BOINC@AUSTRALIA FORUM

Active BOINC projects => EINSTEIN@HOME => Topic started by: Dingo on May 31, 2018, 04:03:50 PM

Title: Project Update
Post by: Dingo on May 31, 2018, 04:03:50 PM
We are in the higheest position in the world that we have ever been.  Well done Team  https://boincstats.com/en/stats/5/team/detail/73 (https://boincstats.com/en/stats/5/team/detail/73)
Title: Re: Project Update
Post by: shift on May 31, 2018, 04:27:13 PM
 :AUS: :rocks :AUS:
Title: Re: Project Update
Post by: JugNut on June 01, 2018, 01:18:40 AM
We are in the higheest position in the world that we have ever been.  Well done Team  https://boincstats.com/en/stats/5/team/detail/73 (https://boincstats.com/en/stats/5/team/detail/73)

Ooh yea!! that's the way ah-ha ah-ha we like it,  ah-ha ah-ha :dance2:
Title: Re: Project Update
Post by: Dataman on June 01, 2018, 01:32:44 AM
We will, we will rock you.
 :rocks
Title: Re: Project Update
Post by: Dataman on June 02, 2018, 05:14:55 AM
GDPR FYI
https://einsteinathome.org/content/upcoming-changes-due-european-general-data-protection-regulation-gdpr
Title: Re: Project Update
Post by: zzrhardy on June 02, 2018, 10:28:03 AM
We are 3rd in RAC at the moment according to https://einsteinathome.org/community/stats  , though I imagine SETI.Germany will mobilise soon as they are very competitive.

In total credit we have overtaken Russia Team for 15th place.
Title: Re: Project Update
Post by: chooka03 on June 02, 2018, 02:06:26 PM
In regards to my comment on the Einstein forum Jugnut, I'm actually undecided if I would give BOINC away entirely. I'm leaning toward the "would give it away" rather than "wouldn't" but I wanted to make a statement on the forum.
I would certainly save a lot of money by giving away this hobby both in electricity usage and machine equipment but I would dearly miss the regular chat on this forum  :thumbdown:

I hope the stats can be maintained. It would be such a shame that these new laws could royally screw this volunteer organisation. It would go down in history with the old adage... "Back in my day.... they didn't have these stupid laws."... etc etc.

There's actually quite a few old guys I talk to these days who just say they are happy to be at the end of their working lives.... or life in general really. The world's gone mad and it seems to be getting worse!

 :oz:
Title: Re: Project Update
Post by: chooka03 on June 02, 2018, 02:24:43 PM
Hey just a question...when you view your computers in your account and you select "tasks" and view say the valid tasks........ Is the run time shown for each task regardless of how many WU's you were running consecutively? Or do I divide it by 3 (because I'm running 3 WU's concurrently.)

Vega = 960/3 = 320/60 = 5.3Min/WU ?

I should know this by now hey lol  Bashhead

If this is correct, the RX480 is much better than a Vega...... unless I'm better to crunch 2 WU's instead of 3 which is what the RX is doing.

RX480 = 555/2 =277.5/60 = 4.62min/WU.

I'm 99% sure I've got this right but thought I'd double check. Once upon a time when I was testing 0.5 vs 0.33 for Vega WU's, I'm sure there was a speed up running 3 vs 2 WU's. Then again, the work units have likely changed a bit since then.
I might just try it.



Title: Re: Project Update
Post by: Dataman on June 02, 2018, 02:28:41 PM
Is the run time shown for each task regardless of how many WU's you were running consecutively?

Yes.
Title: Re: Project Update
Post by: chooka03 on June 02, 2018, 03:26:16 PM
Thanks Dataman.

So I tried 2 WU's concurrently with the Vega. Basically 3 is better.

16min for 3 WU = 5.3min/wu
11min for 2 WU = 5.5min/wu

Probably not the best time to be doing this during the AA  :crazy

I wonder what I can squeeze outa the RX480?

If these numbers come up right (and its very early days), I'd be better to sell my Vega for a profit and buy another RX480 or 580. An R9 280x is a cheaper card still but very power hungry compared to the RX series.

