Folding Forum

new08 wrote:...
But I have seen strictures against OC at various times on F@H discussions- or PG set up tips..

Not lately you haven't. Quite the opposite, for example... http://folding.stanford.edu/English/Compete

Out of curiosity, I searched the forums for 2012 on the word "overclock." I couldn't find any posts by any members of Pande Group containing that term. I found some by Bruce and by 7im, but none of them seemed to be discouraging overclocking. Others can judge for themselves, of course, by reading those posts in context.

My finds:

viewtopic.php?f=55&t=20050&p=218255&hilit=overclock#p218255

viewtopic.php?f=19&t=21589&p=216074&hilit=overclock#p216074

viewtopic.php?f=19&t=21596&p=215923&hilit=overclock#p215923

viewtopic.php?f=38&t=21455&p=214972&hilit=overclock#p214972

viewtopic.php?f=74&t=21208&p=213456&hilit=overclock#p213456

viewtopic.php?f=67&t=21100&p=211345&hilit=overclock#p211345

viewtopic.php?f=19&t=21015&p=210171&hilit=overclock#p210171

viewtopic.php?f=59&t=20878&p=208800&hilit=overclock#p208800

viewtopic.php?f=67&t=20831&p=208098&hilit=overclock#p208098

viewtopic.php?f=38&t=20441&p=203618&hilit=overclock#p203618

I go along with 7im, then. I haven't been on here much this year..
Interesting that PG don't mention the word though..
Newer video cards seem to be getting more stable-and we've had more than enough driver updates!
Good old XP

DoctorsSon wrote:Overclocked is a matter of subjectivity.
What is the difference between a CPU For instance FX-4100 and a FX-4170 other than the factory supplied multiplier setting?
A 4100 can be changed to the speed of a 4170 real quick.
They still are the same chip, just a different factory setting for more $$.

It's called benching or binning. The manufacturer tests the chips under extremes of environmental and other conditions. The ones which pass at higher speed are sold for more $s as higher speed chips. The ones that don't pass are set to lower clock rates. Their testing is more thorough that anything you'll be doing. They have a right to charge more for better chips. You can also be sure that the 4100 failed one of the tests for a 4170.

codysluder wrote:You can also be sure that the 4100 failed one of the tests for a 4170.

This is true early in production of a new line of chips, but later on in production as yields improve that may not apply. Then as they "bin" the chips, once they meet their production goals for the higher speed products, the rest will be sold as whatever speed they need. So you may get lucky and get a 4100 that clocks up all the way to match a 4170, somewhere in between the 4100 and the 4170, or not be able to clock it up at all.

I have a FX-4100 running at 4.9 stable.
It is not a folding comp but it is stable and shows that they don't truly "Bin" the chips otherwise they would have sold it as a 4170.

They all do bin'ing. We simply don't know on what criteria...

You say it's stable, but do you live at the hottest place in the world? Has a nominal amount of dust accumulated in a relatively poorly attached HS (after years of service)? Have you tested it at TDP? The tests required to assure a CPU is stable under the WORST POSSIBLE expected conditions rarely occur on a well-maintained system in a home environment.

I personally say a system is stable if it passes prime95 and fah

iceman1992 wrote:I personally say a system is stable if it passes prime95 and fah

There are better tests than prime95 (OCCT, etc.) and risking WUs by testing with fah is a bad idea, but fundamentally I agree with you. My point is that a manufafacturer requirement to sell a product as stable is more stringent than mine is. They're not going to sell something that we call stable as a bin-ably stable to their requirements.

I just let my desktop (i7 3930k, Linux) fold 24/7...it works fine. I get about 40-50k PPD at stock settings, and in about 2 months I reached rank 15034 of 1645768. Power usage is just fine, too.

Seems there's a few angles to this throughput equation.
First- get people folding more easily. New V7 should help with this.
Secondly get donors to run their gear as hard as they can safely for F@H to get max o/p. [Could provide 20% incr.?]O/C seems quite specialised as an art form rather than a technical issue.
Thirdly, get donors to run longer hours up to 24/7 on efficient rigs, esp in the cold weather where waste heat is usable in the home.
Finally take an interest in what motivates people to fold by more contact with donors to give updates on research in an easy manner.
Because PG seem to have a laid back approach to throughput with software driving most new takeup, I tend to think they have broadly enough effort to maintain current work.
No doubt there are people who can expand on these issues here- that's what the OP is asking for. PR is not high on the Stanford list,maybe.

I'll definitely agree the V7 client is a big step in the right direction. Rather than people having to get a bunch of different clients, it's all nice and centralized. It also means less things to keep track of, in a way.

OCs are fussy things; you can't guarantee how good a part will run. My CPU, for example (a Core i7 2600K) will boost up to 4.3 GHz with no problem and no need to even touch the voltage, but there's some people who can get it to go much higher but are more technically savvy. Enticing people into overclocking is bad, since if you do it wrong, the magic smoke escapes, and somebody has to replace a processor or a motherboard.

The research updates could be nice, but science takes time. Even with us folding the data, they still have to analyze and apply it, so it might be a few months or even years before results go out.

As for the heat... oh yes, I'll be abusing that this winter. Last winter was actually surprisingly warm (it rarely even snowed, and our city gets nine feet of snow a winter on average), so I know this year we're going to be absolutely nailed by it. Always happens.

I very much agree with the sentiment that encouraging folks to OC for better output is not a good idea unless it's being done through a community dedicated to the black arts of OCing PC hardware. That way there is at least a bit of support and (more importantly I think) a recognition of the benefits / limitations of pushing hardware beyond stock settings.

I personally don't use FAH to test my OC's (I typically use a combination of short term (2 hrs) stress tests, GFX stress tests, and finish off with a minimum of 24 hrs prime), I use FAH as proof that my OC's are completely stable once I've got them dialed in. Even with that, because I try to ride on the very edge of fully stable, there are the occasional new WUs which stress something differently and thus start pushing out EUEs or something similar. It's important that folks who are pushing their hardware recognise these OC related issues quickly and take action in order to minimize any damaging effects on FAH's data.

well said sswilson!