CPU bumping the CPU app to the back seat.
 
"Switch application every" is currently 180 minutes, this machine is only attached to Cosmology, Collatz, Milkyway and WUProp so no other CPU tasks for it to switch to.
The CPU usage restriction is also set to 0, but I thought that was basically ignored if BOINC is set to run always?
BOINC is set to use 75% of my cores, leaving one free for my two GPUs and VMs - I don't really want to drop this to 50% as even if Cosmology errors out every few days and say doesn't do any work for a few hours, I'd lose less progress doing that than by dropping my processing power devoted to BOINC by 1/3, if that makes sense. Neither of my GPU projects ever take much CPU usage anyway, my rig generally sits around 80% usage so just around 5% is used to run both GPUs and my VMs.
I just can't see anything that could be pausing, there's nothing else running that would hog the processor unless a VM went crazy, but when I got home and checked on my rig the CPU usage was around 5% - absolutely normal as Cosmology wasn't running properly, so I don't think it was anything actually hogging the processor causing Cosmology to pause.
 
The only other thing I could think of off the top of my head is if the hard drive was either over taxed or starting to fail. I know that some hard drives pass all tests but still don't like CEP2 at WCG. However, that isn't a quick tweak to experiment with. Was it a SSD or traditional?
 
Gilthanis said:
Good job on Cosmology you guys. We have made some pretty good progress and still have a good amount of time to go.
Click to expand...

I swear that's all DooKey, he's killing everyone except a team Intel guy. No matter how many cores are allocated, I can't bring my average up by much!
 
Gilthanis said:
The only other thing I could think of off the top of my head is if the hard drive was either over taxed or starting to fail. I know that some hard drives pass all tests but still don't like CEP2 at WCG. However, that isn't a quick tweak to experiment with. Was it a SSD or traditional?
Click to expand...

Touche, but what would be stealing enough resources to stop BOINC? Perhaps some sort of drive indexing? I guess make sure there is no screen saver on, not even BOINC, as it can be seen as a non-BOINC process and can make resource usage go above the threshold. Hmmmmmm. Tis a challenging question. I shall keep looking for you.

This might help
 
SSD, running on an Intel X25-M which is all good, and there's nothing hammering the drive as my VMs are on a separate X25-E SSD. Just benched this drive and it all looks A-OK, no lockups and it's performing as expected. I don't even have any HDDs in systems around here any more apart from an external 2TB backup drive.

Edit: my screensaver is just blank, I can try setting it to "none" instead, maybe worth a shot. I would've expected it to maybe affect it every time though rather than just a few times in the last week or two. I'll have a read through that link Kai, thanks.
 
Yeah...with GPU crunching, I wouldn't bother with a screen saver. They have been known to cause issues at some projects. I don't remember specifics though. The SSD's shouldn't be having the IO issue then.

Kai, BOINC can thrash your HD depending on the app. CEP2 does a lot of disk writes. So I would not recommend it on a system with SSD's. The more the cores, the harder BOINC can be on a disk. If multiple WU's checkpoint at the same time, it can actually bog the system down enough to error the work units. Another reason WCG doesn't recommend running too many CEP2's on one machine. My i7 could run 8 at a time, but some machines can't do more than 1.

I'm at a loss right now on Captains issue. At least it has lessened.
 
I've changed the screensaver from "blank" to "none", will see how it goes.
It's not a hugely important issue as it hasn't stopped me from crunching, just a few errored WUs and a few hours lost on Cosmology so far is all, nothing disastrous. It's just very odd. I'll keep an eye on it after this latest change and see how it goes, but if it carries on happening then as long as it's only every week or so I can deal with that. If every unit was erroring out like before then it would be more of a problem but at least this system is still contributing like this.

I appreciate the help and suggestions guys.
 
Well, my i7 finally finished the Primegrid WUs, so my Cosmology ppd should double here any time... As long as I don't run out of memory! I started Asteroids back up, but at a lower Resource Share - that seems to keep at least one WU running and keeps the memory usage down a bit, but keeps my from leaving cores/threads from sitting idle!
 
Excellent! Welcome to the race! Now we have enough people to make a much larger difference in our vault position. Well, really, all we need is DooKey with his millions of cores crunching away for us.
 
Kai said:
Excellent! Welcome to the race! Now we have enough people to make a much larger difference in our vault position. Well, really, all we need is DooKey with his millions of cores crunching away for us.
Click to expand...

