Posts by Neo
log in
1) Message boards : Number crunching : Too long calculation for the Sierpinski/Riesel Bases - long (Message 795)
Posted 29 Jan 2015 by Neo


plain simple?

yes - if you call for floating-point performance.

integer speed is just another thing.


and about "K10-based" - those things lack that big L3-cache - pretty ugly....



That's a huge part of it. If you have a AMD cpu that is newer than the Phenom II, it only has half the number of FPU's as it does cores. Try running BOINC at 50% cpu usage. That should dramatically help your crunch times.

I remember when the "Bulldozer" first came out... about 2010-2011?? I went out and bought the new CPU and a new mobo.. only to discover that it was 200% slower than my Phenom II 555....

Neo
AtP
2) Message boards : News : releasing primes (Message 785)
Posted 28 Jan 2015 by Neo
Rebirther,

I would rather crunch w/u's that ONLY have the possibility of being in the TOP 5000 primes list... Albeit, I'm glad to crunch whatever...

The website has come a long way since it's start!

Neo
AtP
3) Message boards : Number crunching : Error on long - but it finished the llr (Message 654)
Posted 16 Jan 2015 by Neo
MAJOR UPDATE: :)

Ok... got home from work and still no SRBase w/u's downloaded....

So, I reset my modem and router, and bingo! I was able to download SRbase work units...

I have "Dishnet" or "hughesnet" (satellite internet) because I live so far away from civilization...

I hope this information helps others.. I don't know for the life of me why that makes a difference, but it apparently did. After reboot of both, I was immediately able to start crunching SRbase w/u's. :) Go me!

Go AtP!

Neo
4) Message boards : Number crunching : Error on long - but it finished the llr (Message 641)
Posted 15 Jan 2015 by Neo
This sounds similar to a problem that's been observed at PrimeGrid and is actually a problem that will affect ALL BOINC projects.

Do you use AVG antivirus? If so, you must DISABLE its "Identify theft" protection. This part of AVG intermittently thinks the network communications with a BOINC server is suspicious and blocks it, resulting in transient HTTP errors. You can't block this by directory; you have to completely disable this feature.



Thanks guys for your responses...

I am not using a proxy. I have no problem receiving workunits from SRBase nor Primegrid, and I've crunched a fair number of SRbase workunits without problem, before and after Rebirther started keeping stats.

I'm not using AVG. I use Microsoft Security Essentials.

The only thing I can think of is that I upgraded BOINC to the newest version on the rig that had these two errors. However, I've returned a decent number of w/u's to SRbase after this upgrade... I only upgraded to get a more accurate estimated time of completion..

I do have a modified config file but the only thing I did was set BOINC to use my two GPUS... nothing else. I got my instructions on how to do that from Gary Craig.

With respect to those two workunits (which were LONG Btw ;) ... BOINC wouldn't let me abort them to clear them out of BOINC... I had to "reset project" for SRBase to get BOINC to clear them out.

A bigger issue is that I have set my preferences for "Resource Shares" to 99% for SRBase and 1% for Primegrid. (this was back when Rebirther was having trouble keeping the server full of work, so I figured I would get all I could from SR Base and crunch Primegrid until Rebirther got the server reloaded)... Both of my systems will crunch Primegrid, and continue to crunch Primegrid despite this resource share allocation, and despite SRBase server having work to send, unless I say "No New Tasks" for Primegrid in BOINC.

My initial thought was Rebirther's internet connection was just over burdened by sending and receiving a bunch of short workunits... but I tried over and over during a 10 hour span...

Woke up this morning and both of my rigs are crunching Primegrid Mega Proths... I have to goto work soon... when I get home I will force my rigs to grab SRBase w/u's and see if this problem happens again.

Neo
AtP
5) Message boards : Number crunching : Error on long - but it finished the llr (Message 638)
Posted 15 Jan 2015 by Neo
I just checked the BOINC event log..

It says (after identifying the w/u) "Transient HTTP error"


Neo
AtP
6) Message boards : Number crunching : Error on long - but it finished the llr (Message 637)
Posted 15 Jan 2015 by Neo
I had a similar problem with:

Work Units:
S185_800-900K_wu_4908_0_0
S185_800-900K_wu_4976_0_0

Crunched 100%... System not acknowledging receipt of the workunits, BOINC can't send them to SR Base.... Been done for 10+ hours now. Tried resending the w/u's numerous times over the 10+ hour stretch.
There is no error from my system; it's DOWNCLOCKED, not overclocked..

Been crunching Primegrid because BOINC isn't getting w/u's from SR Base.

:(

Neo
AtP
7) Message boards : News : Decreasing deadlines (Message 498)
Posted 3 Jan 2015 by Neo
You haven't come close to explaining why a short deadline is needed.


plain simple: only after the last wu of a batch has been returned, the results can and have to be evaluated manually.

after that the next batch can be prepared and started.

not that hard to understand - isn't it?


Good summary. :)

Neo
AtP
8) Message boards : News : Decreasing deadlines (Message 495)
Posted 3 Jan 2015 by Neo
Unless you have MANUALLY started a S/R base yourself, you will never really understand why all the work has to be returned back to the server.

After the results are returned (ALL OF THEM), Rebirther has to remove all the K's from the sieve file for each K that was prime.
It's not just one K... for each K there is, for example, 20,000 w/u's for that K depending on the range of N. After that's done, the sieve file (which is the LLR candidate file) is then sieved to a deeper N in preparation for further LLR testing.

God only knows what Rebirther has to do in order to make all of this work with BOINC and to get us a steady stream of work. Which, BTW, has been increasingly more and more steady.


The process is fun, if you're doing it for the first time, and only working on one (1) base....