Given the Vega 56 outclasses the RX480 in also every way (except speed) I'm not sure what to make of these numbers.
Title: Re: Project Update
Post by: kashi on June 02, 2018, 06:13:56 PM
Anyone else noticed a large variation in the runtime of tasks? My GTX 970 is taking about 50% longer than during a little pre AA trial I did 12 days ago. About 20 minutes a task now and previously it was 13-14 minutes. 20 minutes per task is about 250K per day, on the stingy side of things compared to some other GPU projects. Still suppose the rate applies to everyone at least. Remember years ago Einstein used to have a beta cuda app that was more generous.

I've left 2 CPU threads free, I'll try turning off all Universe CPU tasks for a while to see if it's related to GPU processing being throttled by lack of CPU resources. I know GPUGrid is quite sensitive to that, I only run 4 CPU tasks when running GPUGrid. Also now running those Universe CPU tasks on a Linux VM client, I know this can sometimes slow down GPU processing much more than running the same number of CPU tasks on the Windows host.

Edit: Nope, makes no difference, suspended all CPU tasks for 2 Einstein tasks and they took the same time. Ah well, that's it then, pedalling as fast as I can on these current tasks.
Title: Re: Project Update
Post by: zzrhardy on June 02, 2018, 09:07:54 PM
I belatedly realised I was having issues on Kuroneko with the Vega 64 liquid cooled cards also. The primary one was boosting to 1700+MHz but the other two were idling at about 600MHz. So some work units were averaging 5min and the others 8-9min. I fired up Afterburner and set Core Clock to the base 1406MHz and now all V64 cores are running between 1500 and 1450MHz.

This means WU now just take about 18min across the board running 4WU per card. So about 4:30min per WU/per card -  or a tad higher which is a much better figure.

Another issue cropped up on Chibineko too, it was prioritising CPU tasks over GPU tasks so one of the Titan V's had been sitting completely idle for goodness knows how long. Have turned off CPU work units in the project preferences because they aren't worth idling a Titan V over. PS - the Titan V's are barely faster than the Vega 64's (at these Gamma-ray pulsar binary work units at least).

So all up - hopefully things will be a bit more productive from now on.

Edit: after more experimentation the V64's handle core freq set to 1600MHz in Afterburner without any issue. This is well above base freq but still below max boost freq.
Title: Re: Project Update
Post by: JugNut on June 02, 2018, 10:02:02 PM
@ Kashi: Yea it certainly seems that way.  For me they appear to be about by 25% longer but it's hard to say? All I know is I can't get anywhere near the credit I was getting last time around.  And that's using the exact same hardware too.   Perhaps they've added additional features that require extra processing? 
With a bit of luck these are just some temporary long batches that will return to normal at some point in the future? 
:fingers 
Although that's not the way things usually work,  so it's probably just wishful thinking.



@ zzrhardy: It's good to see you finally got your problems sorted and are now  firing on all cylinders. :thumbsup:
Title: Re: Project Update
Post by: chooka03 on June 02, 2018, 10:16:18 PM
I mentioned the longer run times last week and this is what Gary Roberts replied with -

"If you'd like to know the reason why 220-240 secs has now become 316 secs, you might like to peruse this thread, particularly the recent stuff near the end of the thread.  In the last month, there have been two significant changes in the nature of the data.  The first change, almost a month ago resulted in a significant reduction in the elapsed time to process a task (crunch time).  That has now been reversed, and a bit more for good measure, so it seems :-).  The latest tasks are now taking longer than what was the norm before the speedup earlier in this month.

The longer crunch time is a good thing.  Should take a bit of load off the servers.  They mightn't be quite so frantic in distributing tasks and receiving results :-)"

This was the thread he was referencing - https://einsteinathome.org/content/new-app-or-new-data

I was interested how different a Vega 64 was to my 56. Thanks ZZHardy  :thumbsup:
Title: Re: Project Update
Post by: JugNut on June 02, 2018, 10:40:04 PM
Thanks for the info Chooka you saved me a search. :thumbsup:  + karma



PS Can anyone access boinc-stats?

