| Author |
Message |
|---|
 Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 477,382,143
RAC: 302,741
                               
|
|
For the last several years PrimeGrid has been searching SoB numbers with exponents below n=22M. We have almost exhausted that range of numbers.
Once those are all tested, the next range of numbers we'll be crunching will be from n=27M to n=28M. These will take about 50% longer to crunch than the SoB numbers we're used to. For these new tasks, the deadline has been increased from 40 days to 60 days.
I expect we'll exhaust the old SoB numbers within the next month or two.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
I think that this is great, I can't wait to crunch one or two of these!
____________
 |
|
|
|
|
|
An what is with n=23M to n=26M? Reserved by another project? |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1963
ID: 352
Credit: 6,404,643,830
RAC: 2,581,215
                                   
|
An what is with n=23M to n=26M? Reserved by another project?
Yes, by SoB project
See http://www.seventeenorbust.com/stats/ at Current Test Ranges.
They are in ~26M5 range.
____________
My stats |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 477,382,143
RAC: 302,741
|
|
As I mentioned earlier, we've almost exhausted the current range of SoB candidates and will soon start testing a new batch with 27M < N < 28M. These tasks will take about 50% longer to run than the tasks we've been crunching up until now.
I expect that the new workunits will start being created within the next 24 hours, and will probably start be sent out within 24 to 72 hours.
Credit will be proportionally larger, so I expect each task to be worth about 30K credits. The deadline for the new tasks will be 60 days.
The first of these new SoB numbers is 8,127,821 digits in length.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
That's a big leap in numbers. Do we need any more sieving? |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 477,382,143
RAC: 302,741
|
That's a big leap in numbers. Do we need any more sieving?
No.
____________
My lucky number is 75898524288+1 |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,433,949,218
RAC: 4,124,579
                  
|
|
I'm making an early start on this in preparation for longer waits for my amethyst. Looking at 6 days my end. |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1963
ID: 352
Credit: 6,404,643,830
RAC: 2,581,215
|
|
Got some old SoB task with FFT length: 1920K for example.
New ones for example FFT length: 2304K
In the mix of old ones, stderr.txt states.
BOINC LLR 6.03 wrapper: starting
Major OS version: 6; Minor OS version: 1
FFT length: 2
There is no truncation in copy&paste, it's in the file itself. Just a minor glitch.
While old having numbers in WU name like 71252165, new ones like 220375601 - easy to tell which one is which.
____________
My stats |
|
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 1051
ID: 301928
Credit: 563,881,725
RAC: 1,288
    
|
FFT length: 2
There is no truncation in copy&paste, it's in the file itself. Just a minor glitch.
On these workunits LLR prints "FFT length: 2M" instead of "2048K". I think wrapper blindly fetches the number, not parsing K/M/whatever follows. I don't think this value could be used anywhere so most probably it is just a visual glitch.
|
|
|
JimBHonorary cruncher
Send message
Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,594,363
RAC: 26,927
        
|
|
Just to make it clear, the system doesn't look at that FFT length. We calculate the FFT length of every LLR job ahead of time - that's why the pending credit and show badge pages know the exact amount of credit you'll be getting. In fact the system ignores the Stderr output completely. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 477,382,143
RAC: 302,741
|
FFT length: 2
There is no truncation in copy&paste, it's in the file itself. Just a minor glitch.
On these workunits LLR prints "FFT length: 2M" instead of "2048K". I think wrapper blindly fetches the number, not parsing K/M/whatever follows. I don't think this value could be used anywhere so most probably it is just a visual glitch.
That's 100% correct.
The largest of the older and smaller SoB tasks use an FFT size of 2048K or 2M. That gets reported as "2". It's of no significance to the way the BOINC tasks operates.
____________
My lucky number is 75898524288+1 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 477,382,143
RAC: 302,741
|
|
Two of the new SoB tasks have been returned already! The fastest was returned in just under four days (336954.839 seconds). The CPU is a Xeon E3-1230 v3, which is a Haswell CPU with dual channel memory and it was using the new FMA3 version of LLR.
____________
My lucky number is 75898524288+1 |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,433,949,218
RAC: 4,124,579
|
|
If we can keep them clean + keep the qora down they'll validate quickly. Maybe wording on the project preferences to say only select if you reliable uptime etc. |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1963
ID: 352
Credit: 6,404,643,830
RAC: 2,581,215
|
|
Returned some Longer SoB tasks, let's see how it goes.
http://www.primegrid.com/workunit.php?wuid=386391834
http://www.primegrid.com/workunit.php?wuid=386416558
http://www.primegrid.com/workunit.php?wuid=386416559
This one looks promising since my wingman returned his shorter SoB in 5 days on the same host.
http://www.primegrid.com/workunit.php?wuid=386391834
____________
My stats |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 477,382,143
RAC: 302,741
|
|
Four of the results have validated (two workunits), proving that not everyone waits months for their wingmen!
It's sort of interesting that none of those four results were running the new FMA3 version of LLR. That's merely a coincidence as there are several such results that have been returned and are awaiting validation. There's currently 24 results waiting to be validated, 5 of which used the new LLR.
____________
My lucky number is 75898524288+1 |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1963
ID: 352
Credit: 6,404,643,830
RAC: 2,581,215
|
|
Yeah, this one was completed using stock app and got validated.
http://www.primegrid.com/workunit.php?wuid=386482673
Glad, it worked.
Other 8 are pending and 9 in progress.
____________
My stats |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,433,949,218
RAC: 4,124,579
|
|
Mine'll be finished in an hour:
http://www.primegrid.com/results.php?userid=130544&offset=0&show_names=0&state=1&appid=13 |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1963
ID: 352
Credit: 6,404,643,830
RAC: 2,581,215
|
|
Dave, you need to link to your hosts or WUs, not usersID so others can view it as well.
____________
My stats |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,433,949,218
RAC: 4,124,579
|
|
Oops
http://www.primegrid.com/workunit.php?wuid=386358347 |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,433,949,218
RAC: 4,124,579
|
|
Not a happy bunny:
http://www.primegrid.com/workunit.php?wuid=386358347
Why?
Edit: duh same as below |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 477,382,143
RAC: 302,741
|
Not a happy bunny:
http://www.primegrid.com/workunit.php?wuid=386358347
Why?
Edit: duh same as below
Given that your computer returned the wrong residue (something we can see but you can't), almost certainly the cause is a transient hardware failure, which in turn is often caused by just a bit too much overclocking. (More than a little too much overclocking and the computer wouldn't have made it to the end of the calculation.)
If you're not overclocking and you continue to see problems like this, it's time to start running memory tests, checking the cooling and power supply, etc.
____________
My lucky number is 75898524288+1 |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,433,949,218
RAC: 4,124,579
|
|
I'm at 4.4. Time to rev it down then. It's making a CPU app feel like GFN-WR :). Ty for th fb. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 477,382,143
RAC: 302,741
|
I'm at 4.4. Time to rev it down then. It's making a CPU app feel like GFN-WR :). Ty for th fb.
Be warned that your Sandy Bridge CPU will run a bit hotter, and likely a bit less reliably once we start switching over to the FMA3 LLR app.
(It mostly affects Haswell CPUs, but I've been told that SB and IB CPUs will also see a smaller improvement and temperature increase.)
____________
My lucky number is 75898524288+1 |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3253
ID: 130544
Credit: 2,433,949,218
RAC: 4,124,579
|
|
Pleased to announce I've just had my first long task resolve:
http://www.primegrid.com/workunit.php?wuid=390946339
Now confident they can resolve, + with 49k credit per unit, time to get the remaining 5 in the bag for amethyst :). |
|
|
|
|
Pleased to announce I've just had my first long task resolve:
http://www.primegrid.com/workunit.php?wuid=390946339
Now confident they can resolve, + with 49k credit per unit, time to get the remaining 5 in the bag for amethyst :).
FFT length: 2560K WOW! That's really getting up there. A lot more memory being used, so memory speed AND latency will make a much bigger difference now for SoBs.
____________
Largest Primes to Date:
As Double Checker: SR5 109208*5^1816285+1 Dgts-1,269,534
As Initial Finder: SR5 243944*5^1258576-1 Dgts-879,713
 |
|
|
MonkeydeeVolunteer tester
Send message
Joined: 8 Dec 13
Posts: 548
ID: 284516
Credit: 1,708,258,430
RAC: 3,259,746
              
|
|
Small difference on my AMD FX-8120 from the old to the new... ~330 hours is now ~540 hours.
I guess if I want to keep running SoB in any kind of reasonable time span I'll have to switch out to an Intel for the first time since 2003.
Now to wait impatiently for my one pending unit to validate so I can get SoB Silver. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14037
ID: 53948
Credit: 477,382,143
RAC: 302,741
|
Small difference on my AMD FX-8120 from the old to the new... ~330 hours is now ~540 hours.
I guess if I want to keep running SoB in any kind of reasonable time span I'll have to switch out to an Intel for the first time since 2003.
Now to wait impatiently for my one pending unit to validate so I can get SoB Silver.
You could say that about any LLR project, not just SoB. The same holds for Short GFN also, especially once the next version of Genefer goes into production.
For comparison, the last SoBs I did (long ones) took under 95 hours.
http://www.primegrid.com/results.php?hostid=424445&offset=0&show_names=0&state=0&appid=13
____________
My lucky number is 75898524288+1 |
|
|
