ChrisD,

Having (official "inner circle") access to the optimized app, I am not even using it at this moment, as there's a new Seti Enhanced version coming out at the end of this week. It'll be just plain nonsense to start running 5.05 optimized now and go on a "forget it" mission as the optimized app will NOT auto-update itself. That's a hands on job only, aka you have to do it yourself.

Not exactly the way Beta works.

Beta isn't about fast credits. If you want fast credits, go back to Seti proper and use one of the optimized apps there. If you're only here to get the "volunteer tester" tag, then so be it. But you may want to check up on what Beta means.

---

IMHO regarding a forthcoming relase of the enhanced app. somewhen in the futhure, we better get prepared to provide an optimised app. for the most common OSes, because i think there will be a lot of complains about the long crunching times and we allready seen the behavior here that some are aborting WUs which take "to" long to process in favour of getting short ones.

I think that the aborting of long WUs should be not so often seen if the shorter WUs will claim an appropriate amount of credits/hour, similar to that of longer WUs. Maybe someone will do it once, twice, but if there will be really no motivation to do it...

Peter

ChrisD,

Beta isn't about fast credits. If you want fast credits, go back to Seti proper and use one of the optimized apps there. If you're only here to get the "volunteer tester" tag, then so be it. But you may want to check up on what Beta means.

Ageless!!

This is just too much. This is personal insult!!

I have crunshed SETI WU's since 1999 and at times had 7 machines running. I was forced to reduse the machine count as the electricity bill went skyhigh.

I thought we might have some fun while doing some serious science, but this thread, that wasn't started by me anyway, has turned into a personal attack on me!!

Last time the INTEL Client was changed we all had to abasndon WU's and load new ones. What is the diff??.

Next, If You don't want anybody to ask questions about this 'optimized client', why do You then have to brag about its existence and how fast it works?

Just in this thread alone:

By Mike:

We have an optimized app in inhouse testing.
All looks good so far.

You can expect that with this application you will need less than a half than with the official app.
My machine needs around 13 houres for a 0.189 WU.

and Winterknight:

Are there any optimised clients for the Enhanced project?
I have returned this w/u and it appears slow in comparison with
anothermachine running the same processor!

http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=166974

What is the average time expected for an AMD Athlon64 3500+?

There is a small group testing an optimised version and yes that compter was one of them. If you click on the resultID number the details are shown.

I do not want to drag Mike and Winterknight into this, but their posts are just in this thread alone.

Remember: Lord Tedric started this post asking about the running times. I just pitched in.
Now it seems that I am no longer wanted here.

OK, so be it, but maybe Eric has something to say about this...

ChrisD

---

I think that the aborting of long WUs should be not so often seen if the shorter WUs will claim an appropriate amount of credits/hour, similar to that of longer WUs.

I don't care how long a WU takes to run but it should earn an equivalent amount of credits. In my recent history running the standard un-optimized app I've seen this for credits earned:

Long WU = 4.9 credits/hour
Short WU = 16.9 credits/hour

For each long WU I run I will not recieve 661.9 credits that would have been earned by running short WUs instead.

Something really needs to be done about the credit calculation. I think we've already seen many people aborting long WUs and it's not just because they're time consuming. People do seem to be motivated by the credits received and really don't care that aborting long WUs adversely affects the Seti science progress and is not polite behaviour with respect to the other users.

There is a fine balance between the appearance of fairness and attracting users to do the work. An unequal credit allocation will be percieved as unfair and will cause many people to leave the project or refuse to crunch long WUs.

DC project staff don't get excited about handing out credits because they have a credit making machine in the server closet. They care about attracting users to do their work. At rosetta@home the staff seem to have decided to give whatever credits you ask for. Many people consider this unfair and I think that the result is a relatively small number of users. I believe the appearance of fairness is important.

If a credit adjustment needs to be made regarding the short/long WU discrepancy, the adjustment should be made at the server side. Allowing the Boinc client to make this adjustment should not be permitted because that will result in the appearance of more 'customized clients' that, in my opinion, lie to the server about the work they've done.

All I can say is, ChrisD you need to growup real fast....

---

BOINC Wiki

Bob Guy

Nothing bad... This is also a part of why we are here...

This information should also be in Calibrating Credits, Joe is working with Eric to help make the credits claimed more "correct." So Yes what my Computer ID 2924 can turn out in Seti or Einstein and what it claims and recives in Seti Enhanced is two (or three) different stories (Seti and Seti Enhanced)...

That was part of the Reason I started the New 0.xx Ar Threads to allow different users to identify what they were seeing because of the time/credits involved with WU's with different Angle Ranges (AR)...

I think that the aborting of long WUs should be not so often seen if the shorter WUs will claim an appropriate amount of credits/hour, similar to that of longer WUs.

I don't care how long a WU takes to run but it should earn an equivalent amount of credits. In my recent history running the standard un-optimized app I've seen this for credits earned:

Long WU = 4.9 credits/hour
Short WU = 16.9 credits/hour

For each long WU I run I will not recieve 661.9 credits that would have been earned by running short WUs instead.

