DHEP (Distributed Hardware Evolution) - LIVE

[Bryan]

DHEP is now live and exporting stats to FreeDC. It pays well :D

PROJECT HERE (http://dhep.ga/boinc/)

JAVA must be installed on your computer to run it.

[Ron Shurtz]

Wow! 2 to 2.5 week WUs.
This will be tough with the FB sprints popping up.
Well, I put some on for a slow burn.
Hopefully they can be suspended now and then. :)

[Bryan]

The WU that show on your machines are actually "dummies". You can start/stop/abort them at anytime with no loss of work or credit. The WU downloads some starting parameters and starts crunching looking for a solution to a problem. If it dead ends it modifies the parameters and begins the cycle again. This process keeps repeating. If the search truly can't make any further progress it ends itself and you get another WU.

[Bryan]

Credit clarification: the WU phones home once an hour and reports what it has done. Right now a 2G core/thread makes around 300 credits/hour and a 3.4G thread is producing about 500 credits/hr. So you could lose that amount of credit if you shut down right before it wants to call in. When the project goes from Beta to Production the credits will be reduced to hit the goal of 175/hr for a 2G thread and 300/hr for a 3.4G thread. That will be about a 40% reduction from the current levels. Even so, it pays quite well :D

[Ron Shurtz]

Cool.
I see both WUProp and Free-DC reporting progress now. :)

[Maxwell]

Did they fix the "it freezes the entire computer" issue?

[Bryan]

Did they fix the "it freezes the entire computer" issue?

Don't know, I've never had that problem although some have reported it. There was an issue about suspending WU that apparently caused a computer to become "laggy" to the extreme. That was fixed quite a while ago.

[Username is valid]

Don't know, I've never had that problem although some have reported it. There was an issue about suspending WU that apparently caused a computer to become "laggy" to the extreme. That was fixed quite a while ago.
Yes, the freezing/lagging was related to the verbose logging to the Windows stderr log files.
That was fixed quite a while ago as Bryan says but anyone who still had one or more of those old WUs running could have experienced it.
Now everyone has dumped and reloaded their WUs it is not a problem.

[Bryan]

Hey PDW, how you doing? Good to see you!

[Username is valid]

Never been away, my ID is permanently there under active users !
See Pete has brought more threads to his cause :>

[Bryan]

Never been away, my ID is permanently there under active users !
See Pete has brought more threads to his cause :>

Yeah, it looks like he has added a couple of Thread Rippers to his arsenal. I'm thinking about doing a 64t TR2 but instead of adding I will retire 2 of the X5675 24t machines.

[Username is valid]

Yeah, it looks like he has added a couple of Thread Rippers to his arsenal. I'm thinking about doing a 64t TR2 but instead of adding I will retire 2 of the X5675 24t machines.Retire as in turn them off until you turn them back on again or get rid of ?
Congrats Pete on grabbing the 100M spot =D>

[Bryan]

Retire as in turn them off until you turn them back on again or get rid of ?


Now that is a good question with an unknown answer =)) Come Pentathlon time they might get the cobwebs blown out of them :D

[pinhodecarlos]

Still remember to run this project, when IBM java was quicker 'than standard' java.

[pinhodecarlos]

Last time I ran this project was in 2005 and in one hour of testing the client on three cores I’ve made more 5400x (effort) than my previous 30 days in one core effort. I need to finalise some work for Gaps then I’ll get back to this amazing project which I remember being very fanced about it. I’ve run the standalone client which is faster than boinc one.

Back in 2005 the IBM java was faster than the Sun java, I’ll need to dig out if it still stands. Along side Ubero these two projects were faster with IBM’s java.

[pinhodecarlos]

I’ve played around the two java versions and I didn’t see any difference, maybe I’m not using the valid IBM java version since I’ve downloaded a public version. Also it’s difficult to benchmark the client. Anyway, I’m using the standalone client since I can easily control memory usage. Also I’ve noticed when java service is running the antivirus kicks in so I had to exclude the project folder from it. Still 12 hours to finish my gap work so I can turn on 4 threads of the client with 3.5GB allocated.

[pinhodecarlos]

Where can I find the overall project stats on FreeDC stats page?
The overall stats from DHEP are available here: https://www.dhep.ga/statsrankings.php

[Bok]

I'm only parsing the boinc exported stats from DHEP.

Way back in the day I used to parse those when they were available in cvs format.

[pinhodecarlos]

Last time I ran this project was in 2005 and in one hour of testing the client on three cores I’ve made more 5400x (effort) than my previous 30 days in one core effort. I need to finalise some work for Gaps then I’ll get back to this amazing project which I remember being very fanced about it. I’ve run the standalone client which is faster than boinc one.

