A few questions :) - Page 3

dswanson · 2002-08-20, 04:58

Hi to you too, Garo.

Between you, me, djaquay, and daran, we haven't been leaving much for anyone else the past few months. I started grabbing little ones in May, when there were 786 left to go. 'twas so easy then - I could get 5 or 6 in a single night, and easily keep a couple of weeks backlogged. Now I'm lucky if a couple expire at a time, and I have no backlog at all - every 6xxxxxx number I have is running. I've actually had to start running first-time LL's on a couple of my faster machines

How many of the TPR machines running 6xxxxxx DCs are yours? garo1 and garo3 are obvious, but what about the outlnder's and the others? And how did you manage to collect such a large stash of DCs at the funny hour (1:42-1:43 GMT) on June 5?

Unfortunately I think it's going to be a long time till we see M38 proved. In fact it's going to be quite a while before we even finish DCs through 6000000. Our friend rexmaughan is sitting on 5837891 with 558.0 days to go, and reports back somewhat regularly, so little chance that will expire soon. My guess is that he gets poached before he completes it.

Speaking of poaching, have you noticed that shaneamy has been pseudo-poaching numbers lately? He (she?) waits until a small number is within a couple of days of expiring, runs it, then returns it within hours after it expires. I noticed this a week ago when I got assigned 6195517 and started running it. At my next checkin it was reported that the number wasn't assigned to me because it was completed. Cleared exponent report stated shaneamy turned it in, who was not the person who owned the number when it expired. That cost me a half-day's work. He's since pulled the same trick with 6050507. So if you get assigned a really small number, I suggest you hold off on starting it for a day or so, just in case.

Here's a tip: I'm out of town tomorrow (Tuesday), so the three numbers expiring then are wide open. Grab 'em!

dswanson

garo · 2002-08-21, 06:45

Hey hey,
I'm sitting here with my communicating account open and all I get is a primenet server not available will try contacting in an hour. Very frustrating, knowing that you are not in town today :)
Yeah looks like it will be a while before we get to 6M even but there's nothing to do but twiddle your thumbs. It's somewhat likely that the 6 and 7M milestones comes close to each other. trif who usually ahngs out at the Ars forum thread used to be a champion at collecting small exponents. You'll see her name on quite a few of them still. djaquay of course was prolific till May/June when you starting sweeping everything. In any case, I don't mind if any on us get the exponents as long as it's someone who will complete it in a reasonable amount of time. What I hate it a lucky newbie S0xxxx something as UserID and Cxxxxxxxxxxx as the computer name getting one of these exponents and sitting on them for ages.
So shaneamy has started poaching eh? I guess rexmaughan's completion dates are driving quite a few of us to deperation but still poaching is just not right.
At the moment I have only garo1 and garo3 on sub 7M exponents. I had defiant and intrepid on them as well but they being P4 1800's finished their stash. outlnder and alf are two other team members of ours.
And about the 1:43 GMT exponents. I was just sitting one afternoon when I decided to check the Primenet status page and saw this whole slew of exponents that George had released for triple checking since the two reults did not match. He usually releases them at odd times and there is usually no advance notice. Anyway as soon as I saw that on the status page I jumped into action and grabbed as many exponents under 7.64M as I could. Why 7.64M? Coz that's where the P4 384K FFT range ends. And almost all of them were P-1 done so no time wasted doing that either. Anyway several of them are sitting in our orphanage annie right now and get assigned to memebrs who want them. Since there is NO danger of them being poached I update them once in a while. If you want some - drop me a line at annie@teamprimerib.com Of course since they will still be officially assigned to TPR you will have to put them at the top of your workto otherwise they will get removed.
Garo

Ian_H · 2002-08-22, 23:43

Hi all

Fascinating to have this forum at last. Thanks to those who've organised it!

Apologies if some of the following seems off-topic to some of you - although my understanding is that the Lounge and this thread aren't meant to be too narrow.

As an individual GIMPSer I'm confused by some of what I've read here and made curious by some of the rest. What is the significance of the Sxxxxx / Cxxxxxxxx accounts/computers? GIMPSers from Mars?

I first joined GIMPS quite early on and was incredibly pleased when the acquisition of a Pentium 166 (!) meant I was able to crunch faster and whizz up the rankings. I believe I made the top 100 at one point, but that's well in the past. I stopped for a while, after moving from Germany (where I had ISDN at home) to the UK (where I couldn't even afford dial-up modem access to start with). Some years later, with ADSL and four PCs working for GIMPS (two at home, two at the office) I've just crept into the top 1,000. And that's with 2 x P4 1.7 (1 work, 1 home), 1 x Athlon 1.1 (home) and 1 x P3 500 (work).

