...
But if the calcualtion of the payout at 30:1 contains random errors that spoils the stats. If they randomly payout at 33:1 too often the house goes broke. That is because the game demands that the calculaton of the payout is deterministic even tho the decision to make a payout is not.

I work in a casino and either of those ratios on the roulette wheel would get us in trouble, too much profit not enough chance for the player. You mean 35:1 and 38:1 or 39:1 most likely. 38:1 is a dead even american (double zero) wheel, 35:1 is the standard payout.

Fair comment, you can tell I don't play roulette! - but whatever the odds, a 10% boost in the payout odds would be bad news when that reverses the house's margin.

Edit: A biased wheel is a bigger deal than you would think as well. The advantage can quickly drop from 5.26% to zero or worse if a number is more favorable and a player notices.

Yes, but in general random errors in code that simulates a roulette wheel will not bias the wheel at all. This is the valid point in what Bill says.

Suppose for example that you went 3 slots too far round the wheel each time - you'd have the same randomness apart from the exact sequence of the numbers. Not a problem for the house.

R~~

The only project that I can recollect, that actually came out with a complaint to overclockers was LHC@Home.

Predictor actually segregate WU into platforms - ie one WU gets all its results sent to the same platform, the next WU gets all its results sent to another one.

They do this precisely to avoid differences between platforms.

I'd have said that if a project goes as far as that to try to get homogeneity between processors, you'd be unfair to introduce any unnecessary differences by overclocking.

Back to the topic. I am biased against overclocking in general but not adamently against it.

Ah! John, my sentiments exactly ... though I could never seem to find the words to say it ...

Well, at least I cleaned off one pending task, that has been pending over a year ...

---

So, I guess the answer here is ..... maybe.

dondee

So, I guess the answer here is ..... maybe.

Yes, it's definitely a maybe. :-/

---

---

I'm of the opinion that overclocking does not necessarily cause any accuracy problems.

I overclock some systems because they will take it and still pass diagnostics.

Would the counter argument be to "underclock" all computers to make them more accurate?

Unless the processor is a "test-escape" (i.e. one that has something broken that wasn't detected during testing), underclocking will achieve nothing to improve accurracy [assuming of course that the CPU is run under "correct conditions", such as temperature and voltage]. All processors go through a test to ensure that they calculate correctly at the speed they are meant to run. The only difference with running the processor slower would be the amount of power the processor uses.

What will and won't go wrong in a processor when it's overclocked depends on where the "slow-path" is. The problem with overclocking comes when the data that is SUPPOSED to be ready at the next clock-cycle isn't quite ready yet. Then the processor will fetch the wrong result from the previous stage of the calculatin. This may be one single bit or multiple bits - even the "previous result" in some cases [a latch that didn't get the next clock yet will hold the OLD data]. It's by no means sure that the processor starts going wrong on floating point calculations first either - it could just as well fail on integer operations, including address calculations which means that entirely the wrong data is being used for some calculation. Caches and TLBs (Translation Lookaside Buffer - what the memory management uses to find the physical address from a virtual address) may also fail at higher than recommended speed - which again means that incorrect data is being used in some way or another, and that again can be completely random which particular bits or parts of data is being "hit" by this.

I'm a bit surprised that LHC could say that overclockers had problems in the 6/7th digit of accuracy - maybe they can sort out other miscalcuations more clearly and just error them - I know nothing about LHC, but I'd be surprised that overclocked machines would miss only in the lower part of the result - in my experience, machines that are overclocked may just as well give an error in the most significant bit as the least significant bit or in the middle of the number, and with some more complex math, the resulting erorr could mean completely different results...

--
Mats

---

Dont people do proper prime95 or orthos testing after they overclock? If it can handle those overnight, then i doubt there will be any problems due to overlcocking.

---

Dont people do proper prime95 or orthos testing after they overclock? If it can handle those overnight, then i doubt there will be any problems due to overlcocking.

Don't be so sure about that - it depends on which paths through the processor are hit by Prime95 and which ones are hit by for example Rosetta. Obviously, if it DOESN'T work with Prime95, it's probably not going to work with other apps either.

But each app excercises the processor in a slightly different way, so you may find that one app works fine whilst another one fails, simply because one of them has a slightly different set of instructions, data values, etc.

Obviously, if you want to use your machine with Rosetta, you shouldn't just crank the speed up and hope - you should do your best to ensure that all sorts of verifiable tests are run before you give it a Rosetta work-unit.