EDIT: I can reach it now.  Thanks anyway...
Title: Re: Project Update
Post by: kashi on June 03, 2018, 10:18:02 AM
Thank you both for the replies. Firstly JugNut for letting me know it wasn't just my card acting up ornery and secondly chooka for the explanation.

Hmm, well this "good thing" is about as welcome as stepping in dog poo. "Take a bit of load off the server" No thanks, I'd prefer to push the server to its limits with tiny tasks and gigantor credit for me, ha.

 
:boom:
Title: Re: Project Update
Post by: JugNut on June 03, 2018, 02:02:56 PM
Thank you both for the replies. Firstly JugNut for letting me know it wasn't just my card acting up ornery and secondly chooka for the explanation.

Hmm, well this "good thing" is about as welcome as stepping in dog poo. "Take a bit of load off the server" No thanks, I'd prefer to push the server to its limits with tiny tasks and gigantor credit for me, ha.

 
:boom:

LOL Spot on kashi,  I couldn't agree more...

Actually I don't think anyone would have cared as long as they increased credit in line with the extra runtime, that way they still get the benefit of less server load and we don't lose out because of it.


EDIT: Another problem i'm having is invalids,  i'm getting a lot of them.  While getting invalids is a normal part of doing Einstein i'm getting way more than usual. Since restarting Einstein only 4 or 5 days ago i've already got 49  across all my boxes. If you work that out over the year that's a shed load of wasted time & power with no benefit to anyone.  I'll see if there's anything more that can be done after the AA.

EDIT 2: Now 52 invalids & rising :(
Title: Re: Project Update
Post by: kashi on June 03, 2018, 07:49:12 PM
That's strange, I haven't had a single invalid yet in 144 tasks completed. Yet you've had them on 4 different kinds of cards with 4 different driver versions. And one of them is the same as my 382.53.

So what is it? A bad batch that I evaded by starting late? Windows 7 for you on all boxes whereas Windows 8.1 for me? Perhaps you're running multiple concurrent tasks, that sometimes causes trouble. Different number of CPU cores left free to support GPU crunching? I'm leaving 2 cores free out of 8 to support a single Einstein GPU task. It's all very mysterious.

Hope I haven't mozzed myself and I start getting invalids too. :pcwhack:
Title: Re: Project Update
Post by: Sean on June 03, 2018, 10:33:32 PM
I've only had one invalid (out of 157). Im on windows 7, running two concurrent tasks on a 1050ti with two dedicated CPU cores.

Took me ages to find the link for Tasks on the Einstein@home page, its just below the list for computers in the account tab.  :jester:
Title: Re: Project Update
Post by: Dataman on June 03, 2018, 11:24:38 PM

EDIT: Another problem i'm having is invalids,  i'm getting a lot of them.  While getting invalids is a normal part of doing Einstein i'm getting way more than usual. Since restarting Einstein only 4 or 5 days ago i've already got 49  across all my boxes. If you work that out over the year that's a shed load of wasted time & power with no benefit to anyone.  I'll see if there's anything more that can be done after the AA.

EDIT 2: Now 52 invalids & rising :(
Yes, invalids have always been a problem at E@H but they seem to be increasing. I have 55 out of 5K tasks.  :thumbdown:
Title: Re: Project Update
Post by: zzrhardy on June 03, 2018, 11:33:20 PM
I have 20 work units during the AA period marked as either "validate error" or "completed, marked as invalid" and another 68 from the ramp up period. Titan Vs seems to have a worse ratio than Vega 64 LCs.

Chibineko (Titan V): 1753WU valid and 51 invalid
Kuroneko (Vega64): 2589WU valid and 37 invalid
Title: Re: Project Update
Post by: Dingo on June 03, 2018, 11:54:24 PM
I only have one GPU (1080 Ti) in my PC at home (https://einsteinathome.org/host/12593401) and I am running two tasks at a time.  I have 946 completed and validated, 15 Completed but marked as invalid and only 3 Error while computing.  I can't complain about that.  I did update my Nvidio driver to the latest (NVIDIA GeForce GTX 1080 Ti (4095MB) driver: 397.93) before the AA and I have Windows 10 updated to the latest release.