But it remains fun, which is what George tells us to have!

To those of you organised in teams: why? Not meant aggresively, I'm just curious. I can get my head round the idea of colleagues or fellow students or even the owners of a firm starting a team, but I don't understand why anyone would join TPR rather than just signing up for GIMPS direct.

And where does "Annie" the orphan minder fit in? Is she/it doing anything valuable, or just hoarding exponents which the rest of us could otherwise be testing?

I really don't want to upset anyone, and don't want this posting to provoke replies that will upset the guys running this forum either, so mail me direct using <ian@gmx.at> if that seems advisable. I'm just genuinely interested in the different motivations people have for GIMPSing and, in some cases, for joining teams to do so!

Hoping for interesting replies here in the forum,

Ian Harknett
Yorkshire
England

outlnder · 2002-08-23, 00:13

Ian, welcome to the Forum.

First off, no apologies necessary. We don't take things to heart very much, we just want to have fun.

Why join a team? I joined TPR because I am part of the Ars Technica teams for most all DC projects. I started doing Seti, then found Team Lamb Chop and liked what I found at the Ars Technica Forums. I'm doing this with people I already have a community tie with. As part of a team, I can be in the number 1 position. Alone, I would be only 350 or so. Being Number 1 in all DC projects in an imperitive of ours. We are greedy folk, we are.

Annie is just what you said it was. A way of hoarding the best exponents for our team. Is this good, maybe not, but we only take exponents that became expired. We do not take brand new ones( except in very special cases with the permission of George).

The Sxxxxxxx and Cxxxxxxx accounts are those that someone has signed up for and not even bothered to give their computer a unique name. Usually these people(not all) crunch for a few days, see that it takes too long and go do something else. Thus tying up an exponent for 60 or more days. Some of these people have done manual communications and reserved the exponent for much longer periods of time. It is very frustrating for anyone who wants to see the completion of 5M or 6M exponents, and find out that some are checked out for 500 days with no work being performed on them.

Please, don't ever be afraid of posting your true feelings, either positive or negative. That's what this Forum is for.

Xyzzy · 2002-08-23, 00:27

Also, due to a problem we experienced earlier this year, we cannot unreserve exponents...

So, they get sent to our orphanage... Once there our manager picks through them and assigns them to the the boxes that they will do best on...

Right now I personally am P-1 factoring exponents for the orphanage... I get almost no credit for this (Only .001 P90 years per exponent!) but I have a lot of memory that I can let Prime95 use (It uses 800MB right now!) and by doing so it frees up the exponent for someone who only has 128MB, since the actual LL test is pretty easy on memory...

I'm also trying to see how many "DF" factors I can get... :)

Tasuke · 2002-08-23, 00:36

Many who are of TPR, are because of the Ars Technica community, and now moreso because of the comaraderie(and inability to spell) that we share.

As for ANNIE, the purpose is twofold.

One, we try to keep what we are assigned, and many times we may have partially completed work, and need to hand it off. Also there are those with slower computers that wish to get low expiring numbers, but cannot be up at 0600 utc to do it.

There are also exponents that have a higher value for different platforms. Athlons should be doing factoring or first time ll, but are not penalised like the P4 when doing a 33m. A p4 is better suited to first time or double checking, and takes a penalty(credit wise)when doing a 33m. celerons are better for factoring, because of their lack of memory bandwidth. So if someone wants a small range of a certain type of exponent, and does not have the time themself to do the hunting, annie is a great resource.

And it is all about the fun.

Hades_au · 2002-08-23, 04:32

Quote:

Originally Posted by Ian_H

To those of you organised in teams: why? Not meant aggresively, I'm just curious. I can get my head round the idea of colleagues or fellow students or even the owners of a firm starting a team, but I don't understand why anyone would join TPR rather than just signing up for GIMPS direct.

In a sense, we at TPR have the same commonalities as the colleagues or student groups you describe. We are a group of people with the same interest, GIMPS and distributed computing, as many of us participate in a multitude of other DC projects as well. Forming a team has allowed the members of TPR to gain the benefits of others experience and knowledge, plus the advantage of introducing many new participants to GIMPS in an easy manner. There are many people doing GIMPS due to TPR, that would otherwise never have found it. I know, I'm one. In our home thread at Ars Technica, there is always someone to answer any queries and, as has been stated before, it has a 'community atmosphere'.