--
Mats

---

you need to test the memory too - well you need to test everything to be sure!

the MS Memtest is pretty good at finding errors apparently.

---

I have never had a machine that has passed Prime and memtest for 24hrs crash out on a WU because of the overclock. This includes Seti Classic, Predictor, and Rosetta.

Bad batches of WU's, and application problems, Yes. Overclocking No.

And do we think that all the processors sold are just made for that speed? Supply and demand dictates the manufacturers to program higher clock speed chips to lower speeds to meet the demands for lower speed chips. Yes some are binned because they fail at higher clock speeds. But for the most part every CPU will run at least one clock speed faster at stock voltage to make sure they will run at the lower end of the +/- 5% ATX specification for voltage regulation.

The biggest problem with overclocking is heat, and voltage regulation. If you can keep the core temps down under 50c, and the voltages within the +/- 5% specs (I like to see less than +/- 3%)then you can push as much voltage as you like thru the CPU, within reason.

This starts with the PSU. Skimp here and you are not going anywhere, and take the chance of taking other components out when the PSU fails. Second is cooling, this means better than stock cooling. I like water, but a high end air cooling solution works also if you can put up with the noise, although the new heat pipe based air cooling solutions are very good, and less noisy that the older models. They cannot compare to a quality water option, and I like quiet.

I have a buddy who runs only stock boxes, and some of them are work machines, and he has errors some of the time. He cannot believe that my machines are not crashing WU's and his are. But he is always checking my stats, and is now a believer.

If you use quality parts that are matched to work together, I believe that a properly assembled, and tested overclocked box is more reliable than a off the shelf bargain box.

So should we also ban boxes that cost less than $XXX.XX? Or if you have a box that errors out xx times you are also banned? This would be the next logical step.

I would even venture to say that if Rosetta implements a no-overclocking policy that the teraflops will drop but the errors will continue, at the same or even an increased percentage. Why? Because the owners / builders of overclocked machines are more knowledgeable and monitor them more closely that the owners of off the shelf models. And are usually members of a team, where they can go to get help with their problems.

I can't wait for the fallout of this, first the credit system. Now the banning of overclocking, and all machines that error out xx amount of times. Because this is the best and only way to make sure the project is pure.

We can even take this further and say that if you don't have xyx components in your system you can't play in our Rosetta game. So who is going to decide on the rules for our game? Seems to be more than a few people in here (not just this thread) that would be more than happy to decide whats best for our little game. So why don't you guys start on your x-mass wish list for the project, maybe Santa can bring you what you wish?

Starting to sound like a bad Rudoff nightmare, you cannot play in our Reindeer games.

All overclockers and error givers get out, we don't want you to play in our game anymore. We don't care that you are providing us valuable computer time, you are not pure, and therefore not worthy, so get out! And if you have a couple of untimely power outages, and did not have a qualified UPS, too bad thats an error, get out. And who decides just where the line is drawn? Oh yea the vocal minority. Silly me...

You don't have to remind me that the vocal minority always gets it's way here. Thats already a proven fact. I do believe in this project, but I will not be bullied again. The next set of unreasonable conditions forced upon me will be the last. I will not turn my boxes down to stock speeds, I did not spend $225 a box on watercooling to run at stock speeds.

And since the last time I posted my thoughts here the vocal minority jumped down my throat, and called me a liar, I will not be responding to anyone's responses to my thoughts. They are mine, and I don,t care what a bunch of techno know-it-alls think.

This would almost be funny if it was not so sad.

---

BETA = Bahhh

Way too many errors, killing both the credit & RAC.

And I still think the (New and Improved) credit system is not ready for prime time...

I've had multiple bad experiences with long term overclocking - one using a overclocking board that was recommended by a forum I haven't associated with in years. Ram tested out fine.. temps were cooler than the overclockers.. but my ibm/hitachi HDs didn't like the 5% overclock on the via chipset being used. It managed to corrupt windows so it needed a reinstall to boot up. :) Which was followed by a similar experience on a via chipset Athlon 64 motherboard.

The 5% improvement for 2 weeks to a month wasn't worth it in those cases - after which both systems ran years with no problems at stock speeds. (The joy of proving an incompatibility.. :) And while I know lots of people who have had good experiences OCing.. (I'm probably the only member of TRBCF that runs stock speeds..) - I also know some that would brag about their video redraw rates - and disappear after 20 minutes in Desert Combat matches due to video corruption. And then be gone for 30 minutes getting everything cooled down and set back to stock speeds.

