Optimized BOINC Clients FAQ [Archive] - AMD Users.com

View Full Version : Optimized BOINC Clients FAQ

vaughan

03-04-2006, 01:41 AM

NeoGen has created this FAQ for using optimized BOINC clients

Optimized BOINC clients FAQ

Q: What are optimized clients?
A: Optimized clients are boinc executables (.exe files) and sometimes linked libraries (.dll) that are specially compiled to take advantage of the most advanced features of each kind of CPU. these advanced features include special processor instructions like MMX, 3DNow!, SSE, SSE2, SSE3, etc...

Q: Does it make my machine crunch faster?
A: No, it does not make your machine crunch any faster. The Boinc application itself does not crunch anything, it merely manages the crunching apps, starting and stopping them at defined intervals to ballance the cpu time across all the projects that you are crunching for.

Q: So, what good is an optimized client for?
A: Boinc claims credits for a given workunit based on the time that it took to complete it, and the speed of the machine itself. The optimized Boinc client using the most advanced cpu features gives a higher speed value for the machine's benchmarks, which in consequence makes boinc claim more credit than the regular client on the exact same machine. Besides that, some optimized clients have extra functionalities not provided in the regular boinc client.

Q: So, with an optimized client I get much more credits?
A: Yes... and no. The fact is that there are very few projects where you get the credits you claim. Most of the projects calculate an average of several claims, or grant the credits of the machine making the "middle claim". (Not the highest, and not the lowest). But in general, it is always better to use an optimized client, because in case the project grants the average of claims, one or more high claims can grant higher credit for you and the people that crunched your workunit.

Q: Wait, aren't those inflated benchmark values cheating?
A: If you use a recognized optimized client then no, it isn't cheating. The benchmark values will actually show the real potential of your machine, and not the low values given by the regular client. On the other hand, if you're a hacker and you plant absurdely high values on the boinc source code and compile it, then yes it is cheating. (And yes, it has happened already)

Q: What are those extra functionalities said above, and do I need them?
A: Technically, you don't really need them, the regular boinc client already comes with everything necessary to start working. The extra functionalities can be things like reporting workunits instantly (boinc sometimes takes a while), changing the default communications port (useful if for some reason the default port is blocked or used by another application), changing the default process priority, getting more work, and many more.

Q: Where can I get these optimized clients?
A: See gamer007's List of Optimized BOINC Clients (http://www.amdusers.com/modules.php?name=Forums&file=viewtopic&t=1785) thread for a list of known good optimized clients for almost all platforms/OS'es.

Q: How do I install an optimized client?
A: An optimized client is not really "installed". What you do is install the regular boinc client as you normally would, and afterwards extract the contents of the optimized clients zip files inside the boinc folder, overwriting the files as requested. Then you start boinc as normally, and you execute the benchmarks manually once (Advanced menu->Run CPU benchmarks) just so that it records the optimized benchmark values.

Q: Is there any project that can't be run with an optimized client?
A: Any project that runs through a regular boinc client, can run with an optimized client as well. However any project that runs a custom version of Boinc (The japanese Cell Computing projects for example) cannot run with an optimized boinc client.

NVM

03-04-2006, 02:55 PM

i have to admit, reading this has lessened my anti-optimizing thoughts on this subject. but, something still bothers me about overwriting files,even though they are "recognized".

i sure wish i could get over it though,maybe someday soon. my stance on this is preventing me from helping my team mates on certain projects, and DC junkie that i am, its killing me. i may just have to bite the bullet.

BlackAdder

03-04-2006, 03:54 PM

I've run the optimized clients since they came out, no problems. The Seti client really helps a lot !!! :D

NVM

03-04-2006, 03:59 PM

are you using TRUX?

NeoGen

03-04-2006, 05:32 PM

I too sometimes feel weary of running something that was altered and compiled by someone I don't know. I know a few bits of programming but not enough to understand large projects like this. If I could I would compile an optimized boinc myself.
But we gotta give in sometime, don't we? :roll:

PcManiac

10-22-2006, 10:52 PM

The Link for the list of clients is broken :(

What would be a good one for a core 2?

gatekeeper53

10-22-2006, 10:56 PM

Which one would you like?