Why would someone join TPR, or any team for that matter? Well, it would depend on their 'needs'. IMHO, it's always more fun competing as a team than as a individual. It's a place to converse with like-minded individuals in a less structured manner and to top it all off, TPR have the best GIMPS statistics.

dswanson · 2002-08-23, 06:34

Quote:

Originally Posted by dswanson

Our friend rexmaughan is sitting on 5837891 with 558.0 days to go, and reports back somewhat regularly, so little chance that will expire soon.

Today's report has 5837891 at 41.7 days to go. Rex, if you're out there, thanks for bumping up the priority! :D

Ian_H · 2002-08-23, 08:03

Quote:

Originally Posted by Hades_au

There are many people doing GIMPS due to TPR, that would otherwise never have found it. I know, I'm one. In our home thread at Ars Technica, there is always someone to answer any queries and, as has been stated before, it has a 'community atmosphere'.

Anything that brings more people to GIMPs is welcome and you're obviously right that this has been the case with TPR. It can work in the reverse direction: I only discovered Ars Technica, which seems as though it could be a useful resource, through this forum and the references here to TPR.

I hope that the "someone to answer my queries" bit will apply to this forum as well - don't keep all your best tips to yourselves!

Quote:

Originally Posted by Xyzzy

I personally am P-1 factoring exponents for the orphanage... I get almost no credit for this (Only .001 P90 years per exponent!) but I have a lot of memory that I can let Prime95 use (It uses 800MB right now!) and by doing so it frees up the exponent for someone who only has 128MB

That's certainly a very handy aspect of your team approach. Maybe one day there will be a way of automatically dividing work within a LAN: my Athlon 1100 with 640MB of RAM could usefully 'prepare' expopnents for the P4 which only has 256...

Quote:

Originally Posted by outlndr

I started doing Seti, then found Team Lamb Chop

Er, bye for now guys - I obviously need to do some more research into food... :?

Angular · 2002-08-23, 15:13

Quote:

Originally Posted by Tasuke

... Athlons should be doing factoring or first time ll, but are not penalised like the P4 when doing a 33m. A p4 is better suited to first time or double checking, and takes a penalty(credit wise)when doing a 33m. .

What is this credit penalty? Are you referring the smaller FFT sizes due to SSE2's limited 64-bit precision?

Quote:

celerons are better for factoring, because of their lack of memory bandwidth.

Lack of memory bandwidth? There L2 cache is smaller, and on the celerons < 1GHz used the 66MHz bus which would have hurt bandwidth. However, the new 0.13 micron PIII Celerons have a 100MHz bus with 256kb L2 cache. Furthermore, the P4 based Celerons (now up to 2GHz) use a 400MHz bus. Now the 400MHz bus is slower than the new 533MHz bus that the northwood p4's uses, but is memory bandwidth a significant limiting factor?

Tasuke · 2002-08-23, 15:28

Quite in fact. The Tulatin celerons have HALF the bandwidth of a duron processor, and it shows.

With factoring working from the cache, you will not see a difference between a 1333 tulatin p3 Vs. a 1300 tulatin celeron.

Also the way a celeron's cache is cut, the cache bandwidth is halved, as well as the associativity. The celeron is not a separate processor. it is a P3 or P4 with errors in one of its 2 cache blocks, and the offending block is disable, so intel can sell the bad part. (I would like to point out that I am not bashing Intel on this choice, as it brings the cost of the low end down, and there is nothing wrong with the processor)

As for the the williamette Celerons, the cache is hobbled the same way, Halving the bandwidth to the procesor from cache. If you look at the benchmarks, having a williamette celeron do 33m is a waste of resources, as the more bandwidth the processor needs, the more it starves.

The penalty I mentioned is if you run the same type WU for a set amount of time, over the course of a couple of monthe, you get so many years of credit. The celerons get less credit over time the higher the exponents you are testing.

