Optimize Windows?

[LP]

I'd like to get the opinion of those that run Boinc here.

Optimized Windows Boinc- fair or not?

I'm just wanting your opinion.

[daddygeek]

Yes, it is fair

If you run Linux, Optimized it is the only way to run. Unless you like getting 50% less credit than windows.

[LP]

I didn't say Linux. I said Windows.

I agree with running Linux optimized. It corrects an bug.

I'm asking because it seems unfair to me , for someone to inflate the benchmark in Windows and get 3 times the credit, same as under-value-ing the bench in Linux gives you 50% of what Windows gets.

I'm not going to run Boinc if it's not on an even playing field.

[NeoGen]

Yea, optimized is the way to go, especially on projects where WYCIWYG (What-You-Claim-Is-What-You-Get)

All is fair in Love and Distributed Computing! :lol:

[Keith75]

From everything I have read about it the optimized version is fair and it is the regular version that is screwing you out of points since it doesn't take advantage of your chips features. Be like saying your car should drive with two flat tires in a race designed for motorcycles.

[NVM]

i just modified/hacked DNET so i'm really pulling in the points now,almost double. next i'm going to modify SOB and try F@H as well. is that fair?

[Ototero]

The difference is that if the optimised code is available (publically) to anyone that wants to use it, then that is fine.

If I hack my own version just for my benefit, that is cheating.


Just my opinions.

[LP]

From everything I have read about it the optimized version is fair and it is the regular version that is screwing you out of points since it doesn't take advantage of your chips features. Be like saying your car should drive with two flat tires in a race designed for motorcycles.

It was explained to me that all it does is inflate your benchmark. It doesn't make any more use of your cpu, make it run faster, better, etc, than the regular version.

If it did, as in using flags in F@H turns on SSE, better utilize your cpu, then I could see using it. But not if all it does is inflate your benchmark.

[NeoGen]

But the optimized client does just that what you said.
Check Crunch3r's page of optimized boinc clients. It has boinc versions with SSE and SSE2 enabled.

http://www.guntec.de/Crunch3r/boincx86.html

Seeing that the client doesn't actually do the work, then the end result for the user is an inflated benchmark, so by saying, but the benchmark values are much more in accordance to our machine's real capabilities.

[LP]

Wait til you guys see the claimed credits coming in on Linux machines (mine included) and that they are using "optimized" Windows machines as part of the quorum of 2, in Simap.

I'm waiting for the crying to begin. :lol:

[LP]

Neo, the way it was explained to me, is that unlike F@H, which uses the flags to tell it use certain features of the cpu (SSE2, etc) to actually crunch with, in Boinc programs other than Seti all it does is use it for benchmarking and not to actually do any work with.

I'm finding out that whether they use the quorum system and how many they use for a quorum, will even out some and make others very tilted.

Here's a question to ponder- Linux has a bug that gives an artificialy low benchmark, so they recommend you "optimize" Boinc for Linux. That brings up to an even level with what the same machine would bench if it ran Windows. But even with which version of Windows Boinc? The official Boinc version, I'm assuming. Meaning that those running and optimized Boinc in Linux think they are running on a level playing field, but if they are up against optimized Windows Boinc, then they are in fact being benched at about 1/3rd what the optimized Windows Boinc people are benching?

edited for typo

[NeoGen]

Yes, the SSE enabling on the Boinc client is only for the benchmarks, to give a more accurate display of our machine's power.
To enable SSE/SSE2 in projects, the applications would have to be recompiled accordingly. Which is done only for SETI, and by enthusiasts (not an official thing), as far as I know.

About Linux and Windows claiming credit values... can anyone with several machines/OS'es give it a try and tell us how it goes?

[NVM]

The difference is that if the optimised code is available (publically) to anyone that wants to use it, then that is fine.

If I hack my own version just for my benefit, that is cheating.


Just my opinions.

so, if i hack a client and post it on the 'Net, its ok then?

[LP]

If nothing else, this is an interesting and hot topic, and has piqued my interest in other DC projects, like DPAD. :lol:

[NeoGen]

The difference is that if the optimised code is available (publically) to anyone that wants to use it, then that is fine.

If I hack my own version just for my benefit, that is cheating.


Just my opinions.

so, if i hack a client and post it on the 'Net, its ok then?

NVM, in optimizing an application nothing is changed. You just feed the source code (downloaded from the website) to a compiler with flags enabled to use MMX/SSE/SSE2/etc and the compiler spits out an .exe perfectly optimized and much more efficient.
That is what I consider to be optimization. But now if the person does alter the code for its own benefit before compiling it, say for example adding intructions to multiply x10 the output of the benchmark values, so to claim more credit per workunit, then that IS hacking.

[BlackAdder]

I run optimized clients too, but you are still only get the middle credit when it takes three completed wu's to get credit. On the other one's I think it is fair since everyone else has the same ability to use the optimized clients if they know about them. :D

[daddygeek]

AMD Athlon(tm) 64 Processor 3700+
Memory 1023.48 MB

Win 2000 official Boinc version 5.2.13
_______________________
Measured floating point speed 2254 million ops/sec
Measured integer speed 4185 million ops/sec

Win 2000 optimised Crunch3r 5.2.13
_______________________
Measured floating point speed 3287 million ops/sec
Measured integer speed 10008 million ops/sec

SuSE 10 official Boinc version 5.2.13
_______________________
Measured floating point speed 1236 million ops/sec
Measured integer speed 2265 million ops/sec

SuSE 10 optimised Crunch3r 5.2.14
_______________________
Measured floating point speed 3519 million ops/sec
Measured integer speed 5766 million ops/sec

Need more?

[NeoGen]

Great job there daddygeek! You had the patience to test all 4 combinations! :)

The optimized apps seem to match the numbers on floating point math, but for integers windows still makes twice more. Strange...

[cicide]

Well, I took the time to update all my linux clients to the optimized boinc clients, and WOW. My estimated daily output (for climate prediction) went from ~3500 to ~8000 credits.

[Nflight]

Now I have a question? Is there anything to run a Solaris 9 on?
I have one sitting here staring me in the face per say, and I would like to energize my effort with its capability but alas I have seen little mentioned in the realm of SUN!

Is there a link to a some help or would we have some people on here that have some workings with Solaris?

(Just to mention it is said that Solaris machines run quite a bit faster then Windows)

[daddygeek]

http://www.lb.shuttle.de/apastron/boincDown.shtml#sparc

http://boinc.vawacon.de/

One of these days, I will give Solaris a try.
But it looks like my next project is SuSe 10 enterprise server :lol:

[Nflight]

Thank You very much daddygeek, I also have a laptop with SUSe 9 on it but the processor speed is slower then molassas in January! SO I bet free hand I could use a calculator faster then it can count WU's.

Thanks Again

P.S. No need to ask wikipedia questions we have a Master Level Group of Answer machines in our members! :)

[NeoGen]

We're like a distributed knowledge repository.
Each person here knows a bit about something. If you put us all together you can know alot about almost everything. :)