Posts by Eric Driver

1) Message boards : News : BOINC Pentathlon (Message 3251)
Posted 3 days ago by Profile Eric Driver
Post:
Thanks to all participants!

I had to disable the "accelerated retries" mechanism due to all the abortions. When things stabilize I will turn that back on. This means reissued WUs will have their deadline halved.
2) Message boards : News : BOINC Pentathlon (Message 3248)
Posted 11 days ago by Profile Eric Driver
Post:
All my uploads are stuck...I only have 4000 more


Hopefully they all uploaded by now. I had the same problem last night but when I woke up this morning everything was fine.
3) Message boards : News : BOINC Pentathlon (Message 3245)
Posted 15 days ago by Profile Eric Driver
Post:
Sorry for the big rush on the server.... But I really don't have that many cores running. Blame the others for the overload. ;)

Greetings from Germany
Martin


No need to apologize. It's good to challenge the server every now and then.
4) Message boards : News : BOINC Pentathlon (Message 3239)
Posted 16 days ago by Profile Eric Driver
Post:


Welcome BOINC Pentathlon participants!

I would have posted earlier, but it was hard to reach the server after the initial onslaught. Things seem to be stabilized, at least for the time being...
5) Message boards : News : Batch plan (Message 3238)
Posted 16 days ago by Profile Eric Driver
Post:
When can we get back to subfield 3 and finish it up?


Row 13 of subfield 7 is the last of the batches that were generated years ago, so that's the main reason for closing those out.

Finishing subfield 3 could come after that, but I haven't decided yet.
6) Message boards : News : Batch plan (Message 3236)
Posted 24 days ago by Profile Eric Driver
Post:
Row 10 of subfield 6 is nearing completion. After that we will be moving to row 13 of subfield 7. The first case, DS13x11, was generated before the improved apps, so it will run faster than usual.
7) Message boards : News : Batch plan (Message 3234)
Posted 23 Apr 2022 by Profile Eric Driver
Post:
Can you guys create some progress bars to tell how the search progress is? As a non-math guy, it's hard for me and other people like me to understand what the subfield is, how many subfields are there, and how large is each subfield. Just curious about the it, and I do understand if you guys are busy and don't have time for this.
Go sun devils!

Have you taken a look at the batch status tables:
https://numberfields.asu.edu/NumberFields/batch_status.html

There were originally 7 subfields, now only 3 remain: 3, 6, and 7. Each row in the tables represents a separate search for a specific discriminant bound. The 2 dimensional version of the tables shows the sub-batches better, but the linear version gives the progress information, you just need to scroll down to the line that's active (currently sf6 DS10x12).

I hope that helps!
8) Message boards : Number crunching : Enable server_status.php? (Message 3231)
Posted 27 Mar 2022 by Profile Eric Driver
Post:
Nevermind disregard this found the xml file


You'll notice that the server_status is only updated once per hour so the graph will consist of discrete steps. Work is generated in blocks of 20k whenever the queue (unsent WUs) drops below a certain threshold.
9) Message boards : Number crunching : Changing resources share (Message 3227)
Posted 7 Mar 2022 by Profile Eric Driver
Post:
Can you kindly explain in simple language how to do this so that all projects have the same percentage? Using latest edition of Boinc


You need to go to each project's preferences separately, and set the resource share number. It doesn't have to be a percentage, what matters is the ratio between the numbers (i.e. they can all be 100, which is the default setting).

If you are asking how to change all projects to the same value with one operation, I don't think that can be done.

How can I do this for this project only?


Go to your account page and select "NumberFields@home preferences". There might be separate preferences for home, work, etc; just edit the one you want to change.
10) Message boards : Number crunching : Changing resources share (Message 3225)
Posted 5 Mar 2022 by Profile Eric Driver
Post:
Can you kindly explain in simple language how to do this so that all projects have the same percentage? Using latest edition of Boinc


You need to go to each project's preferences separately, and set the resource share number. It doesn't have to be a percentage, what matters is the ratio between the numbers (i.e. they can all be 100, which is the default setting).

If you are asking how to change all projects to the same value with one operation, I don't think that can be done.
11) Message boards : Number crunching : CUDA work units? (Message 3221)
Posted 8 Feb 2022 by Profile Eric Driver
Post:
Thanks for the update Jim.
12) Message boards : Number crunching : dummies guide to laptop GPUs (Message 3218)
Posted 3 Feb 2022 by Profile Eric Driver
Post:
I need some place to start when considering a used laptop purchase. My understanding that having a windows or linux OS w/ an AMD or NVIDIA gpu really helps w/ number crunching on NumberFields@home and otherBOINC related projects.

It's not a priority in the purchase but it is a consideration. For example, I don't want one so old that OpenCL can't run on it.

Is there a intro guide to this subject? Something along the lines of "GPUs and BOINC projects for Dummies". Maybe a place where I could compare different GPUs commonly found on laptops and their number crunching abilities


I was hoping someone else would chime in, since I don't have a good answer for you. All I can say is the NumberFields app runs best with a 1000 series or better Nvidia GPU or equivalent AMD GPU (but Nvidia is slightly better since it utilizes CUDA instead of OpenCL).
13) Message boards : News : Batch plan (Message 3216)
Posted 28 Jan 2022 by Profile Eric Driver
Post:
We are taking a break from subfield 3 and resuming the search over subfield 6.