2002-08-22, 23:43	#25
Ian_H Aug 2002 Europe 2·3·5 Posts	Some questions/thoughts from a non-team player Hi all Fascinating to have this forum at last. Thanks to those who've organised it! Apologies if some of the following seems off-topic to some of you - although my understanding is that the Lounge and this thread aren't meant to be too narrow. As an individual GIMPSer I'm confused by some of what I've read here and made curious by some of the rest. What is the significance of the Sxxxxx / Cxxxxxxxx accounts/computers? GIMPSers from Mars? I first joined GIMPS quite early on and was incredibly pleased when the acquisition of a Pentium 166 (!) meant I was able to crunch faster and whizz up the rankings. I believe I made the top 100 at one point, but that's well in the past. I stopped for a while, after moving from Germany (where I had ISDN at home) to the UK (where I couldn't even afford dial-up modem access to start with). Some years later, with ADSL and four PCs working for GIMPS (two at home, two at the office) I've just crept into the top 1,000. And that's with 2 x P4 1.7 (1 work, 1 home), 1 x Athlon 1.1 (home) and 1 x P3 500 (work). But it remains fun, which is what George tells us to have! To those of you organised in teams: why? Not meant aggresively, I'm just curious. I can get my head round the idea of colleagues or fellow students or even the owners of a firm starting a team, but I don't understand why anyone would join TPR rather than just signing up for GIMPS direct. And where does "Annie" the orphan minder fit in? Is she/it doing anything valuable, or just hoarding exponents which the rest of us could otherwise be testing? I really don't want to upset anyone, and don't want this posting to provoke replies that will upset the guys running this forum either, so mail me direct using <ian@gmx.at> if that seems advisable. I'm just genuinely interested in the different motivations people have for GIMPSing and, in some cases, for joining teams to do so! Hoping for interesting replies here in the forum, Ian Harknett Yorkshire England

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Two questions:	Dubslow	GPU Computing	1	2011-08-05 18:22
Questions about the QS	Carmichael	Factoring	8	2007-04-10 11:30
5 questions	OmbooHankvald	Factoring	6	2005-08-28 19:31
Questions	OmbooHankvald	Prime Sierpinski Project	2	2005-08-01 20:18
LLR questions	OmbooHankvald	Math	6	2005-06-23 11:42

2002-08-20, 04:58	#23
dswanson Aug 2002 C8₁₆ Posts	Hi to you too, Garo. Between you, me, djaquay, and daran, we haven't been leaving much for anyone else the past few months. I started grabbing little ones in May, when there were 786 left to go. 'twas so easy then - I could get 5 or 6 in a single night, and easily keep a couple of weeks backlogged. Now I'm lucky if a couple expire at a time, and I have no backlog at all - every 6xxxxxx number I have is running. I've actually had to start running first-time LL's on a couple of my faster machines How many of the TPR machines running 6xxxxxx DCs are yours? garo1 and garo3 are obvious, but what about the outlnder's and the others? And how did you manage to collect such a large stash of DCs at the funny hour (1:42-1:43 GMT) on June 5? Unfortunately I think it's going to be a long time till we see M38 proved. In fact it's going to be quite a while before we even finish DCs through 6000000. Our friend rexmaughan is sitting on 5837891 with 558.0 days to go, and reports back somewhat regularly, so little chance that will expire soon. My guess is that he gets poached before he completes it. Speaking of poaching, have you noticed that shaneamy has been pseudo-poaching numbers lately? He (she?) waits until a small number is within a couple of days of expiring, runs it, then returns it within hours after it expires. I noticed this a week ago when I got assigned 6195517 and started running it. At my next checkin it was reported that the number wasn't assigned to me because it was completed. Cleared exponent report stated shaneamy turned it in, who was not the person who owned the number when it expired. That cost me a half-day's work. He's since pulled the same trick with 6050507. So if you get assigned a really small number, I suggest you hold off on starting it for a day or so, just in case. Here's a tip: I'm out of town tomorrow (Tuesday), so the three numbers expiring then are wide open. Grab 'em! dswanson

2002-08-21, 06:45	#24
garo Aug 2002 Termonfeckin, IE 2²×691 Posts	Hey hey, I'm sitting here with my communicating account open and all I get is a primenet server not available will try contacting in an hour. Very frustrating, knowing that you are not in town today :) Yeah looks like it will be a while before we get to 6M even but there's nothing to do but twiddle your thumbs. It's somewhat likely that the 6 and 7M milestones comes close to each other. trif who usually ahngs out at the Ars forum thread used to be a champion at collecting small exponents. You'll see her name on quite a few of them still. djaquay of course was prolific till May/June when you starting sweeping everything. In any case, I don't mind if any on us get the exponents as long as it's someone who will complete it in a reasonable amount of time. What I hate it a lucky newbie S0xxxx something as UserID and Cxxxxxxxxxxx as the computer name getting one of these exponents and sitting on them for ages. So shaneamy has started poaching eh? I guess rexmaughan's completion dates are driving quite a few of us to deperation but still poaching is just not right. At the moment I have only garo1 and garo3 on sub 7M exponents. I had defiant and intrepid on them as well but they being P4 1800's finished their stash. outlnder and alf are two other team members of ours. And about the 1:43 GMT exponents. I was just sitting one afternoon when I decided to check the Primenet status page and saw this whole slew of exponents that George had released for triple checking since the two reults did not match. He usually releases them at odd times and there is usually no advance notice. Anyway as soon as I saw that on the status page I jumped into action and grabbed as many exponents under 7.64M as I could. Why 7.64M? Coz that's where the P4 384K FFT range ends. And almost all of them were P-1 done so no time wasted doing that either. Anyway several of them are sitting in our orphanage annie right now and get assigned to memebrs who want them. Since there is NO danger of them being poached I update them once in a while. If you want some - drop me a line at annie@teamprimerib.com Of course since they will still be officially assigned to TPR you will have to put them at the top of your workto otherwise they will get removed. Garo

