OFFICIAL "SERVER PROBLEMS" THREAD

[retina]

I fairly sure that LaurV gave the correct answer there. But I want to give a "better" answer.

Forget about all those old GHzDays that were run on the ancient 6502 CPUs. Those cycles are now gone. Instead buy more of the latest machines and get that yearly figure as high as possible. If others are getting ahead of you, then just keep adding more machines.

[ATH]

Originally Posted by LaurV

yep, click "customize" (upper left corner), fill the start date like 2000-01-01, (anything older than 2018 or so)

It has to be before 2008 (not 2018) when primenet v5 started.

[ET_]

Originally Posted by ATH

It has to be before 2008 (not 2018) when primenet v5 started.

I started in 1997...

[LaurV]

Originally Posted by retina

Forget about all those old GHzDays

Shoo.. shoo.. let us have our small incentives! Otherwise how could I boast when I will overtake him/them/those SRBase guys?

[ATH]

Originally Posted by ET_

I started in 1997...

Does not matter since your pre v5 data was transferred in 2008. 2008-01-01 should show you the same lifetime stats as 1997-01-01.
Also I do not think you can see any stats for more than 365 days but less than lifetime stats.

[rebirther]

Manual reporting is nasty this time. Iam always running into the server resource error after trying to report 4800 results, down to 3000 now and still the same issue. Before its taking around 10min to report 4800, last year I could report 9000 at once in 5min.


With the current challenge my backlog is huge. Iam expecting to report around 700-800k TF results. Is there a way to speedup things?

[nobreadsticks]

A factor was found for M31817 a couple months ago, which is shown in the history, but its not shown in the list of factors here or here. I also wasn't able to grab a manual PRP test on cofactors as it says "No assignment available meeting CPU, program code and work preference requirements".

Am I missing something?

[axn]

Do you mean
2021-02-25 Seth_Tr F-PM1 Factor: 6932292235325039464332944259181777 / (P-1, B1=10000000000)?

6932292235325039464332944259181777 is a composite factor made up of previously known small factors = 34553263 · 35571407 · 53134391 · 106147811767

[Happy5214]

I reserved 59652389 as a PRP w/ proof DC, but I ran into issues with an mprime 30.6 bug that wiped out the first 10% of work. I decided to up the P-1 bounds a bit, but when I submitted it (without the AID I was assigned for the PRP), it expired the PRP assignment. I've restarted the PRP test and am at 7.4%, so I don't want this reassigned to someone else, but I can't seem to do another manual PRP w/ proof DC reservation for this exponent. Ideas?

[LaurV]

Grrrr... I am angry with the server now, and I will not tell him "happy birthday server" every year on Aug. 18th, anymore!

I just got robbed of ~400GHzDays of TF credit.

I got this assigned for PRP/LL and, as a good person (i.e. which doesn't trust the result there ) I re-did all the TF from 64 to 81 bits, before investing the time in a long PRP/LL test. I do that for all exponents I get for PRP/LL in 100M-digits range, when they are only TF-ed to 77 bits, which takes 1 hour or less in a couple of 2080Ti cards. So better make sure, and do it from scratch. Sometimes, if no free card, I ask somebody else to help, and people here obliged in the past to take my exponents to 80 or 81 bits, to which I am thankful. Now, with the last 6 or 8 assignments, I did it by myself. For this particular exponent, I was splitting the bitlevels in two cards (64 to 80 in one card, and 80 to 81 in another, as the last one takes the same time as all the other together). After about 5 hours, they finished both in the same time, and luckily, the first card found a factor at bitlevel 80.

Now, I have 3 problems here. Actually, four.

First is that I had a hell of a time trying to report the "no factor from 80 to 81". As there was a factor already there, the server totally ignored my report. I succeeded after some effort and different ways, so problem solved. Got the credit, but most important, cleared the information in the DB, so our grandsons who will want to continue searching for factors of this number, won't need to do the 80 to 81 again.

Second, what the hack? I have seen that the factor is recorded like a P-1 factor, albeit I didn't report it as so. Even if that would be true, the 20GHzD given for it is ridiculous, haha, if this would be a P-1, it would worth millions... Look to that k-split! k = 2³ × 5² × 8792212987223. Ten million bucks GHzDays credit, at least. This is a TF from 79 to 80, and it worth about 400 GHzDays of TF credit. I want it back! (otherwise I can not overtake Syd and Andy in top, and they both will laugh at me!)

Third, this is minor, but actually, there is a little bit of a class or two (or more) at the end of bitlevel 80 which were not tested for factors, because the test stopped once a factor was found. So, the grandsons will have to deal with the gap too.

Forth I am angry with James because he keeps the users by their public name. When I change my public name, I am a different person on mersenne.ca, with different status and history. And the "record" factors are not combined between those users (i.e. if I find a record factor, I may not know it is "mine", unless I remember the public name I had at the time). Now, I know that displaying the real (private) user names is not advisable, and that this issue is not easy to solve in such a way to satisfy everybody, but I am still frustrated by it. PrimeNet also handles this in a wrong way, but we got used with it, nothing to do there (i.e. the users' history is done properly, by private user name, so I don't get different statuses every time I change the public name, but the way the history is shown is also done by the private name, when in fact, the name should everywhere be replaced by the current public name, otherwise, if one goes to my history, for example, he can see my login name, even my real name, if I ever used it to report results or as a public name, which is not right). Anyhow. This only FYI.

[James Heinrich]

Originally Posted by LaurV

Forth I am angry with James because he keeps the users by their public name. When I change my public name, I am a different person on mersenne.ca, with different status and history.

I can only work with the data I have. Data is exported from mersenne.org based on public display name. You don't want to have multiple personalities? Don't change your public name.

I don't know why 332689127 shows as P-1 on mersenne.org, but I have manually adjusted the record on https://www.mersenne.ca/exponent/332689127 to be TF.