PcManiac

10-22-2006, 11:08 PM

Which one would you like?
I guess any that would utilize SSE3 and such...

gatekeeper53

10-22-2006, 11:29 PM

I have a 3700+ and the one that I found that works best is at this link:

http://gathering.tweakers.net/forum/list_message/25882224#25882224

I use the SSE2 5.5.0 for windows

Domain name Standby-PC
Local Standard TimeUTC -6 hours
NameStandby-PC Created13 Oct 2006 0:43:39 UTC
Total Credit8,159.74
Recent average credit503.43
CPU typeAuthenticAMD
AMD Athlon(tm) 64 Processor 3700+
Number of CPUs1
Operating SystemMicrosoft Windows Longhorn
Professional Edition, (06.00.5600.00)
Memory1021.88 MB Cache976.56 KB Swap space2300.2 MB
Total disk space74.52 GB
Free Disk Space58.11 GB
Measured floating point speed4075.1 million ops/sec
Measured integer speed13140.53 million ops/sec
Average upload rate25.06 KB/sec
Average download rateUnknown
Average turnaround time0.02 days

Steve Lux

10-22-2006, 11:53 PM

I ran Predictor without an optimized client for over a month on my 3800+ and was averaging about 620-625 credits a day. After "optimizing" as was reccomended to me, my average dropped to about 590-595 a day in Predictor. These things aren't always everything they are expected to be.

gamer007

10-22-2006, 11:58 PM

gatekeeper53

10-23-2006, 01:02 AM

They are Crunch3r's. It seems that anyone and everyone has theaked them to better fit different machine configurations. I have 4 different versions running on 5 machines. I have found about 9 different versions and have tried them on my machines. Some just work better on certain machines. The one on my 4800+ runs like crap on my 3500+'s and vice versa the copy on the 35's stinks on the 48. There are 4 people that I know of that are turning out versions now. I don't have the compiler or I might even try myself. There is a good "How to" on writting a client here:
http://lunatics.at/index.php?op=Articles;article=3

gamer007

10-23-2006, 01:07 AM

Sounds like fun. I should try it sometime. :P

But I don't like installing Visual Studio and then having the stupid "Just-in-Time Debugger" always poping up whenever an error occurs in any program.

Evil-Dragon

10-23-2006, 08:33 AM

Imo clients like Crunch3rs BOINC are forcing most project admins to move to fixed based credit to stop people from cheating. For example, how do we know that crunch3rs client isn't just over exaggerating the figures? Does anyone know what there real integer speed is for a default clocked processor? And does crunch3r's client approximately match that figure or is it above?

A lot of projects are starting to ban 5.5.0 for cheating credit. I don't think it is a good idea for AMD Users to start being labelled as "cheaters"

I'm not saying that Crunch3r's client is the only way that people are cheating for more credit, since you can easily edit client_state.xml to give higher benchmark scores.

I've always been an honest person and just stating my opinion on the matter.

Brucifer

10-23-2006, 03:45 PM

Sigh....... /rant-on

One of the original goals of the boinc effort was to eliminate the "cheating," a laudible effort. Unfortunately that saying of "When you're up to your ass in alligator's, it's hard to remember the initial objective was to drain the swamp" has proved over time to have quite a bit of merit. Too much internal tunnel vision. Excellent concept. The point awarding system is great in theory, all automated dependent of each computer. But another old saying of "what one man maketh, another man tearith apart" is quite appropriate too. This whole points issue is the super sized BOS (ball of $h*t) of boinc. It's all well and good for some to sit and say "ahhhh, it's the science" when in reality a great bulk of the people wouldn't be crunching without a points system. The science attracts them to the project, but the point game keeps them crunching. The poll held elsewhere on here regarding fixed or dynamic points shows just what the percentage is of people that would like fixed points. Why don't they just switch to it?? Well the main reason I can fathom is that then they wouldn't be able to set up relative points between projects, etc. I guess that's fine for some beanie sitting in a chair somewhere. But for me I'd much rather have the fixed points. I'd crunch boinc efforts then on linux. Points would appear as they are "earned" rather than waiting to be validated. Save all that "indexed" stuff for determining how many work units need to be downloaded to a system. Works great for that. But until this whole points thing gets fixed to the point that people can't mess with it, or cheat, it's just gonna keep going on. So my point? In so far as points manipulation goes, boinc isn't really any improvement over the old seti classic which had one *small* thing over boinc.......... and that was that the client worked for everyone.
/rant-off

