CPU Time versus Credit

Jord
Joined: 26 Jan 05
Posts: 2,952
Credit: 5,723,975
RAC: 308

I have to ask Al, do you have

Message 97431 in response to message 97427

I have to ask Al, do you have any spare socket 478 (chipset i848/i865) motherboards with FSB800 and hyper threading lying around? My Pentium 4 3.0GHz is at the moment not running BOINC, since memory slot 2 has a problem. When I put memory in there, I see BSODs ranging from 0x0A, past 0x19 and 0x7F to 0x8E. Seti will constantly plonk with 0xC5 errors.

So I now run with only 1 GB RAM and to make sure my system is stable enough to do something on it, not run BOINC. See if I care that any work on it goes badly past deadline. Redundancy, anyone? ;-)

Will go hunt for a new mobo today. Before I get complaints why don't I buy something new, let me tell you that if I could I would, OK? But since my budget allows for a maximum of 60 euros, I can't even go for an i945 chipset motherboard as that would mean I still need new DDR2 RAM, a new SATA2 (SATA/300) drive and most probably a new videocard.

Thor
Thor
Joined: 21 Oct 06
Posts: 2
Credit: 686,855
RAC: 0

RE: I think it's

Quote:
I think it's fascinating as to why people contribute such a valuable resource for nought exchange in material terms.

Spoken for myself:
1. It doesn't really cost me much. (It's just the difference in power consumption between the "idle" and "load" status)
2. I like it when my comp is working. She's running all day anyways (when I'm at home) since I don't have a stereo or a TV she's basically the center of "entertainment" ;)
3. Boinc is a funny software. I can study, watch videos, hear music, even play most games without being really bothered by it. Only when I need really good FPS i stop it temporarily.
4. Its a nice software to let the system work for hours / days and test overclocking stability & temperatures. Yes I know there are specifically designed stress tests, but with Boinc the "stress test" is even producing results another guy who's way smarter than me can use. Two flies with one grenade ;)
5. Silly me I forgot the most important reason of all: I like to feel helpful. I like the fact that I can help scientists with my humble possibilities.

Quote:
With E@H I'm pretty sure it's largely the cachet of the Einstein name per se [...]

Guilty as charged. However, I find the concept of gravitational waves quite interesting. But come on.. I had to chose something.... ._.

But Oh my that's severely off topic. Sorry

Greetings
Thor

P.S.: Yes, I have a little brain damage due to an overdose of Sci Fi, as to why I call my comp a "she" and pretend she's an AI :X

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,334
Credit: 1,857,978,901
RAC: 1,507,737

RE: I think it's

Message 97433 in response to message 97428

Quote:
I think it's fascinating as to why people contribute such a valuable resource for nought exchange in material terms. With E@H I'm pretty sure it's largely the cachet of the Einstein name per se, however that is backed up by the very real continuation of his work born over a century ago.

My math skills aren't good enough to do this for real.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,149
Credit: 129,254,071
RAC: 13,336

RE: My math skills aren't

Message 97434 in response to message 97433

Quote:
My math skills aren't good enough to do this for real.


Ah, but actually in a way you do, by ( probably inadvertently ) continuing another tradition even older than general relativity, and therein lies a tale .... ;-)

Originally 'computer' meant a person who did do mathematical calculations, as a paid job. So back in Brahe/Kepler's day they had crews of people working under their supervision, doing all the math to finally lead to conclusions re. planetary orbits. On or about that time logarithms appeared and cleared an enormous logjam. So instead of having to do utterly brutal long multiplications/divisions to however many digits, you'd just convert the operands to their logarithmic equivalents, add/subtract them in a few easy steps and apply exponential ( or anti-logarithm ) to get a result. Note that as trigonometric functions especially can be represented as power series then the true value of logs really kicks in ...

So here we are now owning and operating gadgets that do what was once a profession performed with ink, paper, candles and exceptionally rigorous patience ( hence 'burning midnight oil' ). We volunteer/delegate their services for E@H - with all the methodology and data supplied. All we need to do is maintain them with sufficiently energetic electrons, a nice home to live, a conducting conduit to communicate with, blow out the dust bunnies occasionally .... :-)

So we caretake machines to do what was once a wetware function. Think of E@H as Einstein's computing crew as per the olden day definition.

Cheers, Mike.

( edit ) Another point about logs, demonstrating that a quantitative change can have a qualitative jump in the type of problem that can be tackled. Logs brought many calculations to within the span of a human lifetime, and so it is with E@H : the reason LIGO needs help is the size of the data set, and you have to go through it all to find the answer.

I have made this letter longer than usual because I lack the time to make it shorter. Blaise Pascal

archae86
archae86
Joined: 6 Dec 05
Posts: 2,938
Credit: 3,758,655,160
RAC: 5,348,801

RE: Originally 'computer'

Message 97435 in response to message 97434

Quote:
Originally 'computer' meant a person who did do mathematical calculations, as a paid job. So back in Brahe/Kepler's day they had crews of people working under their supervision, ...

And still into the 1940's, save only that by then they generally had electromechanical desk calculators to multiply their own pace somewhat. Also, by the 1940s, they were quite generally women, both in the USA and the Soviet Union.

Mike Hewson
Mike Hewson
Moderator
Joined: 1 Dec 05
Posts: 6,149
Credit: 129,254,071
RAC: 13,336

RE: RE: Originally

Message 97436 in response to message 97435

Quote:
Quote:
Originally 'computer' meant a person who did do mathematical calculations, as a paid job. So back in Brahe/Kepler's day they had crews of people working under their supervision, ...
And still into the 1940's, save only that by then they generally had electromechanical desk calculators to multiply their own pace somewhat. Also, by the 1940s, they were quite generally women, both in the USA and the Soviet Union.


