New user for manual results (mfaktc), any word on account verification time frame so I can submit?

[Realizm]

Hello,

I signed up yesterday and wanted to give the GTX 1060 and RTX 2080 Ti a try using mfaktc. I only managed two results before manual submissions were disabled, apparently someone was submitting wrong results or some such.

I'm now prompted with me needing to have an administrator review/verify the account before I can submit manually. My mersenne.org info should be the same as my forum account for any administrator(s) looking into this. Any help getting this going would be appreciated (or a time frame on the verification).

Thanks a bunch!

[kriesel]

Originally Posted by Realizm

Hello,

I signed up yesterday and wanted to give the GTX 1060 and RTX 2080 Ti a try using mfaktc. I only managed two results before manual submissions were disabled, apparently someone was submitting wrong results or some such.

I'm now prompted with me needing to have an administrator review/verify the account before I can submit manually. My mersenne.org info should be the same as my forum account for any administrator(s) looking into this. Any help getting this going would be appreciated (or a time frame on the verification).

Thanks a bunch!

Welcome. While you're waiting, keep that RTX2080Ti cranking TF! The current precautions were introduced in response to the issue described in post 153 at https://www.mersenneforum.org/showth...=24103&page=14

[Madpoo]

Originally Posted by Realizm

Hello,

I signed up yesterday and wanted to give the GTX 1060 and RTX 2080 Ti a try using mfaktc. I only managed two results before manual submissions were disabled, apparently someone was submitting wrong results or some such.

I'm now prompted with me needing to have an administrator review/verify the account before I can submit manually. My mersenne.org info should be the same as my forum account for any administrator(s) looking into this. Any help getting this going would be appreciated (or a time frame on the verification).

Thanks a bunch!

Yeah... we have one yahoo trying to spoil it for everyone else.

Just keep cranking away. I'll check your account. Right now we're experimenting with a few different things to make sure the impact is minimal. The trick of course is trying to determine what subtle things differentiate a real user from the idiots.

We'll probably need to make a few structural changes here and there so in the meantime please be patient with us. I'd say try again about this time tomorrow and we should have things a little bit more settled.

[chalsall]

Originally Posted by Madpoo

Yeah... we have one yahoo trying to spoil it for everyone else.

Notify the police. Register the complaint.

Seriously.

[Realizm]

Originally Posted by Madpoo

Yeah... we have one yahoo trying to spoil it for everyone else.

Just keep cranking away. I'll check your account. Right now we're experimenting with a few different things to make sure the impact is minimal. The trick of course is trying to determine what subtle things differentiate a real user from the idiots.

We'll probably need to make a few structural changes here and there so in the meantime please be patient with us. I'd say try again about this time tomorrow and we should have things a little bit more settled.

Thanks, much appreciated. It really sucks that some people have nothing better to do than to cause trouble and problems for others. I wonder what sort of cranial malfunction makes someone get joy from that, boggles the mind.

I'll keep the 2080 Ti (OC) rippin', so far real-time output shows about 4095 GHz-day/day. The little 1060 GTX is only doing 444 GHz-days/day, poor thing. Going to see what some 970's and 460's will do, though not having much hope for the 460 at this point after seeing the difference between the 2080 Ti and 1060. At least they'll be doing something rather than sitting there, so, might as well give it a go too.

What is everyone else doing for manual tracking of results? I'm super new to this so I've just been getting manual assignments and working on those in the worktodo.txt, then when the results.txt gets decently large I'll append that to a file on my Google Drive and do the same for all systems. Any GUI or automated method that'd work better perhaps?

Thanks guys.

[Prime95]

There is a program called MISFIT that automates GPU TF on Windows. If you cannot find it in the forums, someone here can give you a pointer.

[kriesel]

Originally Posted by Prime95

There is a program called MISFIT that automates GPU TF on Windows. If you cannot find it in the forums, someone here can give you a pointer.

The attachment in post 3 of https://www.mersenneforum.org/showthread.php?t=23371 has a summary of available client management software including Misfit, and links to download, and discussion threads. Post two is similar, about the cpu or gpu Mersenne hunting software applications themselves, which is regularly updated as the main applications' authors release new versions.

[kriesel]

Originally Posted by chalsall

Notify the police. Register the complaint.

Seriously.

Indeed. Around here, if it was before the initial complaint, it's as if it did not occur.

[kriesel]

Originally Posted by Realizm

Thanks, much appreciated. It really sucks that some people have nothing better to do than to cause trouble and problems for others. I wonder what sort of cranial malfunction makes someone get joy from that, boggles the mind.

Joy is not necessarily the result for the source. Consider that the estimate of sociopathy in the general population is about 2%, and probability of having some mental issue or other (anything from mild arachnophobia to severely disabling or life threatening illnesses) is variously estimated at ~25% to 60%. While the % incidence is not zero in this community, I would not be surprised if it was lower here than for the general population, and/or more frequently directed in useful and low-harm ways.

I'll keep the 2080 Ti (OC) rippin', so far real-time output shows about 4095 GHz-day/day. The little 1060 GTX is only doing 444 GHz-days/day, poor thing. Going to see what some 970's and 460's will do, though not having much hope for the 460 at this point after seeing the difference between the 2080 Ti and 1060. At least they'll be doing something rather than sitting there, so, might as well give it a go too.

You may find it enlightening to compute an estimated cost per unit output ($/Primality test, or $/ThzD of TF) for your specific utility costs and other factors. I found orders of magnitude difference across my equipment list. Some older gear we already own and have already fully depreciated can be so power-inefficient per unit output that it's more cost effective to use cloud computing or save the power expense for purchase of a replacement gpu. (Assuming here that you're paying for the electricity.) Louisiana might be a good location for a server farm, and definitely not Hawaii. https://www.electricchoice.com/elect...ices-by-state/

What is everyone else doing for manual tracking of results? I'm super new to this so I've just been getting manual assignments and working on those in the worktodo.txt, then when the results.txt gets decently large I'll append that to a file on my Google Drive and do the same for all systems. Any GUI or automated method that'd work better perhaps?

Thanks guys.

I'm still doing manual reserve and report. It takes time. One of these days I'll get that better organized. I wonder why you're using a Google drive rather than a shared drive on your own system. Maybe you're running GIMPS code on systems in multiple locations? Have fast reliable internet? Using google drive for backup?
One thing I've implemented recently is that when the batch wrapper files for my gpu application installations head toward their exit point, after the application exits for any reason, they append a record to a file whose name is constructed from system name, gpu name, application name (creating the file if it does not exist), logging the date and time and source of the exit message, in a single file server directory. I call the folder appwatch. So I have one place to look, for all apps on all gpus on all systems, for what might need attention. Whenever I see and address any issue logged there, I delete the relevant log file. This makes it easy to check, with a glance, more frequently than daily, if anything needs attention, and what. Normal state is just a 00readme file, and a file from which I copied and pasted bits of batch file code to add that functionality to each of the individual installations' wrapper batch files.
I also have all gpu apps logging to files by redirection from the wrapper batch files. This catches all error messages, timings, etc which I may want to use later.

(If still reading...) Welcome, and have fun, without going broke.

[petrw1]

Originally Posted by Realizm

RTX 2080 Ti

Yeah!!
Welcome to the club.
I got one late last Fall and it screams in TF.
I too am in the 4,000/day range.

Find info on misfit here
https://www.mersenneforum.org/showthread.php?t=17915
It works very well. A lot of us GPU'ers use it.

When you configure it, because the 2080Ti is SOOOOO fast, consider letting GPU72 provide the work (I think it is under the Work Fetch config tab). The GIMPS TF assignments are at a lower bit level and your GPU could crunch through thousands per day.

If you need help with the config let me know; I can provide pictures when I get home.

GPU72 is explained in great detail here:
https://www.mersenneforum.org/forumdisplay.php?f=95
Webiste: www.gpu72.com

In a nutshell it is a sub-system managing TF workload in what is referred to as the "Leading Edge"; that is, using GPUs to complete higher level TF just ahead of the LL and DC workers to shorten their workloads.

There is much more to add once you have digested this sufficiently.

[Realizm]

Originally Posted by kriesel

I'm still doing manual reserve and report. It takes time. One of these days I'll get that better organized. I wonder why you're using a Google drive rather than a shared drive on your own system. Maybe you're running GIMPS code on systems in multiple locations? Have fast reliable internet? Using google drive for backup?
One thing I've implemented recently is that when the batch wrapper files for my gpu application installations head toward their exit point, after the application exits for any reason, they append a record to a file whose name is constructed from system name, gpu name, application name (creating the file if it does not exist), logging the date and time and source of the exit message, in a single file server directory. I call the folder appwatch. So I have one place to look, for all apps on all gpus on all systems, for what might need attention. Whenever I see and address any issue logged there, I delete the relevant log file. This makes it easy to check, with a glance, more frequently than daily, if anything needs attention, and what. Normal state is just a 00readme file, and a file from which I copied and pasted bits of batch file code to add that functionality to each of the individual installations' wrapper batch files.
I also have all gpu apps logging to files by redirection from the wrapper batch files. This catches all error messages, timings, etc which I may want to use later.

(If still reading...) Welcome, and have fun, without going broke.

Using Google drive because I'm working at two different companies and I have to go between those locations and home. If it was all at one location I'd just put it on my home server (which would be handy and preferred). Google Drive is just convenient since it is already used for other things and the machines are running Windows except for the server.

Thanks for the ideas on file management. I feel a bit limited working in Windows but I might be able to get something going. Probably something simple, like making individual directories in GD and having the machines auto-copy the files for each machine into their respective directory. Then maybe have a script at the end of the day that takes all the individual results.txt files and appends them all into one big one and removes the old ones. This way I'd be able to just pull everything from one file. It wouldn't do much good for error check/reporting but it would get the results gathered up at least.

For distributing stuff in the worktodo.txt, that will be a bit more complicated. Maybe dump a bunch of work into a central work.txt to be distributed between all machines, hopefully based off a weight (could be the GHz-days/day value, maybe). Then a simple C program (haven't done Windows scripting in forever) could total up the available machine power and machines and distribute work properly. For example, GD/pc1/ghzd.txt that will have a value of 4090, GD/pc2/ghzd.txt will be 444. If work.txt has 200 lines of TF and it is copying work to pc2 it could take 444/(4090+444)*200 (GHz-day value / total GHz-days available * work available), which could round up to 20 lines. Then pc1 would get the other 180 lines. I don't know if that would realistically distribute the work equally but it's gotta be better than nothing.

It is probably more work than it is worth though. Would be neat to have the Prime95 application include the highly optimized GPU cores and have it manage everything automatically.

Let's see if my account even gets verified first. :)