Unbalanced Scoring

[FordGT90Concept]

They attract bonus points for their 'betaness' and lack of maturity (hence need a large amount of monitoring and hand holding by the donator), and also provide a massive computational boost to the project.

Bonus points are just another flaw in the system. I know, for a fact, that in most cases, 8 applications is more efficient than 8 threads + 1 main thread. The main thread tears the thread running on the same core down by at least 2/3s while the individual threads show little improvement. This is observed on simple ADD and FPADD machine code. Again, the points system becomes extremely inconsistent across various hardware and software so, it is my conclusion that they serve absolutely no purpose anymore--except, of course, to annoy me.

You are not putting the time into your machines to keep them running and contributing to the project.

You are in no position to judge that.

What is stopping you from going out and grabbing an 8800GS/9600GSO for about $100 and adding 4Kppd to your output? NOTHING.

I don't buy hardware to fold. That statement basically means to me what a drop of water means to an ocean.

I am not all that happy that my highly OC'd quads (that I sunk alot of money into BTW) were being killed by a single 8800GT. So I went and bought a few GPU's, dropped the second SMP instance off and now my boxen do double their previous output.

By overclocking your system, you are vastly increasing the odds that an inaccurate result will be produced. Because WUs are linear, a fault early on would be multiplied on every subsequent WU produced. This is a fallout of relying heavily on a points system. It is like someone speeding to get to their destination quicker only to get in a wreck and be paralyzed for life. Remember that lovely story where the tortoise beat the hare?

They certainly do complete the WU's dished out well within deadlines, but there is one simple solution to slow machines. Tighten deadlines so that lesser machines cannot achieve them, then lock that Install ID from being assigned any more WU's after 5 missed deadlines. This has the affect of requiring a new install of the client to return it to active duty. I'd also like to see V5 clients not assigned any more work units in about 6 months time too.

That creates more work for the contributors, not less. It is F@H's responsibility to notify the user if the hardware no longer meets their needs so it can be removed from folding serivce. This could effectively be done by a little pop up message or a message added to the application event logs.

[bowman]

How does preventing the machines from doing work help find a cure for cancer? If client A is a GPU and client B is a CPU, client A is going to complete a certain number of WUs regardless of whether or not client B is working on WUs. However, if you take client B away, that's a certain number of WUs that aren't going to be used in order to contribute to the project. How does reducing the amount of added work help? That doesn't make any sense. Is scoring more important than finding a cure? You guys are hopeless.

He didn't even MENTION scores in that context! He was on about slow machines that slowed down the entire distributed network. There's no sense in work being done if it's slow, then they might as well get a Blue Gene.

[mdk777]

Look everyone,

We have explained the logic, benefit and considered reasoning behind the current point system.

Guru and FordGT90Concept do not want to understand.

Or they might claim to understand but disagree.

For once I also have to agree with 7im, there is really no point in continuing the discussion with people who do not have any intention of listening.

Turning the discussion into some class envy, reactionary, anti-technology, (the world is not really helped by advances in science, it just makes us all slaves to buying new equipment and planned obsolescence theory) just doesn't accomplish anything.

Remember that lovely story where the tortoise beat the hare?

Holy hand grenades Batman! Have you heard nothing about the project? The turtle gets Alzheimer's disease and dies a slow and painful death! You just really don't care what the real goals of the project are do you?

Good intentions don't count, results count.

Someone please close the thread so I don't have to read "Our suggestions are genius, you are all just fools for scorning our manifestly perfect plan....the world will regret that they ignored GURU...."
Really people like this start to scare me.

It does come down to your world view about "fairness"

Lenin would look at the situation and say 10,000 workers are producing as much as the other 260,000 combined. However, it is only because they have the Capital (GPU equipment) The other 260,000 workers deserve just as much credit (time recognition for their CPU TIME). Much time and energy would be wasted "redistributing" the wealth.


Henry Ford would say: I bet I can make a cheap machine so EVERYONE can be as productive as those 10,000 workers, and EVERYONE can share in the 1000X increase in production and wealth!!!

[Guru]

I appreciate the fact that new versions have been created to keep up with technology, but I'm not convinced that CPUs are being used to their maximum potential. The reason that I say this is the math involved. If floating point numbers are being used in calculations, then it hasn't been optimized for a CPU. CPUs will handle smaller calculations much faster and a large calculation can be simplified into a couple small calculations, which would result in a CPU being used more efficiently.

Think about it like this. A processor has 64 bits and lets say a quadcore runs at 3.2GHz. Now compare a 512 bit GPU that runs at 600 MHz. The CPU will receive 512 bits in the same amount of time that a GPU will process 512 bits. Now take into account that quadcore has 4 separate cores... Something is not right if the physical qualities of a processor indicate that it should be faster than a GPU, but is in fact slower. I am under the impression that this is caused by the type of math involved. This is a programming logic that can be changed, but it would require fundamental work on the structure of the entire program, which could be a lot of work. Nevertheless, it would then take advantage of the CPU efficiently.

Regardless of that, comparing all attributes of the device folding instead of just scores would at least make things more realistic without making any changes to the way the software works.

[Guru]

Lenin would look at the situation and say 10,000 workers are producing as much as the other 260,000 combined. However, it is only because they have the Capital (GPU equipment) The other 260,000 workers deserve just as much credit (time recognition for their CPU TIME). Much time and energy would be wasted "redistributing" the wealth.

Guru would say that they should receive the pay they originally contracted for and the new equipment will obviously earn a great deal more since it is accomplishing more. However, you shouldn't remove the labels that tell you whether or not you are looking at a worker's pay or a GPU's pay... Label each as what they are so that no one has to guess.