I tried to use my daughters PC  (https://einsteinathome.org/host/12599702)which has an ATI 5870 I think and 2 out of three have had errors.

Title: Re: Project Update
Post by: kashi on June 04, 2018, 12:19:06 AM
....Took me ages to find the link for Tasks on the Einstein@home page, its just below the list for computers in the account tab.  :jester:

Yes, that's the better way to look at your tasks. Clicking the "Your results" box link in BOINC Manager takes you to a Tasks page that is so lacking in information as to be basically useless. What were they thinking when they designed that, it's totally ludicrous.

At least the one accessed via the Account>Dashboard shows the number of tasks in each category. It would be annoyingly short on detail though for anyone that crunched more than one application.

The horrid reduction in available crunching information on the eviscerated Einstein website is one of the reasons why I don't crunch it ever now except for AAs or if someone had called a "mini aa". For comparative hardware performance details used by anyone enthusiastic about maximising their contribution it's almost as execrable as WCG now and hasn't even got badges to compensate. 
Title: Re: Project Update
Post by: JugNut on June 04, 2018, 01:32:31 AM
Agreed Einsteins website redo was/is a disaster. It's only a improvement for mobile phone users.

Back to invalid situation, it's certain that the more WU's you run in parallel  the more invalids you'll get.  I've also noted that the more video cards you use per box can also affect the amount of invalids you get.(On top of what the extra cards would produce on their own)
But up till now the trade offs were well worth it. Up until now it has always been more efficient to run the extra WU's anyway as even with the slightly higher amount of invalids you still came out well ahead of running just one at a time.  But it's now very close to the point where thats is no longer true.  What's changed? Who knows?  Whatevers changed I seem unable to fix.  It would help if I knew what the problem was but with invalids there's no error number to guide your way, it just trial & error(no pun intended).   I'm sure I could reduce the amount of invalids by going back to running one at a time, but as things stand now it would still slash my over all output/credit but if this problem persists i'll soon have no other choice.

I done the usual things one does in these situations such as lower memory clocks, then core clocks,  add/remove voltage ect ect but unsurprisingly it didn't work.  I say unsurprisingly because it's happening on all my boxes so it would be some coincidence if all my cards started failing all at once,  but I guess anything's possible.   

What do you guys think? After all in the last 5 days I completed 4,300+ WU's,  of that 1059 are pending with 53 invalid.  Although at this rate it looks like my invalids will reach triple figures by the end of the AA. Overall my invalids have increased by 75% from the last time I crunched Einstein which was just under two months ago.

Any more thoughts on the subject would be kindly appreciated.

Crunch-on  :crazy
Title: Re: Project Update
Post by: Dataman on June 04, 2018, 05:03:22 AM
Ditto here. I am annoyed with my performance. With 13 cards running I thought I could do ~4+ million/day ... I am not even close. I have not been able to OC the cards at Einstein for a long time but now it seems worse. I have ~ a 1% invalid rate.

Ditto as well on their page layout. One needs a friggin' map to find the simplest things.  :thumbdown:

And for my final whine of the day, the heat is back! Send us more clouds please.  :rofl:

 :cry
Title: Re: Project Update
Post by: kashi on June 04, 2018, 12:16:19 PM
Yes invalids are annoying and much more so on power hungry GPUs. Remember one of the previous beta applications started giving me invalids or errors once which was frustrating because the standard application was stable but had a poorer credit rate. Folding had a big problem with invalids on Maxwell class GPUs which stopped me crunching there completely.

Website layout is an initial annoyance but mostly overcome with familiarity. It's the loss of functionality that is deplorable. I mean the main reason to go to a project's website is to see your tasks/statistics and compare performance of your hardware to others. But the Top Computers list only shows the first 50 hosts making it useless for anyone that doesn't have one of the highest performing cards. Is it worth firing up my old 7790 box or whacking a 5870 in the dual box for the AA? I'll just look it up and see. Aarrgh, no I won't, the essential info is missing. Oh well, better just give it a miss then, they're probably poor Einstein performers anyway.