The main reason for this is we have not found anything new in about 6 months. It's hard to put a percentage on it, but I am 99.99% confident that this search over sf3 is complete. It seems resources would be better spent on a search that will actually find new fields.

We can always finish sf3 at a later time, if for no other reason than to say it is complete with 100% confidence.

In my opinion it is either 100% complete with 100% confidence, or mathematically incomplete. If there actually isn't anything more to discover, that means we set a bound that is too loose, which could be interesting to investigate.

I also want to ask a question. Are there any peer-reviewed publications/grants/awards/anything related to NumberFields@home 2019-2022? If we kept finding nothing, using the resources more wisely can be a great idea.

We are using the best known theoretical bounds. So yes, they might be too loose, but someone smarter than me would have to find a way of making them tighter.

We had a journal article published back in 2019 for degree 7 fields. The plan was to eventually publish the results of the current degree 10 searches. We would most likely complete the search over sf3 before publication.
14) Message boards : News : Batch plan (Message 3214)
Posted 22 Jan 2022 by Profile Eric Driver
Post:
We are taking a break from subfield 3 and resuming the search over subfield 6.

The main reason for this is we have not found anything new in about 6 months. It's hard to put a percentage on it, but I am 99.99% confident that this search over sf3 is complete. It seems resources would be better spent on a search that will actually find new fields.

We can always finish sf3 at a later time, if for no other reason than to say it is complete with 100% confidence.
15) Message boards : News : 2021 Year in Review (Message 3213)
Posted 20 Jan 2022 by Profile Eric Driver
Post:
If search 6 is an order of magnitude larger than 3. And since 3 will have taken at least 1 year to complete (if we sum the number of days spent working on only field 3). This would imply search 6 will take at least 10 years to complete.

Is the above logic correct?


The logic is correct. The majority of that time will be spent in the last tier of the search. The good news is >99% of the fields are usually found in the lower tiers. So by the time we get there we will be able to say we have found >99% of all degree 10 fields unramified outside {2,5}. With the other subfields complete there are some tests that can be done to check for completeness of this final subfield, which will give us even more confidence on completeness. In summary, after about 2 years we should be able to say it is complete with a very high degree of confidence. Of course, there is also the fact that CPU and GPU processing power goes up with time, which will also help (by analogy, 5 years ago I would have predicted sf3 would take a decade to finish and now it is about done.)
16) Message boards : News : 2021 Year in Review (Message 3211)
Posted 16 Jan 2022 by Profile Eric Driver
Post:
It is interesting, what search is bigger: Search 3 or Search 6 or Search 7?


Search 7 should be similar to search 3, and search 6 should be about an order of magnitude larger than 3 and 7.
17) Message boards : Number crunching : Cross-project statistics (Message 3207)
Posted 8 Jan 2022 by Profile Eric Driver
Post:
how can I allow that:(


Go to your account then select NumberFields@home preferences. Under primary preferences look for this check box:
"Do you consent to exporting your data to BOINC statistics aggregation Web sites?"
18) Message boards : News : 2021 Year in Review (Message 3203)
Posted 3 Jan 2022 by Profile Eric Driver
Post:
So another year has come and gone.

This was not the best year. We started the year with a severe hard drive crash that took weeks to recover from. After that, the year was spent on the search over ℚ(√2) (subfield 3). This search should be complete within several months. A single search seems like a small accomplishment, but I should point out that this is a gigantic search that just several years ago seemed out of reach.

Other than that, in November support was added for Intel GPUs. The app worked for the most part but performance was disappointing. We look forward to the new Intel Arc GPUs coming out this year which should perform much better.

Thanks everyone for your contributions and have a wonderful New Year!



If the Intel GPU application is stable you may get more usage if you take it out of BETA status. Some people do not want to run Test applications and do not have that option checked in their Project preferences.

Thanks
Bill F


Hey Bill,

Although it is stable, with the current Intel GPUs, run times are worse than using a single CPU core. For that reason, I can't justify removing the Beta status.

I'm considering getting a new Arc GPU when they become available so I can test out the app first hand.
19) Message boards : News : 2021 Year in Review (Message 3200)
Posted 1 Jan 2022 by Profile Eric Driver
Post:
So another year has come and gone.

This was not the best year. We started the year with a severe hard drive crash that took weeks to recover from. After that, the year was spent on the search over ℚ(√2) (subfield 3). This search should be complete within several months. A single search seems like a small accomplishment, but I should point out that this is a gigantic search that just several years ago seemed out of reach.

Other than that, in November support was added for Intel GPUs. The app worked for the most part but performance was disappointing. We look forward to the new Intel Arc GPUs coming out this year which should perform much better.

Thanks everyone for your contributions and have a wonderful New Year!
20) Message boards : News : Happy Holidays (Message 3198)
Posted 22 Dec 2021 by Profile Eric Driver
Post:
uhm can you change the name to NDMathematicsVN please (if you want) because I've just changed my name :) Thanks you so so so much Eric


ok


Next 20


Main page · Your account · Message boards


Copyright © 2022 Arizona State University