So there are a number of folks that are just not cut out to run above stock speeds. More power to those who are able to OC error free, though. :)

---

To those who are really interested in knowing if overclocking afects the science just look at results of CASP 7, those machines that produced the best fit. The largest proportion of them belonged to people that overclock.

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

To those who are really interested in knowing if overclocking afects the science just look at results of CASP 7, those machines that produced the best fit. The largest proportion of them belonged to people that overclock.

Only 5 on the first page, that i know of personally. I don't have to look any farther to satisfy myself.
Machines predictor Opteron 170 @ 2800Mhz. 24/7 cruncher.It now crunches WCG with 100% valid results, what ever that means.
meatgrinder2 12/08/2006 21:32:04 0:166:07:43:06 518,857 1,035
166 days 7 hrs. 43min. 06 sec.= runtime WCG
opps almost forgot 518,857 points that i don't care about generated :)
heater

---

Ah, but but you know how to overclock properly and safely (well unless blowing out watercooling pipes on their new computer Jose ;)

I see it this way,

If you overclock too far your results will fail, crash at some points and so you do not effect the results. Just your own experience and throughput, but hten you'll also find other parts of your computer are unstable.

Don't know what the guy that went off on one is on about, who mentioned banning overclockers?
Most homebuilt computers have software that comes with it to do the overclocking automatically now. Press go and it tries it out for you.

For a start Gil necromanced this thread which is almost a year old!

---

Team mauisun.org

Don't know what the guy that went off on one is on about, who mentioned banning overclockers?

My advice is, out of fairness to the project, don't overclock on a project that does not apply redundancy. At present this implies don't overclock on Rosetta.

Just my two pence worth, we will have to wait till NewYear for the official project policy on this.

River~~

---

BETA = Bahhh

Way too many errors, killing both the credit & RAC.

And I still think the (New and Improved) credit system is not ready for prime time...

Stevea, River was stating his personal opinion. He does not speak for the project. And he made that post back in 2005, so his opinion may have changed a few times since then.

Ah, but but you know how to overclock properly and safely (well unless blowing out watercooling pipes on their new computer Jose ;)

I see it this way,

If you overclock too far your results will fail, crash at some points and so you do not effect the results. Just your own experience and throughput, but hten you'll also find other parts of your computer are unstable.

Don't know what the guy that went off on one is on about, who mentioned banning overclockers?
Most homebuilt computers have software that comes with it to do the overclocking automatically now. Press go and it tries it out for you.

For a start Gil necromanced this thread which is almost a year old!

Overckocking requires a commitment to equipment.(Softeware @#$#$###$$$*&^) To learning and asking questions before doing something that will turn your new and most powerfull CPU into a nie key chain. Overcloclking is nurtured in a community of overclockers....

So, you got those who use opti clients out of here ; Now I see people are going atfter overclockers?

What next?!!!!!

Are they going to prohibit machines whose owners have sex ....

---

This and no other is the root from which a Tyrant springs; when he first appears he is a protector.”
Plato

@stevea,

as AMD_is_logical said it was his opinion BUT it was also a YEAR AGO, did you not read the last part of my post ;-)
When it was posted was also not long after the project had been going.


@Jose,
Exactly, nTune that Nvidia provide for its motherboard chipsets, ASUS, ABIT all do similar wrap arounds for their motherboards, even Intel are getting in on the act.

Press a button, it goes through tests altering FSB/HTT/AGP/PCI, RAM timings, graphics cards active overclocking etc.. all at the push of a button or a tick of a box.
the problem comes when people 'trust; this and then go blaming Windows for being crappy for crashing all the time. This is still overclocking, but not in the way enthusiast overclock, They at least understand what the problems are going to be.

Myself, I overclock whatever I can, my Laptops PentiumM is overclocked from 1.7GHz to 1.9GHz. The only computer I have that is not overclocked is my older PentiumIII-m computer since I just cannot find out how to do it. I did use to overclock it's M6 graphcis card though :-D

I had peltiers in my Pentium 75 and K6's ...

---

Team mauisun.org

As Bill said back in '05, there's overclocking and OVERclocking. If your hardware can handle it, then there's no harm in it (it's beneficial to the project). The problem is it might not be obvious that your hardware can't handle it...

---