Albert Einstein would say, “Great spirits have always encountered violent opposition from mediocre minds.”

[Guru]

He didn't even MENTION scores in that context! He was on about slow machines that slowed down the entire distributed network. There's no sense in work being done if it's slow, then they might as well get a Blue Gene.

Wait a second. You're telling me that someone isn't smart enough to prevent scenarios where a single machine is slowing down the entire network without retiring the slow machine? Are you serious? To think that many of you are infatuated with their skill level and then to see that statement written down is just hilarious...

[FordGT90Concept]

For once I also have to agree with 7im, there is really no point in continuing the discussion with people who do not have any intention of listening.

I think you are not listening. The point system as it stands furthers your objectives and as such, you are biased to my point of view.

Holy hand grenades Batman! Have you heard nothing about the project? The turtle gets Alzheimer's disease and dies a slow and painful death! You just really don't care what the real goals of the project are do you?

Or a cure is made for Alzheimer's using corrupted information from someone who overclocked their computer and it ends up killing thousands before the computational flaw is discovered. I have seen what overclocking can do to computers. They may seem 100% stable but over a trillion clocks, it may not hold true. The score system drives those that contribute for the sake of points to push their systems to the breaking point. Ultimately, it may prove counter-productive for the project.

Someone please close the thread so I don't have to read "Our suggestions are genius, you are all just fools for scorning our manifestly perfect plan....the world will regret that they ignored GURU...."
Really people like this start to scare me.

If you can't take the heat, get away from the fire.

[mdk777]

I have listened.

You make what you think are debate points to show your "wit"?

But you make no real statements about how your improved time keeping system would work.
You make no references to any successful model or demonstration case of your concept.
You show no benefits.
You cannot name a single time in the history of the earth, where rewarding people for showing up, improved society.

All you do is make nonsense comments about how you could run the servers better and how it should be just a few lines of code to solve this "problem"

I have listened and you are correct:

I am very biased towards system that have worked, are working, and will continue to work.
I am very biased against social re-ordering, that for whatever original good intentions, have never worked in the past, are not working now anywhere they are being tried, and are doomed to fail in the future.

You debate for the sake of debate, and so infuriate those who really care.

I think there is a word for such a person.

TROLL perhaps?

[mdk777]

Wait a second. You're telling me that someone isn't smart enough to prevent scenarios where a single machine is slowing down the entire network without retiring the slow machine? Are you serious? To think that many of you are infatuated with their skill level and then to see that statement written down is just hilarious...

GURU, you give away your ignorance here.

It has happened. just recently the GPU severs were clogged by a half dozen folders cards EUE in constant loops.

Despite the considerable skill and attention of the entire PG staff, it is the nature of DC projects to face very difficult scheduling /work server load balancing problems.

In fact, this is the key concept in almost all computer science today, how can software, and hardware interconnects schedule communications, packets of information, to utilize massive parallel computing power. From making programs work with two, 4, 8 or 16 cores on up to supercomputers, how to best schedule that information flow is the single great problem of computer science today.

If you are that "someone smart enough" people will be quoting you in the future instead of Albert Einstein.

[Guru]

All you do is make nonsense comments about how you could run the servers better and how it should be just a few lines of code to solve this "problem"

Your entire post and many other posts you've created are nonsense comments... You did nothing to edify the discussion nor to solve the dilemma. Your only purpose has been to spout insults and to say that the system is good the way it is. There is a word for that as well...

FANBOY

[Guru]

Wait a second. You're telling me that someone isn't smart enough to prevent scenarios where a single machine is slowing down the entire network without retiring the slow machine? Are you serious? To think that many of you are infatuated with their skill level and then to see that statement written down is just hilarious...

GURU, you give away your ignorance here.

How so, by laughing at the fact that the system has a serious problem with tracking? If device types, speed, and WU completion rates were being tracked, (as I've been stating since I got here), THIS WOULD RESOLVE THAT PROBLEM AS WELL... Looks like I am the genius that I claim to be... I've came up with a single solution that fixes multiple problems with the software that none of the developers have put into the system since the start...................

If you are that "someone smart enough" people will be quoting you in the future instead of Albert Einstein.

If my idea is implemented into the system, then perhaps people will be quoting me. I wonder if I will get credit for my idea... lol Somehow I doubt it, but that's not my goal, nor do I really care. Seeing it implemented would be satisfactory. My goal is to help contribute in a way that makes the most people happy while improving the system and getting more results. My solution would accomplish that.

[FordGT90Concept]

But you make no real statements about how your improved time keeping system would work.
You make no references to any successful model or demonstration case of your concept.
You show no benefits.
You cannot name a single time in the history of the earth, where rewarding people for showing up, improved society.

Once you have thread time, you can derive a lot of values from it. I already covered much of this several pages back. There is two sides to this coin and you are failing to see one of them. Yes it takes some time to implement but it can also produce significant yeilds, especially in terms of accuracy of findings.

You debate for the sake of debate, and so infuriate those who really care.

If you truly believe this topic is "for the sake of debate" then you fail to see the point we (and others before us) are trying to make. If you don't like the prospect of improvements, then perhaps you don't belong in this thread.

[mdk777]

GURU,

You missed my point.

EVERYONE and I mean EVERYONE, that deals with Computer science, knows of this problem that you think you have "discovered".


"discovering" war or world hunger will not win you the NOBEL PRIZE.

Your solution: wish it didn't exist.

Sorry, not going to win any prizes either.

[alancabler]

Just found a few old PIIIs w'mobos in my dept. o' things and stuff- a 450, 500 and 550.
Think I'll fire 'em up just to hack off some poseurs in this thread.

[FordGT90Concept]

Don't bother. Pande only cares about GPU clients.