2002-08-23, 00:13	#26
outlnder Aug 2002 2×3×53 Posts	Ian, welcome to the Forum. First off, no apologies necessary. We don't take things to heart very much, we just want to have fun. Why join a team? I joined TPR because I am part of the Ars Technica teams for most all DC projects. I started doing Seti, then found Team Lamb Chop and liked what I found at the Ars Technica Forums. I'm doing this with people I already have a community tie with. As part of a team, I can be in the number 1 position. Alone, I would be only 350 or so. Being Number 1 in all DC projects in an imperitive of ours. We are greedy folk, we are. Annie is just what you said it was. A way of hoarding the best exponents for our team. Is this good, maybe not, but we only take exponents that became expired. We do not take brand new ones( except in very special cases with the permission of George). The Sxxxxxxx and Cxxxxxxx accounts are those that someone has signed up for and not even bothered to give their computer a unique name. Usually these people(not all) crunch for a few days, see that it takes too long and go do something else. Thus tying up an exponent for 60 or more days. Some of these people have done manual communications and reserved the exponent for much longer periods of time. It is very frustrating for anyone who wants to see the completion of 5M or 6M exponents, and find out that some are checked out for 500 days with no work being performed on them. Please, don't ever be afraid of posting your true feelings, either positive or negative. That's what this Forum is for.

2002-08-23, 00:27	#27
Xyzzy "Mike" Aug 2002 2⁵·257 Posts	Also, due to a problem we experienced earlier this year, we cannot unreserve exponents... So, they get sent to our orphanage... Once there our manager picks through them and assigns them to the the boxes that they will do best on... Right now I personally am P-1 factoring exponents for the orphanage... I get almost no credit for this (Only .001 P90 years per exponent!) but I have a lot of memory that I can let Prime95 use (It uses 800MB right now!) and by doing so it frees up the exponent for someone who only has 128MB, since the actual LL test is pretty easy on memory... I'm also trying to see how many "DF" factors I can get... :)

2002-08-23, 00:36	#28
Tasuke Aug 2002 1001000₂ Posts	Many who are of TPR, are because of the Ars Technica community, and now moreso because of the comaraderie(and inability to spell) that we share. As for ANNIE, the purpose is twofold. One, we try to keep what we are assigned, and many times we may have partially completed work, and need to hand it off. Also there are those with slower computers that wish to get low expiring numbers, but cannot be up at 0600 utc to do it. There are also exponents that have a higher value for different platforms. Athlons should be doing factoring or first time ll, but are not penalised like the P4 when doing a 33m. A p4 is better suited to first time or double checking, and takes a penalty(credit wise)when doing a 33m. celerons are better for factoring, because of their lack of memory bandwidth. So if someone wants a small range of a certain type of exponent, and does not have the time themself to do the hunting, annie is a great resource. And it is all about the fun.

2002-08-23, 15:28	#33
Tasuke Aug 2002 2³×3² Posts	Quite in fact. The Tulatin celerons have HALF the bandwidth of a duron processor, and it shows. With factoring working from the cache, you will not see a difference between a 1333 tulatin p3 Vs. a 1300 tulatin celeron. Also the way a celeron's cache is cut, the cache bandwidth is halved, as well as the associativity. The celeron is not a separate processor. it is a P3 or P4 with errors in one of its 2 cache blocks, and the offending block is disable, so intel can sell the bad part. (I would like to point out that I am not bashing Intel on this choice, as it brings the cost of the low end down, and there is nothing wrong with the processor) As for the the williamette Celerons, the cache is hobbled the same way, Halving the bandwidth to the procesor from cache. If you look at the benchmarks, having a williamette celeron do 33m is a waste of resources, as the more bandwidth the processor needs, the more it starves. The penalty I mentioned is if you run the same type WU for a set amount of time, over the course of a couple of monthe, you get so many years of credit. The celerons get less credit over time the higher the exponents you are testing.