Naw, we need 100's of DooKeys with their millions of cores. :D
 
Gilthanis said:
Naw, we need 100's of DooKeys with their millions of cores. :D
Click to expand...

Better get your butts in gear then we lost 16 points today
The Lattice Project (Down 9.433963 points)
Docking@Home (Down 8.771930 points)
Loos like a couple of teams played slap [H] today :eek:
 
This happens when most of the team are currently doing a push or not active...

Then again, if points are the concern, the low Vault point reward projects like Rosetta and WCG never get any serious progress. We will bounce back in due time. Especially if we make POGS next months push. They are worth more than both of those projects right now.

With all of the BA talk going on....sounds like possibly a lot of folks jumping ship to WCG. Take a look at my new thread here by request: http://hardforum.com/showthread.php?t=1801908 It is still a work in progress, but I think many will find it useful.
 
Is there a good DC Vault for beginners post or wiki somewhere? It may be useful to those wandering around looking for project direction.

To be honest, while I understand the gist of the Vault, I never got the project weight part of it and what makes certain project monthly pushes more desirable than another project (other than easy overtakes).
 
Not really a single post. I will try and dig up the forum post over at Team Ninja's forum where they mention the equation they use for the point system. In layman's terms though, the less popular a project is, the more points it is worth. One of the reasons for this is to encourage more participation at other projects to help give them support.

If etothe3ric is still around, he could edit the first post in this thread with some of that info. I will try and write something up soon on it unless someone else beats me to it. I'm kinda in the middle of finishing up the All Inclusive DC Project List at the moment. :)
 
The vault itself is a magical place ran by elves utilizing unicorns to acquire the scores from each and every project. Flying unicorns of course. But seriously, I can't actually explain the Vault itself or how the scores are calculated or any of the particulars. But I can tell you that the pushes are largely based on what we feel we want to increase. It's usually suggested and either voted upon or we simply add up the highest amount of suggestions and take the one that's suggested the most. There are also times when a push is based on a certain cause that has a personal element. In summary, the vault is difficult to decipher, but we love science here!
 
And some people don't bother with the push and do their own thing. You really can't go wrong because as long as your attached to the project with our team, the contribution continues even if you didn't know you were helping. Like FAH is part of the Vault. Everyone who folded for Team33 contributed to those Vault points or maintaining those points.
 
I can dig up the Vault formula particulars tomorrow when I get to work. As an engineer, I a while back created a spreadsheet (of course) that could tell me how many points we would make by jumping a certain number of places. It's what I used to use when helping plan our monthly pushes before all the hardware failure forced me to take a break...

But for the "short" version from memory - take the number of teams the Vault sees in a project (usually, but not always the same as Free-DC, never understood why) and divide that number into 10,000. That's how the points per place are divided up within a project. So a project with only 30 members, like Prime Sierpinski Problem - PRP, would be 10,000/30, or 333.33 points per place. #1 gets 10,000, #2 gets 10,000 - 333, or 9,667, etc. A project with 90,000 or so teams (I'm guessing here) tracked in the Vault, like F@H, would be 10,000/90,000, or only 0.11 points per place. There, #1 (us, of course) gets the 10,000, but #2 (evga) gets 10,000 - 0.11, or 9,999.89, etc.

This is why it's important the pushes look at three key points to maximize our Vault scores:
  1. How many other teams are there? The less teams, the more points for each spot we move up, and vice versa.
  2. How active are the teams above us? Projects where everyone above us are fairly active means we may not gain many (or any) spots, even with a month long push.
  3. How far is the gap to the next few teams? Projects with huge gaps minimize the number of spaces we can move up. That's okay if the ranks are 50-100 points per rank we move up, but if if a single rank isn't worth much, we're better off elsewhere.
A good example of a project that a push would do no good because of #1 is WCG - we already have a score of 9,980 there, even if we're only ranked #60 or so. Each place is only worth 0.33 points, so the work to move up isn't worth a push. For #2, SIMAP would be a pretty good example. Assuming we put 125K ppd up, like we're doing now in Cosmology, we could do just under 4M points in a month. We currently have 4.8M points, so we'd end a push at about 8.5M or so. that could move us up almost 20 places, which is about 4 points per place, or 80 Vault points total for the month. But of those 19 teams currently below 8.5M, only 8 are at 1,000 ppd or less. They other 11 are active (although none are near 125K ppd), but a few would stay out of reach, and a few more would pass us back again as soon as the push was done. I'll skip #3, as you're already likely asleep or TL;DR...