Something really needs to be done about the credit calculation. I think we've already seen many people aborting long WUs and it's not just because they're time consuming. People do seem to be motivated by the credits received and really don't care that aborting long WUs adversely affects the Seti science progress and is not polite behaviour with respect to the other users.

There is a fine balance between the appearance of fairness and attracting users to do the work. An unequal credit allocation will be percieved as unfair and will cause many people to leave the project or refuse to crunch long WUs.

DC project staff don't get excited about handing out credits because they have a credit making machine in the server closet. They care about attracting users to do their work. At rosetta@home the staff seem to have decided to give whatever credits you ask for. Many people consider this unfair and I think that the result is a relatively small number of users. I believe the appearance of fairness is important.

If a credit adjustment needs to be made regarding the short/long WU discrepancy, the adjustment should be made at the server side. Allowing the Boinc client to make this adjustment should not be permitted because that will result in the appearance of more 'customized clients' that, in my opinion, lie to the server about the work they've done.

Al

---

Thanks to Paul and Friends
Please consider a Donation to the Seti Project

For those that can see me/Crunch3r and crew and wonder what TMR is up to...

166417

This workunit shows one of TMR's machines... Computer ID 4495 a 2.8 GHz Intel
My computer for this one is Computer ID 4995 which is an AMD 64 3200 @ 2.2Ghz...
The third machine Computer ID 4012 is an AMD 64 3200

So there is other testing besides this small group...

Has anyone seen 5.06 yet?

Al

---

Thanks to Paul and Friends
Please consider a Donation to the Seti Project

Yeah, mine is almost the same as official ones except the name tag, result dump (sometimes do this because older days there often used to be discrepancies between hosts. Since version 4.09 I haven't seen discrepancies at all), disabled graphics, and disabled memory check. I do not tune so eagarly that may affect the result. Around version 4.09 I found ICC might miss some of Gaussians when aggressively optimized and I sent a fix to Eric as "inaccuracy proof" and it's implemented in the official source (it's not commented in the source). So mine can be regarded as plainly "compiled with ICC and IPP" and anyone can reproduce it.

And note my 2.8G P4 is HT enabled; ie crunching 2wu's at once. That's how ICC/IPP work for Intel processors well...

Another interesting thing is my 2.4G (running at 2.5GHz) single threaded P4 takes almost the same time for a wu as 2.8G HT;ie the speed is almost half of that of 2.8G HT! Maybe it's because 2.8G P4 has dual channel memory while 2.4G P4 has single channel and smaller cache. For Einstein@home the difference isn't so large.

EDIT: some additions

---

Luckiest in the world. WMD = Weapon of Mass Distraction

TMR

Thank You!

You have been very quiet and left the talking to me... I cut my teeth on your Optimized Apps...

So, I presume that this means as the Enhanced App moves Users will have options... Or can we hope that part of this will be rolled into the Current Enhanced Application?

Yeah, mine is almost the same as official ones except the name tag, result dump (sometimes do this because older days there often used to be discrepancies between hosts. Since version 4.09 I haven't seen discrepancies at all), disabled graphics, and disabled memory check. I do not tune eagarly. So it can be regarded as plainly "compiled with ICC and IPP" and anyone can reproduce it.

And note my 2.8G P4 is HT enabled; ie crunching 2wu's at once. That's how ICC/IPP work for Intel processors well...

Another interesting thing is my 2.4G (running at 2.5GHz) single threaded P4 takes almost the same time for a wu as 2.8G HT;ie the speed is almost half of that of 2.8G HT! Maybe it's because 2.8G P4 has dual channel memory while 2.4G P4 has single channel. For Einstein@home the difference isn't so large.

Most old crunchers know that Graphics means less workunits... Then Memory bandwidth is one thing that Eric owes us graphs for... will have to bump that thread...

Al

---

Thanks to Paul and Friends
Please consider a Donation to the Seti Project

some additions:

What I'm up to? Recently I am up to supporting Japanese woman speed skater Tomomi Okazaki at Torino Olympic, who missed the medal at 500m race by 0.05 second (2 feet) in total time!! She will have 1000m race, though she is usually better at 500m race. I won't watch the next race live because it'll be in the midnight here in Japan. I'm her fan since she had won the bronze medal at Nagano Olympic 1998.

These days I use Intel C++/Fortran compilers for scientific programs and multimedia programs such as scilab, octave, xmms, xine, lame, transcode, etc. Intel compilers are more accurate and faster than gcc/g++/g77 with proper options. But I found Intel Fortran had a bug with -xP option (sse3) and reported it to intel Premier Support. Intel had reproduced the bug, so I guess it will be fixed, but it's another story.

And I have to go to the class day of my younger daughter's kindergarten next Friday.

getting close to crap? Not to me :)

TMR

Applause! Yes sometimes Science gets in the way of Life...

