News:

If you are a member of the Team on BOINC you still need to register on this forum to see the member posts.  The posts available for visitors are not posted to much by members.
 Remember to answer the questions when Registering and also you must be a active member of Team BOINC@AUSTRALIA on BOINC.

Main Menu

AQUA Project Discussion

Started by Dataman, August 12, 2009, 03:06:09 AM

Previous topic - Next topic

Dataman

I must say that I have always had some strong reservations about this project. Normally I do not run this type of project as my interest tends to be in biology, chemistry and physics. I think I understand what they are trying to accomplish but, in my opinion, I think my efforts are put to use better in projects investigating cures and a better understanding of the physical world. Again ... just my opinion and personal preference.
I've also had some problems with how MT was introduced and its effect on other projects running with AQUA. BOINC has never been too good at handling Long Term Debt (especially with GPU) and MT still seems to get more than its "fair share" of runtime (at least on my farm). From the "for-profit" nature of the company, the questionable marketing tactics and the science itself, AQUA leaves me lacking.
I have been running AQUA solely because I want to see us solidly in 9th position world wide. While shameful in some respects, it does satisfy some aspect of my competitive nature.  biggrin However, when we are there, I will be leaving AQUA to go back to some projects in which I have more interest.
I would like to hear what others think about the project, its science and tactics. Thanks!


kashi

I understand what you are saying about having doubts sometimes about being influenced by credit rate in your choice of project. However that is the advantage of BOINC, you are free to choose whichever project you wish to support. With home computers you are the one who pays, so it is your choice. All projects are happy to accept your contribution of computing resources and power usage.

These doubts can arise both ways. Not only if you are scoring higher credits on generous projects while neglecting projects that you consider to be more beneficial for the health and wellbeing of many people. But also if the projects of your choice have poor credit then your personal ranking and that of the team may go backwards, your competitive spirit may be bruised and you may not be as motivated to upgrade to faster computers and/or GPUs.

Therefore when doubt is involved you can never win. Better to make your choice and not have second thoughts. If you are hard on yourself then you may be tempted to judge others. Don't have regrets but crunch on and enjoy the positives. For generous projects revel in the credit that is rapidly swelling your total and that of the team. Have fun and go to town. For more humanitarian type projects be proud of your contribution signifying not just credits but generosity of spirit and have faith that it will help achieve a better world long after the credits are forgotten.

Personally I prefer higher fixed credit projects but still do a bit of some lower paying benchmark based projects now and again. What I find most enjoyable of all, regardless of credit rate or quality of science is supporting the team in an AA. Once the vote is counted it removes any need to choose a project for myself for 3 weeks. The team has spoken and it is onward and upward time.

I don't know about the science of AQUA, but the introduction of multithreaded applications to BOINC is interesting. Although there have been teething troubles, now that I have discovered how to specify the number of cores AQUA uses it is very useful for scheduling to run with MilkyWay ATI. This configurable CPU usage percentage would also be handy on heatwave days as it is much more convenient than having to go into the BIOS and reduce the clockspeed every time there is a scorcher. Multithreading would be brilliant for projects like CPDN where contention issues cause a severe slowdown or instability when running tasks on more than 2 cores. Also it enables greater efficiency on recent and upcoming Intel CPU models that use hyperthreading and have a smaller amount of L2 cache.

Tixx

i feel boinc should police all projects credit allowances.. if anything they should level the platform and make every project allocate same amount of credit.. that way people would be less inclined to chase high numbers, but take general interest in the cause of the project they crunch for..

Wang Solutions

I think it all tends to balance out - when I am crunching a high paying project, that gives me more opportunity to crunch more favoured, low-credit projects without harming my output so much. So while I am dong a bit of AQUA, I can do more of WCG, Malaria, Spinhenge etc than I would otherwise be able to "afford" to do. Without the high paying project, I would have to take some resources off the lowest payers and put them on higher payers to keep my output "competitive". (Not that I have any hope of being competiitive with the likes of Beakerulz!)  biggrin

Mike Mitchell

I think the credits should be produced from a central point, so every project can at least have a chance of getting equitable reward. It may even be a way to reduce the cheating by some users too.
AA's > 1-Malaria 2-Tanpaku 3-Riesl Siev 4-Seti 5-ABC 6-Einstein 7-WCG 8-Seti 9-QMC 10-WCG 11-Cosmo 12-ABC 13-MilkyWay 14-3x+1 15-Rosetta 16-ABC 17-MilkyWay 18-Einstein 19-WCG 20-WCG 21-Poem 22-Rosetta 23-Docking 24-Spinhenge 25-Alternate 26-Simap 27-Alternate 28-Constellation 29-WCG 30-Edges 31-Alternate 32-Pogs 33-WCG 34-Seti 35-Pogs 36-Poem 37-Pogs 38-Asteroids 39-Pogs 40-Simap 41-Pogs 42-Seti


Tixx

yeah.. if you take the factor of varied credit allocation out, then better benchmarking / competition will prevail. that way having better hardware and more of it would give you a better advantage then others.