See? Simple and short, right? :D

- Professor Jathanis, BS in BS
 
Gilthanis said:
And some people don't bother with the push and do their own thing. You really can't go wrong because as long as your attached to the project with our team, the contribution continues even if you didn't know you were helping. Like FAH is part of the Vault. Everyone who folded for Team33 contributed to those Vault points or maintaining those points.
Click to expand...

And those rogues that have favorite projects and stay with them after we've moved on to the next month also keep us from slipping as fast when a push is over...
 
Not Asleep;Did Read. Thanks guys, it makes sense now.
 
Thanks Jathanis for the explanation. It was well done. I also appreciate not having to dig up the info again. Maybe we should make you the DC-Vault rep. :)
 
KMac said:
Is there a good DC Vault for beginners post or wiki somewhere? It may be useful to those wandering around looking for project direction.

To be honest, while I understand the gist of the Vault, I never got the project weight part of it and what makes certain project monthly pushes more desirable than another project (other than easy overtakes).
Click to expand...

KMac, phoenicis also gave us this nifty link many pages back that is also really helpful for planning. It at least lets you know the values of taking each spot at the Vault projects.

http://digik-oz.nl/dcVaultStats.aspx?team=[H]ardOCP

I noticed that they didn't list Asteroids as having a GPU app, but they do. It just really isn't worth running on your GPU's since there isn't really much gain due to how the app is written and it uses typical GPU power consumption while it does.
 
I can dig up the Vault formula particulars tomorrow when I get to work. As an engineer, I a while back created a spreadsheet (of course) that could tell me how many points we would make by jumping a certain number of places. It's what I used to use when helping plan our monthly pushes before all the hardware failure forced me to take a break...
Click to expand...

Since I was already browsing, I found the thread anyways. :D

http://www.team-ninja.com/vbulletin/showthread.php?t=36977
 
You gotta love when Windows updates restart your system and thus shuts down your VM's...slow production day in DIMES for me.
 
Windows bugged me for activation on my DIMES VMs after I had to reboot my machine, didn't realise and lost half a day's work - boo!
I threw another quad on Cosmology, just waiting on some more cases to fire up another quad and an IB i3, plus a GTX 460 and possibly another quad but at least a dual. Apparently nekkid systems look untidy or something :confused: hence the need for cases
Pulled my 5770 off Collatz as it's needed for something else at the minute, is the GTX 460 any good for Collatz or is it best suited to something else? Haven't run an Nvidia card on any BOINC projects yet. Oh, is an 8400GS worth running at all, on anything? It's currently just my test card as it's the only thing I have that will display on the only monitor I have (needs dual link DVI) but if it'll crunch anything worthwhile it can go in one of these systems.
 
I would try the 460 and any other nVidia card at GPUGrid and see how they perform. Mid and low end cards should be set for short run work. It is a bio project after all. I would also set them up to run POEM (at least once it has GPU work again) because it is the other bio science that supports them.

If you don't have a preference on the science, GPUGrid and POEM (once they have work again) pay out points quite well. Milkyway could use the double precision capabilities of the card. SETI is terrible point payout. Einstein's points aren't the greatest either. Collatz would work fine. (last I knew this project awarded AMD cards best)

The 8400GS is worth running if you compare it to CPU's. If you have the spare PCIe slot, let it run. But certainly don't invest money buying those cards. I don't think the 8400GS will run GPUGrid any more. I know it still did about a year ago even though it was recommended. Jathanis recently fired up a few 210's I sent him (those are rebadged 8400's) and noticed they were about on par with some modern processors. Since they consume less electricity than many processors, they are still efficient enough to use. But again, only if you don't have better and only if you already own them. Certainly there are better bang for your buck out there. I am to the point where I'm putting my 8400's into borged boxes I donate or systems I give a free upgrade too...
Also, some 8400's (there are 3 types by the way) have 8 CUDA cores and some have 16 like the 210's do. This can also make a performance difference.

8400
8400GS
8400 GS
Yes that space makes a difference.
 
Last edited:
Yeah, that's the situation - I have it sat here, just used as a test card, I wasn't intending on buying it. If it makes sense from a power/performance perspective, I'll stick it in a spare slot, and if/when I pick up something better I'll swap it out. I'm unsure which rev this is as it's not in a system at present and it has no info on the card itself, but I saw there are some with 8 and some with 16 cores. I'll have to check next time I use it.

I'm keeping an eye out for any good deals that come up locally or on the FS forums I frequent, for any GPUs - my thoughts being that if I'm running some quad core systems anyway, and if I can pick up some half decent GPUs at a good price, they're worth throwing in where possible. Not making it a priority, just watching and waiting.

Thanks for the info on the 460, I have no preference on the science just whatever helps our Vault ranking, so I'll throw it on GPUGrid and POEM when I have it up and running. I run Milkyway on my 5850 and it's doing pretty nicely there. I think I have a 9800GT around here somewhere, will that do GPUGrid and POEM too?
 
http://www.gpugrid.net/forum_thread.php?id=1150#10607

Captain, that is the place to check I would assume. I know the card worked there at one time, but do not know if it still does. I would give it a shot and see. Since POEM does not have GPU work right now, I can't contest on what will work at this time. Previous work units supported most GPU's.
 
Last edited:
From my experience, yes on POEM and no on GPUGrid - I have a 9600GSO and a 9800GTX+ that both work on POEM just fine - about 50-60K ppd for the 98 and about 15-20K for the 96 (compared to 250K ppd for my GTX680). As for GPUGrid, according to their forums last November:

Recommended:
GTX690, GTX780Ti, Titan, GTX780, GTX770, GTX680, GTX670, GTX660Ti, GTX660, GTX650Ti Boost, GTX650Ti, 2GB versions of the GTX650, GT640 (Kepler) and GT630Rev.2.

Still works reasonably well:
Geforce GTX 590, 580, 570, 480, 560 Ti (448), 470, 465, 560Ti, 460, 550Ti
Tesla* C2050/C2070 (Workstation), M2050/M2070, S2050 (Data Center)
Quadro Desktop* Quadro 6000, 5000, 4000, 2000, 2000D

Should still work but are now inefficient (slow, power inefficient and more likely to fail):
Geforce GTX 295, 285, 280, 275, 260-216 (55nm), Tesla10, Tesla20, 460SE, GTS450, Quadro 600, 480M (CC2.0)*, Quadro Mobiles; 5010M, 5000, 4000, 3000, 2000

Recommended Desktop OEM Cards:
GTX760Ti, GTX760, GeForce GTX 660 (OEM) - Two versions 1152:96:32 (256bit bus) and the lesser 1152:96:24 (192bit bus), GTX 645 (GK106) 576:48:16, GT 640 (GK107)
OEM GTX cards tend to be mid-range and GT cards are entry level cards, however the 760Ti and 760 are high end.

GPUGrid is now CC1.3 or higher (CC3.0, CC2.1 or CC2.0). I would add that most 65nm versions of CC1.3 cards will probably not work.​

Linky: http://www.gpugrid.net/forum_thread.php?id=2507
Note what I put in bold about CC version - this is why the old 9XXX series cards don't work any more.
 
Thank you for the list! It'll be helpful to get a farm set up correctly. So, POEM push next month?
 
I'm guessing POGS would probably be the most points... POEM hasn't started testing their new GPU app yet. Probably another 10 days or so and that will be on a test server. So, probably even longer. CPU work wont get us very far in the Vault.
 
POGS units seem to be going quite fast lately, I agree with that statement. They would probably be a good project to go after. Asteroids needs to fix their gpu app for it to be viable. I'd think a lot of teams will make a push on that once that app is more efficient.
 
I don't think they are all too concerned about the GPU app. I think it was given as an option for those that don't support other projects or don't care about points. It in essence acts like having an extra core. Very inefficient yes, but that matters to some. Right now I allow it to run on my GPU's so that I can keep an eye on the progress and to help get the project to my 1 million point goal. :)
 
We gained 23 vault points in POGS today but it hasn't updated yet. POGS is out of work until Thursday due to upgrades.
 
All right, due to the fact that Cosmology has been giving me more issues than it's worth due to my low memory issues, and that I'm a day's worth of WUs away from a nice even 250K points, I'm going to head over to POGS over the next few days and catch up with the rest of you guys so I'm running full speed ahead for if/when the February push officially starts. And I sure like those 30MB WUs after so many 400-600 MB varieties!

Oh, and not sure why, but all three of my machines got a load or POGS today - maybe the upgrades got finished early? Server status page currently shows 70K+ WUs ready to send...
 
You must log in or register to reply here.
Back
Top