Wonder if all the projects will remove hosts info due to that EU privacy thing? That would be almost enough to put the kybosh on BOINC for me. Think I'm already stuffed at WCG because their new terms require a current email address which they won't get from me. If I change my old BOINC email my CPID will become split, totally rooting my stats.

Ahh, that's better, now that my whinge is clarified and expanded, how is your pending going? Mine's currently at 42% of completed tasks which is not too bad for Einstein I suppose as traditionally it's a very high pending rate project. Wondered why my first full day's crunching yielded much less than I expected/calculated, because I'd forgotten about Einstein's pending.

On a positive note, I'm quite enjoying the extra warmth from GPU crunching now the nights are cooler. Very cosy in my bedroom now and only a small increase in noise.
Title: Re: Project Update
Post by: Dataman on June 04, 2018, 12:49:38 PM


Wonder if all the projects will remove hosts info due to that EU privacy thing? That would be almost enough to put the kybosh on BOINC for me. Think I'm already stuffed at WCG because their new terms require a current email address which they won't get from me. If I change my old BOINC email my CPID will become split, totally rooting my stats.


On a positive note, I'm quite enjoying the extra warmth from GPU crunching now the nights are cooler. Very cosy in my bedroom now and only a small increase in noise.
I blacklisted WCG today. Until now it only contained Russian projects. I also went in my profile there and checked "Hide my info" and my info disappeared from the forum. I am done there.

Kashi, I will trade my extra warmth for your cool! I will pay for the shipping.  :rofl: Was 36C here today. GPU temps are reasonable (~56C) but CPU temps on some of mine are above 80C. I hope I do not have to shut down cores to get the temps down.  :thumbdown: It is late evening here so no worries until tomorrow afternoon. Crunch onward ...

 :greet
Title: Re: Project Update
Post by: tazzduke on June 04, 2018, 10:42:12 PM
Greetings All

Think we need to take a back step and have a good a proper look at what this is all about.

As per the following link from Einstein, its all about anonymous access to their stats.

https://einsteinathome.org/content/upcoming-changes-due-european-general-data-protection-regulation-gdpr (https://einsteinathome.org/content/upcoming-changes-due-european-general-data-protection-regulation-gdpr)

Also from Willy over at Boincstats via this link https://boincstats.com/en/forum/5/11967,1 (https://boincstats.com/en/forum/5/11967,1)

There is some legal stuff to take care of which takes some time. After that BOINCstats will be able to update WCG stats again. Until that's done, no updates.

So all in all this really affects the owners of the stat sites more than us - Legal mumbo jumbo.

To tell you the truth, it will not stop me from crunching.

They only thing that will stop me from crunching is when the power goes out lol.

Cheers  :holy-moly :niceday :golf :panic: :rocks
Title: Re: Project Update
Post by: chooka03 on November 07, 2018, 06:01:24 PM
LOL.
I just noticed that I've moved into 4th highest average Einstein credits. My 2 X V56's have overtaken Armin Burkhardt speaking for MPI/FKF's 4 x GTX1080's.     :BigGrin

https://einsteinathome.org/community/stats
Title: Re: Project Update
Post by: Dingo on November 07, 2018, 06:38:18 PM
 :BigGrin  well done chooka  :party: :party1: :party:
Title: Re: Project Update
Post by: Dataman on November 30, 2018, 02:50:22 AM
READ THIS

https://einsteinathome.org/content/gdpr-new-website-features-and-statistics-export-option

You must signify that you agree before you can access your account. You must ALSO go into your security settings and change the new setting to allow export to stat's services. The default is NO. If you participate in Albert (their test project) you must go in and do likewise there.