gatekeeper53

10-23-2006, 04:13 PM

The thing is that all the points granted so far and $1 will get you a cup of coffee. I see the points as being like a shiny new toy that does nothing but keep an interest. No matter the system used if it is used around as large a group of geeks as make up the crunching world there will be someone that finds a way to circumvent it. It seems to me that the tweaking and system talk that we all have in common is a far greater draw than points that really have no worth. Why else would we spend thousands of dollars,pounds,euros,... to keep doing something? Sometimes I can find a fishing buddy on here there is always someone that shares an interest in computers. In real life I get a lot of blank stares if I talk about a new system I'm running or built. Not to mentioned what they think when they find out I have a larger number of computers living in my house than I do people. For all I care they could do away with the points today because by tonight we will have found some other yardstick to use to race. As for being a "cheater" nah nah na nah nah lol

Strongbow

10-23-2006, 05:30 PM

Brucifer

10-23-2006, 06:39 PM

The thing is that all the points granted so far and $1 will get you a cup of coffee. I see the points as being like a shiny new toy that does nothing but keep an interest. l

Prescisely the point. Without the interest and participation then DC would dwindle significantly. And how else do you measure participation between teams? No teams then? These teams would not be anywhere near where/what they are without a point system. And Berkeley realized the validity of having points, otherwise they wouldn't have been built into the system.

Brucifer

10-23-2006, 06:50 PM

I'm with Terry on this one, especially around the point (no pun intended) he made on "the tweaking and system talk that we all have in common is a far greater draw than points that really have no worth." - as I'm crunching due to having a computing hobbie and the points system simply allows me to gauge how much more I can push out of a system compared with other like minded enthusiasts (i don't care if I'm top or bottom, although top does have a feelgood factor - apparently!!! :icon_cry:)- the science is just an added bonus but in all honesty I have no clue what some of them are doing with the results, I enjoyed HashClash but I don't understand it and even though the forum was rife with members complaining about points cheating the project was effectively completed simply due to us members throwing our systems at it and there is no actual harm to the science whether we are optimized or not!

Right then, time to put my money where my mouth is and shutdown all my systems, get into BIOS and tweak these mothers till they melt!!!

Go look at all of the posts................ how much is about getting points on various projects, and then compare that with how large?small the "hardware" thread is, etc.

LOL If points don't matter then why is the team on all the projects, why do we have all the threads on projects, why don't we just withdraw from it all and see how far this site goes as a gossip/hobby. Why do so many people worry about getting more systems bought/borged? Why don't lots more people just take the "anonymous" path of dumping in all their machine time and results into the big void in the sky so they or anyone else has no idea what they have or are doing?

It's because of the points. Period.

gatekeeper53

10-23-2006, 07:38 PM

The points keep many people here. The thing is the one's that have farms and really do a disproportionate amount of the work would be the ones that would just find another yardstick to enable a race. If you look there are over 800,000 people registered with Boinc. Most of those are running their home system and have little intrest in points. To many of them it's just what it was meant to be; a pretty screensaver. My wife says she likes me doing this as when I was driving drag cars it scared her, and was more expensive. It's a competitive thing with many of us and the gissip/hobby is part of it. If there was another way to put man and machine against one another I believe many of us would be splitting our time as we do with different projects. As it is this is what we have. But, the notion that you can make a points system that no one would get around is laughable. If you could write something that bulletproof you are wasteing your time and should be writting security programs. This is great fun as long as people don't start taking anything to seriously. The only serious thing in all this is the science that gets done faster than it would without all us grownups playing with our toys.

Strongbow

10-23-2006, 08:34 PM

Go look at all of the posts................ how much is about getting points on various projects, and then compare that with how large?small the "hardware" thread is, etc.

LOL If points don't matter then why is the team on all the projects, why do we have all the threads on projects, why don't we just withdraw from it all and see how far this site goes as a gossip/hobby. Why do so many people worry about getting more systems bought/borged? Why don't lots more people just take the "anonymous" path of dumping in all their machine time and results into the big void in the sky so they or anyone else has no idea what they have or are doing?