Back in 2005 the IBM java was faster than the Sun java, I’ll need to dig out if it still stands. Along side Ubero these two projects were faster with IBM’s java.

I was wrong on my assessment led by wrongly reading the stats. Previously the total effort was the same as the evaluated circuits which for my case I had 400M. After 15 hours I’ve reached 144M which the rate will come down from those 5400x. I slept on the subject and I couldn’t understand where were the gainings since 13 years ago (32 bits vs 64 bits client, number of cores, frequency, productivity of single core, java improvements).

[pinhodecarlos]

I'm only parsing the boinc exported stats from DHEP.

Way back in the day I used to parse those when they were available in cvs format.

Do you want me to contact Michael on the subject?

[Bok]

Do you want me to contact Michael on the subject?

I've been working with Michael for the past few months, in fact I'm the one who pushed him to convert to a BOINC project in the first place :)

So no, I don't really intend to parse any more non boinc stats.

[pinhodecarlos]

I've been working with Michael for the past few months, in fact I'm the one who pushed him to convert to a BOINC project in the first place :)

So no, I don't really intend to parse any more non boinc stats.

Great to hear that Phill!!!!
Upset with no more non boinc stats.....
I love this project but I will stick with the standalone client.

[Bryan]

I just hit 100M so I'm done for a while :D

[pinhodecarlos]

I just hit 100M so I'm done for a while :D
For the equivalent effort units that will take me 10 years.

[John P. Myers]

I'm thinking about doing a 64t TR2 but instead of adding I will retire 2 of the X5675 24t machines.
Fresh off the presses. Boost clock on the 2990X is 4.2 GHz.
2961

[pinhodecarlos]

I’ve reached 1000M circuits evaluated, next goal 10,000M.

[pinhodecarlos]

Since I love this project I am running the standalone client and in a process of requesting to Michael to get all my points onto BOINC points in this way I will help the team although I am already helping on the overall standard credits.

[pinhodecarlos]

Was wondering how much power is needed to get by Grid5000 team on the overall stats. I can only produce close to 200M circuits daily, not sure how much it is in effort units!!!

[pinhodecarlos]

I’m in a monologue but I was wondering how many threads are you guys running.
Right now I’m only running 5 threads of the standalone client on my ivy bridge laptop and currently on position 33 on the daily stats: https://www.dhep.ga/statsfull.php?title=Most%20Powerful%20-%2010/8/2018&colname=todayevals&coldes=Circuits%20Evaluated&arrow=0

I’m curious about this since I bet the client is quicker than using Boinc wrapper. On the other hand I don’t think I’m getting the same boinc credits as if I was running the boinc client.

PS(I’ve just confirmed boinc credits are fair to others)

[shiva]

I'm running 64 threads. when I can keep them running.

[Maxwell]

I'm running six on a six-core AMD. ~30k/day. I'll probably run it to a million then work on other things (slowly)

[pinhodecarlos]

At current pace we will get Grid5000 in 505 days.

[pinhodecarlos]

At current pace we will get Grid5000 in 505 days.

374 days.

[pinhodecarlos]

I've noticed I produce more if I lower the coms with the server, also it will give a higher chance to find a knack. Allocating more memory is also an advantage.

start /low /b /wait java -Xmx2000m -server -Djava.security.policy=java.policy -Djava.rmi.server.useCodebaseOnly=false -jar ITClient.jar rmi.dhep.ga dubhnahUidhe Carlos.Pinho -i 5 -c 10 BoincUserStandalone_XXX SETI.USA

[pinhodecarlos]

I’ve reached 1000M circuits evaluated, next goal 10,000M. Almost half way there but in 3 days I will need to shut down the computer for a week.

[pinhodecarlos]

Almost half way there but in 3 days I will need to shut down the computer for a week.
Back to it

[pinhodecarlos]

Hey Mumps_[MM], is this you: https://www.dhep.ga/statsisland.php?island_id=7005. Don't forget to join us.

[Mumps]

Just dabbling. I was wondering how I could be 229th on "the team" this morning. :) Thanks for the reminder pinhodecarlos!

[pinhodecarlos]

Just dabbling. I was wondering how I could be 229th on "the team" this morning. :) Thanks for the reminder pinhodecarlos! Ok, also check my post #34 but someone who has a bunch of cores probably you will get some knack after all, got two so far with only 5 threads running.

[pinhodecarlos]

Watch my numbers running 4 threads of the standalone client with the latest java version on Win64. Slow ivy bridge laptop.