EDIT: Per Einstein:
"Update: 2018-11-29 - The changes to the Web site have been rolled-out. The statistics export option will go in effect December 17th."
Title: Re: Project Update
Post by: JugNut on November 30, 2018, 11:45:53 AM
READ THIS

https://einsteinathome.org/content/gdpr-new-website-features-and-statistics-export-option

You must signify that you agree before you can access your account. You must ALSO go into your security settings and change the new setting to allow export to stat's services. The default is NO. If you participate in Albert (their test project) you must go in and do likewise there.

EDIT: Per Einstein:
"Update: 2018-11-29 - The changes to the Web site have been rolled-out. The statistics export option will go in effect December 17th."



Yep I couldn't read the post you linked to or even go to my account page without agreeing to Einstein's new terms&conditions. LOL 

I wonder what Einstein's stats will look like after this all kicks in?
Title: Re: Project Update
Post by: chooka03 on December 20, 2018, 07:20:28 AM
WOW....I can only assume the export data thing has kicked in as our team numbers on BOINCstats just shrunk dramatically!
My guess is anyone who didn't change their settings to allow stat exports now no longer shows up on BOINCstats.

https://boincstats.com/en/stats/5/user/list/0/0/73/#3        <---- I think that will work unless it's specific to my account?
Title: Re: Project Update
Post by: Dingo on December 20, 2018, 01:43:48 PM
Yes I think the majority of users attch to a project and never change thir preferences again.  It looks as if we are down to 29 users in Einstein. Acording to Formula BOINC our total credits were 5,254,430,782 yesterday and BoincStats has maintained the total, as of now anyway.
Title: Re: Project Update
Post by: chooka03 on December 20, 2018, 02:03:42 PM
Oh I didn't even think about Formula Boinc!
Gee that will be interesting.
Title: Re: Project Update
Post by: JugNut on December 20, 2018, 02:18:28 PM
Oh yea me too. Sheesh all that work we put in. *shudders in fear*

@Dingo: Maybe create a subject about the issue in the news section. That way over time when team members drop by they'll hopefully see the news and change their settings accordingly.
I guess you could also email team members?


Title: Re: Project Update
Post by: Dingo on December 20, 2018, 04:34:51 PM
No good sending an email as I only get email address from those that tick the box.  I tried to send a Team Message on Einstein but the web page is broken.  I have put in a post on there Message Board  (https://einsteinathome.org/content/web-page-error-send-message-team)about the problem.

On a side note LHC@HOME only shows 2 users.  I will send a message to all the projects on the list
Title: Re: Project Update
Post by: chooka03 on December 21, 2018, 10:11:55 AM
No good sending an email as I only get email address from those that tick the box.  I tried to send a Team Message on Einstein but the web page is broken.  I have put in a post on there Message Board  (https://einsteinathome.org/content/web-page-error-send-message-team)about the problem.

On a side note LHC@HOME only shows 2 users.  I will send a message to all the projects on the list

LHC only shows 2 users?
Nothing's changed for me. I can still see all my rankings.

Haha on WCG I've gained 213 places in one day. That will help my terrible global ranking!  :wink
Title: Re: Project Update
Post by: Dingo on December 21, 2018, 01:28:57 PM
No good sending an email as I only get email address from those that tick the box.  I tried to send a Team Message on Einstein but the web page is broken.  I have put in a post on there Message Board  (https://einsteinathome.org/content/web-page-error-send-message-team)about the problem.

On a side note LHC@HOME only shows 2 users.  I will send a message to all the projects on the list

I got a reply from the Admins to basically say there is no way to send an email till it is fixed sometime next year.  It is good thay the Admin did get back to me and I can see that there explanation looks feasible.

Quote
I'm afraid something is broken (software bug) on our end. The problem is that the URL for generating a direct message (DM) uses people's IDs that are comma separated: e.g., userids=1,2,3,4,5, etc. With 1300+ users, that URL is just too long for the Web server to handle.

In fact this "Send Private Message to Team members" functionality needs some attention. I'm sorry to say this won't be looked at into the new year. Until then I don't have a good solution for you.

You can contact members individually by going to the Member list, clicking on a username, and sending message. Obviously its impractical to do this for your team.

Einstein@Home Project