My mom did that in her youth - for an accounting department - with the title of 'comptometrist'.

Cheers, Mike.

I have made this letter longer than usual because I lack the time to make it shorter. Blaise Pascal

Whiskymania
Whiskymania
Joined: 2 Sep 05
Posts: 8
Credit: 10,030,266
RAC: 0

RE: RE: Cross-project

Message 97437 in response to message 97427

Quote:
Quote:

Cross-project credit allocation is insoluble - not technically, but sociologically. You just won't get agreement on the scientific or other relative merit of the utility of the investigations that different projects perform.

It's paradoxical ( good old human nature .... ) but you'll find no shortage of people who will demonise a virtue. So what some will view as a terrific range of choice of enterprises to enjoin - currently quite a smorgasbord really - others will bemoan the lack of parity between them.

Quote:
Aside : My pet peeve, as a moderator, is the 'hostage taker' paradigm - "I'll take my valuable farm/services/battlecruisers elsewhere, and your project will suffer, if the following demands aren't met .....". Sharp punters may note the number of times, in reply, I've had to obfuscate the phrase "go on then, p*** off". :-) :-)

Cheers, Mike.

LOL...

Regarding you're Quip-of-the-Day, yes, I noted that! ;-)

As far as the other observation goes, that's the fallacy that indicates a lot of people can't be objective about the credit subject, and that includes technical types.

The question has nothing to do about the relative merits or value of the science, either technically, socially, or morally. A fundamental principle of BOINC is that all science is created equal in terms of credit.

It's simply a matter of how much work did the computer do. Unfortunately, there is no way to directly read out the amount of computational work a computer does (an OPometer). The best we can do is to infer the amount of work from its derived quantity, power. That's when all the fun starts. :-D

Of course, we had a very nice discussion about virtually every aspect of the subject back during the summer of '08 here in NC. Not much has changed since then. ;-)

Alinator

Hi again

I didn't wanted to offend anybody. I just wanted to point out, that the correction factor should be more higher, arround ≈2 could do it right I guess.
Thats, for any reason I don't know, cause my iMac with core i7 CPU need at least 2 times as long for am S5GCE WU than for an ABP WU. Compared that to the credit given, which was starting with less credit than an ABP WU and after a while has been corrected by the factor 1.6, I would like to earn just what time consumption needs. If it needs two times longer, twice as much credit should be given. If it need ten times longer, just ten times more credit should be given. Is that really not possible?

Besides that: My name is Markus and I'm from a small country in central Europe called Switzerland. You might know my country since UBS has made world-wide news. I'm 42 and married with my lovely wife. So my opinion is allways a "his" opinion ;-)

Have a nice day
Markus/CH

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,334
Credit: 1,857,978,901
RAC: 1,507,737

The problem is that there's a

The problem is that there's a very large variation in relative performance between different architectures. IIRC (I was looking at S5GCE, ABP2, as well as the now defunct S5R6 apps) My Intel systems earned more credit on one of the current apps, my AMD systems on the other. Unless and until the disparity in relative performances between CPU architectures in the current apps is resolved significant variation in which app earns the most credit for any computer will be impossible to remove.

DanNeely
DanNeely
Joined: 4 Sep 05
Posts: 1,334
Credit: 1,857,978,901
RAC: 1,507,737

actually my cross platform

actually my cross platform analysis was farther up thread.

The AMD systems were relatively close in relative performance per app with at most a slight advantage to S5GCE. All the intel systems are earning significantly more credit with the newer apps than with the old S5R6 and I strongly suspect the s5r6 credit rate is the correct one. A core1 CPU is only about as fast per clock cycle as an a64 on most apps; but in the new apps my 1.7ghz core1 duo is beating my 2.4ghz A64x2.

When credit rates are adjusted again I expect to see a significant drop in the rates earned by my intel systems. Hopefully this will be done concurrent with new apps that reduce the AMD penalty they currently have. I suspect the problem is that the new apps are written entirely in C++ with the result the GCC's results are rather variable. S5R6 (and the few apps prior) all had the key portions done hand optimized assembly for maximum performance on all systems. The E@H team doesn't want to do this until the apps are mostly final because writing assembly is much harder to do.

The fact that the AMD systems have maintained something much closer to 1:1 than the intel ones is probably due to the fact that the projects clusters are all built around AMD so that's what they're using for the benchmarks they calibrate the rates with.

Stranger7777
Stranger7777
Joined: 17 Mar 05
Posts: 436
Credit: 375,770,082
RAC: 147,619

Looking throughout this

Looking throughout this thread, I suddenly understood that newer WU's give us almost equal credits but take each time more time to complete a WU for the current search than for the previous seaches. This means that new search WUs contain less data (in cobblestones) or (and I slowly come to believe in that) newer clients are less and less optimized than previous ones, thus making less useful work with each new version or run.

For example, let's assume that modern Core Quad ten times more powerful than first generation P4. Given that, we should receive ten times more credits using the same app. And that is true, because P4 takes 10 times longer to proccess the same WU. It is ok, but....
If we compare S4 WUs with new S5GCE WUs, we will found that the same processor gives us more credits per hour with older app than with the new one.
What does it mean? And why does this happen? May be 2 years ago we used more scientifically-powerful applications? Or may be new programmers can't reach the same level of optimization in new apps? It will be interesting to listen for any opinion and any interesting explanation.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.