It's because of the points. Period.

Don't get me wrong about points, points do matter just like you stated -- without points there would probably be no teams and without teams there'd quite possibly be no amdusers.com. Points are important and makes a project competitive between the members and that can only be a good thing for the 'whatever' project. I am certainly not against points and am certainly not against gaining as many points as I can by O/Cing my systems and letting them crunch 24/7/365.

EDIT : Why am I crunching Tanpaku at the moment? - simply to help the team gain the #1 points position!!!

My post was really around why I do these projects.

Regarding cheating - I strongly disagree that optimized libraries such as Crunch3r's old optimized code, which is just a better design, should be labeled cheating I would say cheating is when someone friggs with their benchmarking figures so much so that they don't actually benefit the project whatsoever and are just in it for the points - as an example the guy who gained 775,040,862,421.17 credits for one 11 minute WU in HashClash that time, now that's cheating!

gatekeeper53

10-23-2006, 09:59 PM

I agree anything that speeds up my machine therefore speeding up the results to the project to me is fair game. I quit crunching RS because the wu's didn't run well on my setup but gave me a lot of points. If it's not doing some good there is just no fun in it. I enjoy squeezing all I can get out of my machine and if all there was, was to install the boinc client and let it crunch I don't think I would have been here long. The way it is the projects benefit and I have a good time.

Brucifer

10-23-2006, 10:28 PM

But, the notion that you can make a points system that no one would get around is laughable. If you could write something that bulletproof you are wasteing your time and should be writting security programs.

That wasn't the point I initially was getting at. It was one of *berkeley's* goals. My point is that the current boinc points system sux, and would be better served by just using a fixed point system. As for making something people can't get around so easily, all they need to do is use a crypto based work unit naming/assignment routine. Nothing that can't be done now, and it isn't going to get cracked in a day or two.

drezha

10-23-2006, 10:56 PM

But, the notion that you can make a points system that no one would get around is laughable. If you could write something that bulletproof you are wasteing your time and should be writting security programs.

I must admit I'm yet to see someone get around the FAH points system.:icon_salut:

The benchmark machine data already also means that some projects get a nice bonus on it and others suffer a bit but it adds to the game:blob3:

Evil-Dragon

10-23-2006, 11:11 PM

Oh dear, i've dugg up a can of worms on this one. I take back me using the word "cheating". This arguement was not intended, i was merely commenting on what had been said about Crunch3r's BOINC on project forums. I'm hoping once the BOINC code is optimised and released properly then hopefully it will put an end to the "Optimised/Unoptimised" client threads that are apprent on practically every project forum.

At least we can be happy in the knowledge that AMD processors have been a cut above Intel for a long long time (even though the Duo is gaining good benchmarks) I'm still hopeful about AMD's Quad processor system... i still think it's gonna blast Intel out of the water.

My apologies for this :icon_cry:

Evil-Dragon

11-01-2006, 09:31 AM

To help this thread rather than cause more problems on it, please find a link to more optimised clients:

http://boese.dnsalias.com:6969/

I'm using the 5.7.0 version.

Lagu

11-01-2006, 04:06 PM

There must be somthing wrong. Have you type in wrong words? I can´t load this page.:icon_cry:

Evil-Dragon

11-01-2006, 04:24 PM

Nope thats the url Lagu, works for me in firefox.

daddygeek

11-01-2006, 04:47 PM

I can´t load this page.:icon_cry:

Nor, can I

Evil-Dragon

11-01-2006, 05:34 PM

I think i'll just upload them to my ISP webspace. Please don't hammer this link otherwise i'll have to remove it (my ISP will get annoyed)

http://evildragon.myby.co.uk/boinc/boinc_5.4.11_optimized_win_client(binaries_only).r ar (http://evildragon.myby.co.uk/boinc/boinc_5.4.11_optimized_win_client%28binaries_only% 29.rar)
http://evildragon.myby.co.uk/boinc/boinc_5.7.0_optimized_beta_win_client(binonly).rar (http://evildragon.myby.co.uk/boinc/boinc_5.7.0_optimized_beta_win_client%28binonly%29 .rar)

gamer007

11-01-2006, 10:51 PM

How's the 5.7.0 optimized client benchmarks compared to non-optimized?

mitchellds

11-14-2006, 03:07 AM

Gamer,

Are you using that 5.7 version along with Crunchers 5.5 windoze code, or just the 5,7 client? I see that using the 5.7 dropped benchmark scores compared to the 5.5 cruncher code. What have you found ?

gamer007

11-14-2006, 03:17 AM

I'm still using windows 5.4.9 client with Crunch3r's 5.5.0 optimized. I've been happy with Crunch3r's optimized clients for the last while. :)

NeoGen

11-14-2006, 07:06 AM

Me too... I'll probably only move away from Crunch3r's when boinc v6 is out. :)

Evil-Dragon

11-14-2006, 08:55 AM

5.7.0 is a champ and optimised too. I've been using it for weeks with no issues what so ever. Unfortunately it does not include the simple interface till 5.7.2.

mitchellds

11-14-2006, 02:49 PM

Hi Evil-Dragon.

So did you find that the 5.7.0 code is less optimal or more optimal than crunchers 5.5 code? It seems to be less optimal from my first couple of days running it.

gatekeeper53

11-14-2006, 03:39 PM

I tried it for a couple of days and it cost about 9 points an hour per CPU.

Lagu

11-16-2006, 08:05 PM

Talking about optimised clients:

Riesel Sieve has several.
Seti has several too.

But I have pondered if these optimised clients are specially developed for their own tasks? I have Cruncher´s client on my AMD and it seems to work with every project. If I prefer an optimised client from SETI or RS will it work with other project too?

NeoGen

11-16-2006, 08:35 PM

There's a little difference there... Crunch3r's boinc is an optimized boinc client while RS or SETI have optimized applications.
Optimized applications speed up real processing of workunits (for example from 40 mins to 30 mins), while the optimized client only improves the benchmarks of your machine.

EDIT: And yes, optimized applications are only suitable for their own projects. You can't use the SETI application in no other boinc project

spikey_richie

11-17-2006, 02:56 PM

I got a new company laptop this week, but I got stuck with a Celeron M

It's the Yonah core, which supports SSE2 and SSE3. Core speed is 1727mhz

Which optimized client would be best for this core?

P.S. I just updated my AMD 64 3400+ (which is running oc'd) to an optimized client and doubled by benchmark results

PcManiac

02-25-2007, 09:49 PM

http://calbe.dw70.de/

here are the latest clients from Crunch3r's clients I believe

I am going to try out the 5.9.0.32 one...

Edit:
OK here is the difference on my lappy:

Before:
Memory702.48 MB Cache976.56 KB Swap space1622.95 MB Total disk space74.53 GB Free Disk Space16.23 GB Measured floating point speed1313.43 million ops/sec Measured integer speed2508.88 million ops/sec Average upload rate5.83 KB/sec Average download rate44.3 KB/sec Average turnaround time0.2 days Maximum daily WU quota per CPU100/day

After:
Memory702.48 MB Cache1953.13 KB Swap space1622.95 MB Total disk space74.53 GB Free Disk Space16.18 GB Measured floating point speed3394.89 million ops/sec Measured integer speed7239.11 million ops/sec Average upload rate5.83 KB/sec Average download rate44.3 KB/sec Average turnaround time0.2 days Maximum daily WU quota per CPU100/day

gamer007

02-25-2007, 10:22 PM

Tried them a while ago on NanoHive. Doesn't like it. It finishes WUs in minutes when it's supposed to be finished in hours.

drezha

05-01-2007, 01:59 PM

Will this work on normal PCs?

I want to use it on ym Windows box bcause looking around, my 3500+ was being beaten by a 2800+ Semperon!

Mine (http://boinc.bio.wzw.tum.de/boincsimap/show_host_detail.php?hostid=68332)
Semperon (http://boinc.bio.wzw.tum.de/boincsimap/show_host_detail.php?hostid=48354)

Obviously if it reports wrong results I dont want it.

AMD-USR_JL

05-01-2007, 08:37 PM

Crunch3r's boinc won't make you crunch any faster than normal, it will just inflate your benchmarks, making you request more credit per WU.

If you want to crunch faster though, Akosf made some optimized apps for SIMAP somewhere in this thread.
http://boinc.bio.wzw.tum.de/boincsimap/forum/viewtopic.php?t=96&highlight=akosf

I'm not sure if they still work or if they have gotten integrated into the offical SIMAP app, but you can try it. Just make sure to make a backup of your BOINC directory before doing anything.

drezha

05-01-2007, 09:19 PM

I was being beaten by benchmarks. I know I cant crucnh faster.

Yes the appilication in SIMAP did take on board his optimiseastions.

Currently running the truxcroft client now actually.

AMD-USR_JL

05-01-2007, 10:15 PM

Try Crunch3r's then. Just look at how much higher my 2.6 GHZ P4 is than your 3500+.
http://boinc.bio.wzw.tum.de/boincsimap/show_host_detail.php?hostid=66311 When you install crunch3r's it should be even higher than that.

I had problems with Crunch3r's 5.9.0.32, i think it's a developmental build. The stable version has worked perfectly though.
http://calbe.dw70.de/boinc/5.8.11.32_windows_intelx86.zip

drezha

05-01-2007, 10:20 PM

Ah I'm fine with the Truxcroft for the minute. ;)

I installed that and whilst not technically optimised, it got doubled my integer speed. (but my floating points remained similiar)

I'll leave this going foer a bit, see what happens, and then try the cruncher one :)

EDIT: Just noticed the Cruncher one can return result immdiatly as well..I'll give it a whirl now :)

drezha

05-01-2007, 10:44 PM

Double post BUT the crunchers client was awful!

Ran it twice and got:

01/05/2007 23:37:23 3779 integer MIPS (Dhrystone) per CPU
01/05/2007 23:37:23 2018 floating point MIPS (Whetstone) per CPU

01/05/2007 23:35:25 3752 integer MIPS (Dhrystone) per CPU
01/05/2007 23:35:25 1998 floating point MIPS (Whetstone) per CPU

Running the Truxcroft client, I get
01/05/2007 23:40:43 7347 integer MIPS (Dhrystone) per CPU
01/05/2007 23:40:43 2150 double precision MIPS (Whetstone) per CPU

I'll stick with the Truxcroft core.
http://boinc.truxoft.com/

apple1

11-29-2007, 10:01 AM

HELLO!
thanks,it's very clear.I like this.
thanks:(

Borfil

03-24-2008, 04:11 AM

How you find that core, is it stable?

vaughan

11-20-2008, 10:41 PM

Download and install the faster "optimized" client from here (http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=433).

Pick the "Alternative download location" link if using Windows (haven't tried the Linux version) and it should take you to a page where you can download a 120Kb sized file in Zip format. Stop BOINC. Unzip the file into the Milkway folder in your BOINC Projects folder. Easy to find this folder location if you are using BOINC v5.x.x but darned difficult to find if using BOINC v6.x.x . Now re-start BOINC and allow Milkway to get work.

The server is getting hammered but eventually you should get some tasks. My Intel T2300 1.6 core duo laptop has gone from 8 hours a task to approx 12 minutes :icon_eek:

Borfil

11-25-2008, 11:04 AM

Wow, that is a real big change. even eliminating the graphics and cleaning the code. you are talking of 1oooo% increase. that mean that either the MW wu's and / or boinc were not as compatible. Are you getting good results, i mean the end product of our work precise or does it just passing the chksum and been validated, but the data later is trash? just thinking, because it so big the diff. I will try to down the linux version to see what it does in one of the old 1.7 single core i have.
tnks Vaughan
Rafael

vaughan

11-25-2008, 11:25 AM

Hi Borfil - the MilkWay project admins keep playing with the apps and the points awarded. The optimised app helps slow old computers but is useless with modern fast multicore CPUs. They say they are going to release a new improved version of the official optimised app but it hasn't appeared yet. My Q6600 takes about 7.5 hours on each core to run the current MW tasks. My old core duo 1.66GHz laptop took much less time to complete these tasks with the optimised app. The points were awarded so I presume the "science" was OK.

Borfil

11-26-2008, 05:38 PM

hi, I Triedthelinux version on an older Laptop. di d not get but a few seconds imporvemet in 1 and 2 crash. so return to the I686 version.

vaughan

12-16-2008, 10:50 PM

Not sure if we have covered optimized SETI clients but they are available here. (http://lunatics.kwsn.net/) :)