Rebirther: If I oversimplified the above, my apologies.

Neo
AtP
9) Message boards : Number crunching : Work available?? (Message 482)
Posted 3 Jan 2015 by Neo
Robert,

I crunched all those w/u's...
Server empty..

:P

Neo
10) Message boards : Number crunching : Too long calculation for the Sierpinski/Riesel Bases - long (Message 472)
Posted 3 Jan 2015 by Neo
The massive difference is due to the AVX instruction set in the intel cpu's.

You will see the same result if you crunch on Primegrid.

Neo
AtP
11) Message boards : Number crunching : Work available?? (Message 424)
Posted 1 Jan 2015 by Neo
Happy new year and now it´s startin to roll :-)

Tasks ready to send 35,723
Tasks in progress 6,038



Yeppers, for the last 24 hours there's been plenty of w/u's to crunch!!

Thanks Rebirther!

Neo
AtP
12) Message boards : Number crunching : Sieving (Message 396)
Posted 31 Dec 2014 by Neo
There's a challenge on Primegrid fast approaching. My cores will be dedicated to crunching for that challenge...

After that challenge though, I will take two bases significantly deeper... I will P/M Rebirther for his recommendation on which two bases he thinks we have a good chance of knocking 'out of the park'.

It's a good way to bring a TOP 5000 prime to the 'surface' :)

-Neo
AtP
13) Message boards : Number crunching : Raspberry Pi (Message 391)
Posted 31 Dec 2014 by Neo
Do enlighten me....


finish prescool, learn just a little bit about processor architectures and you will figure out with no problem why LLR on ARM is just - well - not that bright of an idea.


Frank,

I've seen your type on Mersenne before, and so have others on this project; assholes aren't tolerated for long. You made a smart ass comment regarding a good faith question that I made in a different msg thread re: work unit times, and now your just being a dick to one of my teammates who asked a good faith question in order to utilize equipment that he has in order to benefit this project.

I don't know jack shit about mathematics; I got an "A" in college algebra, and that's about it. But I am really really fascinated with the science behind PrimeGrid and SRBase. I think this project KICKS ASS. All my questions, and FURTHERMORE, AND FOR YOUR INFORMATION, NOW AND IN THE FUTURE, any question by anyone of my teammates, IS FOR THE SOLE BENEFIT OF FURTHERING THE GOALS OF THIS PROJECT. Don't ever mistake it. There's a reason why the processing power behind this project tripled in the last three days, or work now runs out in mere hours. Surely you're smart enough to figure that out. P/M Rebirther if you need further explanation.

AtP brings it hard, whether it be crunching workunits, or developing all the SOFTWARE BEHIND crunching the workunits. Are you aware of that? Go ahead and run your mouth now... better yet, go dust off your systems from 1990.


IF YOU ARE NOT DOWN FOR THE CAUSE, GET OFF THIS PROJECT.

Neo
AtP
14) Message boards : Number crunching : Work available?? (Message 370)
Posted 30 Dec 2014 by Neo
What is the point in waiting to send out new WUs, until the processed ones are returned? They are not depending on each other, are they?
It seems like wasted time to me.
Some units will never be processed, so they have to time out.
Again, total waste of time.



When a prime is found for a certain base, those additional w/u's for that specific K for which the prime was found can be removed from the LLR candidate file. When N gets larger this becomes far more important.

No nobody wants to crunch w/u's that have no importance...

As for those hosts that have "slower" computers.... Perhaps they should modify their preferences to crunch the shorter tasks? Primegrid forbids people with Pentium D processors from crunching SoB and GenferWR w/u's...

I know this project is expected to "last years" for all the bases up to 1030, but I am hopefully foolish that we can start knocking down bases within 2-3 days... We just need the work and deeper sieves.

Rebirther & all the guys over at Mersenne Forum who made this project possible - THANK YOU. This project kicks ass!

GIVE US MORE WORK!!!!!

Neo
AtP
15) Message boards : Number crunching : Sieving (Message 368)
Posted 30 Dec 2014 by Neo
Rebirther,

Ok, I can/will do that. Are there any bases in particular that we are working on here at SRbase that you feel need to be taken significantly deeper?

Again, I was thinking of those bases where five or less K's remain (i.e., where N is above 500K)

Neo
AtP
16) Message boards : Number crunching : Work available?? (Message 363)
Posted 30 Dec 2014 by Neo
Why is there a 24 hour deadline for S873 (n-50-100K)? when R663 has a 4 hour deadline for the same range of N?

If users can only get 2 w/u's per core, why the huge time period to return a work-unit. I read in another thread that no new work is loaded until all w/u's are returned....

Why not just set a 30 minute deadline?

Neo
AtP
17) Message boards : Number crunching : Sieving (Message 361)
Posted 30 Dec 2014 by Neo
Rebirther,

I fear to speculate how difficult it would be to set up a sieving effort via boinc, but ... there's been a TON of computing power added just within the last 3 days to this project .... but,

I think it a worthy endeavor to start eliminating bases outright... and hence my question as to what is being done on the bases with 5k's or less in regards to sieving since N is getting larger and tests are taking longer??

If BOINC sieving isn't feasible at this time, I am familiar with SRsieve and can do some manual sieving; I have a couple of i5 cores to donate...

Neo
AtP
18) Message boards : Number crunching : Very long download times (Message 331)
Posted 28 Dec 2014 by Neo
35+ minutes to download a 5 mb file is pretty unacceptable. (srbase2_x64_00001.zip).

Waiting to crunch workunits... but still waiting for download to finish.

Neo




Main page · Your account · Message boards


Copyright © 2014-2021 BOINC Confederation / rebirther