[QUOTE=LaurV;412758]The project is better if any user clears how many exponents he can, how fast he can, with the hardware he has. Period.
If your hardware can clear more exponents per day (week, month) doing TF, then go for TF.
[/QUOTE]As James already replied this would mean all GPUs would do only TF on low levels : no need to stop at 1 G exponents. This would maximise the number of cleared exponents indeed.[QUOTE=VBCurtis;412734]You appear to claim that solving for the most-efficient path should take into account the typical hardware used for each task, rather than judging by what a single piece of hardware can do for each task. How would you go about accurately doing that? How do you value 1 hr of GPU time vs 1 hr of CPU-core time (or 1 hr of a full CPU)?
I don't see how making those judgments is an improvement over pretending one's own GPU is the only item that is going to work on a particular number, and determining the most (expected) time-efficient way to go about testing that candidate. The solution is likely different on CPU vs GPU (for instance, it may be optimal to P-1 higher bounds on a GPU than a CPU because the code is relatively faster than TF code on a GPU, or vice versa), but it seems folly to declare one solution that holds for all of GIMPS for any machine.[/QUOTE]The reasoning you apply (and encoded in the Prime95 program) that a signle machine will do all work on an exponent, this has been almost true in the beginning of GIMPS and is still visible in the very poorly P-1'd exponents. But now different machines do the TF to different levels, other machines do P-1 and still others do the first LL and the double check(s) on a single exponent.
It will indeed be difficult to have an accurate typical hardware (one of the reasons being that the hardware mix continuously changes). But there are trends and one can calculate approximate TF levels, P-1 bounds based on that. Especially now that people specialise on particular types of work.

Jacob

[QUOTE=S485122;412776]But there are trends and one can calculate approximate TF levels, P-1 bounds based on that. Especially now that people specialise on particular types of work.[/QUOTE]

Completely agree. Economics has been called a "Bastard Science" (even by The Economist). Rightly so.

I particularly enjoyed the lessons learnt by the mistakes made by Reinhart and Rogoff of Harvard University in their spreadsheet (which, because they were PhD's, was assumed to be correct).

Interestingly, fiscal policy was informed by this (mis-) information in many nations (to their detriment).

[QUOTE=chalsall;412778]Completely agree. Economics has been called a "Bastard Science" (even by The Economist). Rightly so.[/QUOTE]

Economics is really a humanity. Economics is fundamentally about human behaviour.

[QUOTE=Mark Rose;412799]Economics is really a humanity. Economics is fundamentally about human behaviour.[/QUOTE]

So, then, a "social science". No where near as accurate as Asimov predicted in "The Foundation".

Read: not worth its weight in salt.

[QUOTE=chalsall;412802]So, then, a "social science". No where near as accurate as Asimov predicted in "The Foundation".

Read: not worth its weight in salt.[/QUOTE]
Isn't the Foundation serie based on some sort of psycho-history? That is at least what I can remember (I read the first 3 books of the serie in Dutch some 10 years ago).

[QUOTE=VictordeHolland;412818]Isn't the Foundation serie based on some sort of psycho-history? That is at least what I can remember (I read the first 3 books of the serie in Dutch some 10 years ago).[/QUOTE]

Yes.

"Psycho-history" is remarkably close to what we now call economics. Equally non predictive.

[QUOTE=S485122;412776]As James already replied this would mean all GPUs would do only TF on low levels : no need to stop at 1 G exponents. [/QUOTE]
You (and James) miss the fact that [U]yes[/U], the best way to clear exponents in 1G range is - [B][U]still[/U][/B] - trial factoring. Why don't you do LL in 1G if you think otherwise? :razz: The goal of the project is to find primes, and you only find primes by doing LL. This LL is more "reasonable" to be done for lower expos (this was davieddy's argument, wasn't it? haha), for whatever reasons, therefore lower expos is where you have to clear more exponents, by any meanings, with [U]your[/U] hardware. Of course I am not against cooperation, see my activity in the forum and in the project, for years. But if it takes me 3 hours to do a P-1 in 70M and I found a factor every 25 trials (say 4% chance to find a factor), than it makes no sense to TF at a bitlevel higher than 75, if that would take me more than 1 hour. Because in P-1 case I can find 1 factor in 75 hours, and in the TF case I would find less than a factor in 75 hours. With the same hardware and power consumption. Of course, if that hardware is a newest Tesla and I can do one 70M LL test in 37 hours (just an example!), then neither the TF nor P-1 would make sense, because I could clear that expo in 74 hours by doing one LL and one DC to it, and THAT is a sure thing (not probabilistic, like finding factors) and additionally, it may bring me a prime.

What I want you to understand is that I am not arguing against you. What you say is ok, and again, I am not against cooperation.

But I want that [U]every[/U] user understand the goal, and make this type of calculus for him/herself, for the hardware [U]he owns/use[/U] and see for himself how he can help better. And when I say help, I say "help the project" [U]and[/U] "help yourself". Help better, feel better, do whatever you like. But think first. And that should be the attitude.

Don't forget the main goal is to find primes. Not to clear exponents. But somehow, they go together - lord works in mysterious ways :razz:

edit: Disclaimer: the numbers used are only examples for calculus

[QUOTE=LaurV;412830]... that [U]every[/U] user understand [color=red][b]the[/b][/color] goal ...[/QUOTE]I think this is where you are perhaps creating a problem. There is no one single goal. People have different goals. Some just want to test their hardware for errors. Some want to find lots of factors. Some want to find primes. Some want to be the top of the producers report. Some just want to be part of the community. etc. People will optimise for [color=red][b]their[/b][/color] goal, whatever that goal may be.

That is exactly what I was saying. You piked on my English :yucky: I was just arguing for "the group is doing well when the individual does what's better for the group [U]and himself[/U]" before, and you and others jumped on me about the "greedy" part. It was my reply to it. This is not a complaint, I like this discussion. We are all in a "violent agreement" here. I will follow whatever is decided, in my own way (i.e. doing work at which I "feel" I am more efficient).

[QUOTE=LaurV;412834]You piked on my English[/QUOTE]Okay, I didn't realise it was just a translation thing. But for others reading the sentence it is perhaps good that the confusion is cleared.

[QUOTE=retina;412835]I didn't realise it was just a translation thing[/QUOTE]
[COLOR=White]Honestly it wasn't. And I didn't really think that you picked on it. But every time when we are cornered, we blame the fact that we are not native speaker... Where is that picture from Mike? (pity we can't make that picture to be white color only, too)

:kitten:[/COLOR]