The Balance between "what we want and what we do," gets to be hard at times...

I think the Important Part is that You be there with Your Daughter! Science can wait!

some additions:

What I'm up to? Recently I am up to supporting Japanese woman speed skater Tomomi Okazaki at Torino Olympic, who missed the medal at 500m race by 0.05 second (2 feet) in total time!! She will have 1000m race, though she is usually better at 500m race. I won't watch the next race live because it'll be in the midnight here in Japan. I'm her fan since she had won the bronze medal at Nagano Olympic 1998.

These days I use Intel C++/Fortran compilers for scientific programs and multimedia programs such as scilab, octave, xmms, xine, lame, transcode, etc. Intel compilers are more accurate and faster than gcc/g77 with proper options. But I found Intel Fortran had a bug with -xP option (sse3) and reported it to intel Premier Support. Intel had reproduced the bug, so I guess it will be fixed, but it's another story.

And I have to go to the class day of my younger daughter's kindergarten next Friday.

getting close to crap? Not to me :)

Al

---

Thanks to Paul and Friends
Please consider a Donation to the Seti Project

TMR

Applause! Yes sometimes Science gets in the way of Life...

The Balance between "what we want and what we do," gets to be hard at times...

I think the Important Part is that You be there with Your Daughter! Science can wait!

Al

Thanks!

And another addition, I'm self-studying physics (Quantum physics and relativity) recently out of interest. Intel compilers are used for scilab like this picture which describes the probability density of an electron around hydrogen atom at state (n,l,m) = (3,2,1). But this doesn't require speed so much for me.

What a phreak I am...

And note my 2.8G P4 is HT enabled; ie crunching 2wu's at once. That's how ICC/IPP work for Intel processors well...

Another interesting thing is my 2.4G (running at 2.5GHz) single threaded P4 takes almost the same time for a wu as 2.8G HT;ie the speed is almost half of that of 2.8G HT! Maybe it's because 2.8G P4 has dual channel memory while 2.4G P4 has single channel and smaller cache. For Einstein@home the difference isn't so large.

EDIT: some additions

Hi TMR,

I've seen that too.
While using caching functions, it's really important to have high memory bandwith.

My P4 3.2 GHz running at 3.75 GHz really suffers from having only a single channel
setup.

regards
Crunch3r

---

ChrisD,

Beta isn't about fast credits. If you want fast credits, go back to Seti proper and use one of the optimized apps there. If you're only here to get the "volunteer tester" tag, then so be it. But you may want to check up on what Beta means.

Ageless!!

This is just too much. This is personal insult!!

I have crunshed SETI WU's since 1999 and at times had 7 machines running. I was forced to reduse the machine count as the electricity bill went skyhigh.

I thought we might have some fun while doing some serious science, but this thread, that wasn't started by me anyway, has turned into a personal attack on me!!

Last time the INTEL Client was changed we all had to abasndon WU's and load new ones. What is the diff??.

Next, If You don't want anybody to ask questions about this 'optimized client', why do You then have to brag about its existence and how fast it works?

Just in this thread alone:

By Mike:

We have an optimized app in inhouse testing.
All looks good so far.

You can expect that with this application you will need less than a half than with the official app.
My machine needs around 13 houres for a 0.189 WU.

and Winterknight:

Are there any optimised clients for the Enhanced project?
I have returned this w/u and it appears slow in comparison with
anothermachine running the same processor!

http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=166974

What is the average time expected for an AMD Athlon64 3500+?

There is a small group testing an optimised version and yes that compter was one of them. If you click on the resultID number the details are shown.

I do not want to drag Mike and Winterknight into this, but their posts are just in this thread alone.

Remember: Lord Tedric started this post asking about the running times. I just pitched in.
Now it seems that I am no longer wanted here.

OK, so be it, but maybe Eric has something to say about this...

ChrisD

No personal attack intended but they are hardly bragging at all. Lord Tedric asked about optimized clients so they just provided basic information, saying that yes there is testing going on and then trying to be nice and give a general idea of how it seems to be going and the overall difference it seems to be making so far. If you think that's bragging then I recommend clicking here Definition of Bragging

I have to admit that I tend to agree that creating an optimized client should probably wait until after Beta as there will be less quirks that pop up but if the testing and bug reporting of the optimized client helps in the development of the official client then I'm all for it.

How did I get involved in all this. I answered Lord Tedric by stated yes, there was an optimised app being tested by a small group.

BUT I AM NOT ONE OF THAT GROUP, my faster than average times is because I run a fast cpu on a good mobo and know how to tweak it for best performance.

Andy

I started this thread with a simple question as to whether or not optimized clients existed for Seti Beta - this question was answered a long time ago and I am happy with that.

Lets end this thread now as the origonal question was answered and the resultant messages are technically off thread!.

---

jsprague78

Welcome to Seti Enhanced Beta

Al

---

Thanks to Paul and Friends
Please consider a Donation to the Seti Project