SkyStation

I have to say I totally agree, I don't like it that crunching on WCG gives a lot less credit than if I devoted the systems to AQUA (much prefer WCG), while haven't done much reading on AQUA I see that they are going to make results public even tho they are commercial company and the results they will be generating will help other entities who produce or are considering producing quantum systems can benefit from. It has to be hard for a commercial company to be producing something so much on the bleeding edge and while not for a community or academic project our crunching may be allowing a company to explore the bleeding edge whereas they might not have the resources to be able to and other companies may not even start.

If they are successful, other companies will sure follow, if they did not try could probably be stuck with academic institutions trying it, nothing against them but I think having a commercial company trying gives them a slightly different set of goals that may give them the motivation to succeed into making it commercially available instead of proving a theoretical point and both being worthy goals.

My 2c :)

Dataman

#7
I thought I would pass this on for what it is worth. I have 3 GTX260; all are identical PNY's, all clocked the same, all on Q6600 processors and win7 64-bit. They are all on BOINC 6.6.36 with nVidia 190.38 and are running GPUGrid. One is paired with Milkyway CPU and runs GPUGrid between 1k-2k seconds.  One is paired with CNPD and runs GPUGrid in 1k-2k seconds.The other is paired with AQUA and runs GPUGrid in 5k+ seconds. (no bonus points  :thumbdown:) When I suspended AQUA and started Yoyo it runs GPUGrid in 1k-2k seconds.  :thumbsup: I suspect AQUA MT is taking so much CPU that GPUGrid is not getting enough. Has anyone else noticed this?
:aus1:


veebee

Hmm, I have my i7 running AQUA and GPUgrid... (a GTS250) same with a Q6600 (GTX 275).

the AQUA credits are so worth it though.. my daily credits have gone wild since starting AQUA/ GPUgrid combination under WIN 7 x64.... the loss from dropping AQUA to increase GPU grid wouldn't be worth it (for me) I dont think...

kashi

Quote from: Dataman on August 24, 2009, 06:09:39 AM
.....I suspect AQUA MT is taking so much CPU that GPUGrid is not getting enough. Has anyone else noticed this?.....

Yes. I run AQUA on only 7 of 8 cores for just that reason. MilkyWay ATI pays a fair bit more than AQUA so this is the preferred configuration for my combination of projects.

Lack of CPU resources can slow down GPU processing quite a lot but I have never seen a slowdown as great as you mentioned. When I did Einstein on 8 cores, MilkyWay ATI processing time inreased by 50%. I now always leave a CPU core free to support the GPU.

It can make quite a difference on Collatz GPU too. I have looked at a few contributors there and noticed that it is quite common that they bottleneck their cards by not providing sufficient CPU resources. This is more noticeable on faster/higher model nVidia cards because it results in them processing tasks at the same speed or even significantly slower than slower/lower model nVidia cards.

One person who posted in the forums there was crunching Collatz on his dual core Opteron CPU as well as on his video card at the same time. I gave him the information to be able to work it out for himself but I thought it was kinder not to inform him that he would achieve greater efficiency and receive more credit if he left his CPU idle while he was crunching Collatz on his video card. He was confused about how the CPU time reported related to total GPU processing time and his way of thinking about BOINC was still highly CPU-centric, so it is quite possible that I would not have been able to make him understand anyway.

Many top producers on GPU applications do not share on the project forums the settings and tweaking tips that can help to enable the best performance. I don't know whether this is due to the competitive nature of BOINC, both on a personal and team level, the difficulty sometimes encountered in explaining things to people or just a lack of time.

It seems that many believe that GPU crunching is a bonus that is almost totally separate from what is being done on the CPU. Perhaps this is due to the way it has been marketed and promoted. Kind of like how BOINC is supposed to run only on spare cycles and so doesn't use any more resources than just having your computer on. Haha. In practice however sometimes the GPU applications are so efficient that running almost anything on the CPU just reduces total efficiency, particularly for an older dual core CPU.

AQUA is a CPU application that can be an exception to that, depending on the type and number of video cards in the box. On my setup with a Nehalem at default and a single HD 4890, AQUA gives about 33% of the credit of MilkyWay ATI but about 25% more than Collatz GPU. If I had an additional HD 48xx card however Collatz GPU would give more than AQUA. Fortunately I have the CPU resouces and correct settings to be able to run AQUA and a GPU application concurrently with minimum slowdown. The generous credit of some of these projects enables me to do a bit of WCG as well.

AQUA is superior to GPUGrid in credit rate so it may not be worth trying <ncpus>5</ncpus> option in a cc_config.xml file or configuring AQUA to run on 3 cores of your Q6600s to give more CPU support to your GPUGrid processing.

Dataman

Thanks Kashi. +1

QuoteKind of like how BOINC is supposed to run only on spare cycles and so doesn't use any more resources than just having your computer on. Haha

That has always given me a chuckle too ... just like electricity is free.  :rofl:

:aus1: