How about your provide numbers and tests for your own 'beliefs'?

 

Yeah +1, no disrespect to Martin and Noel, but they always seem to try to disprove every theory they see. Just because it isn't logical doesn't mean it's not possible.

And looking at Bens figure, those results are fairly linear with a few spikes.

Yeah +1, no disrespect to Martin and Noel, but they always seem to try to disprove every theory they see. Just because it isn't logical doesn't mean it's not possible.

And looking at Bens figure, those results are fairly linear with a few spikes.

 

Arjen...think about it, what are the possible explanations for what Martin's and Noel's testing demonstrated, which was exact 1:1 scaling w/ overclocking, albeit a narrow range in my case, 3000 to 3710Mhz. I'm just puzzled about why your data set was so very different fundamentally. What's your best guess as to the source of the difference? You heard Martin's findings, exact 1:1. And 1:1 I think would be the default guess in how this would function, especially for an application that is allegedly 'cpu bound'. That is the first guess I would have without having any empirical evidence to go on. I'm only interested on one thing: what is the truth. I don't care if I win or lose an argument that is not the point. I just care about getting to the bottom of it, and even more importantly, why is it as it is? Answering the why question will help us generalize to other systems and so find these reports useful in our own practices.

 

So again, why is is that you're seeing completely, seemingly random scaling in performance per clock speed, whereas we're seeing exactly 1:1 scaling regardless of clock speed? What are the potential explanations? Is it SB versus IB or C3Q? Is it something in the video driver perhaps? These are questions that when answered give us important hints as to what's important and what really isn't, for example in this very question: does it really yield much greater than 1:1 scaling to go from say, 4.7 to 4.8Ghz? Or is it really just the 2% change 1:1 would involve. That a big question, that deserves exploration. Is it worth the extra .3 volts? People will consider whether or not the risk reducing the life expectancy of their CPU based on the truth of this issue. So Arjen, don't take it personally. See if you can come up w/ some explanations, and if not that, some ways to test to try to determine the truth, and what really matters. You were the dude who posed the question originally!

 

Don't forget to enjoy the awesome performance you already have. I hope I get there some day!

 

Peace Out!

How about your provide numbers and tests for your own 'beliefs'?

 

I did, in the thread I mentioned. Not super scientific of course, but then neither are your figures

 

Yeah +1, no disrespect to Martin and Noel, but they always seem to try to disprove every theory they see. Just because it isn't logical doesn't mean it's not possible.

And looking at Bens figure, those results are fairly linear with a few spikes.

 

Actually, if it's not logical, then it generally isn't possible. The universe we reside in IS logical. the universe functions according to the immutable laws of physics. :smile:

 

Neither Noel, nor I, disagree just for the hell of it, just to wind you up. No, on the contrary, we do so because we, as individuals, have found the exact opposite to your hypothesis. Our experiences tell us different Simple as that.

 

Scepticism is a healthy thing. It's VERY important to attempt to disprove a theory. If it passes the test, and can't be disproven, we have all learnt something new. NEVER accept anything of import without definitive replicatable evidence.

 

If a reasonable number of individuals, all provide definitive evidence that you are right, both Noel and I will reverse our opinions. And thus... we will have learnt something new.

<br />So again, why is is that you're seeing completely, seemingly random scaling in performance per clock speed, whereas we're seeing exactly 1:1 scaling regardless of clock speed? What are the potential explanations? Is it SB versus IB or C3Q? Is it something in the video driver perhaps? These are questions that when answered give us important hints as to what's important and what really isn't, for example in this very question: does it really yield much greater than 1:1 scaling to go from say, 4.7 to 4.8Ghz? Or is it really just the 2% change 1:1 would involve.<br />

 

Hey Noel, I just found a statement made by you.

 

Not very well I'm afraid. 4.7Ghz is really not enough to push FSX where it needs to go. You only will have reached the Promised Land when you get to at least 4.8Ghz as this is the spot at which you overcome in the inherent timing issues that manifest themselves w/ clockspeeds under 4.8Ghz. Let's say you are in the NGX and sitting on the taxiway at KJFK in heavy weather. You're going to sitting there feeling terrible when you see the frame counter is at 23FPS. Now, jack that thing up to 4.8Ghz, and all of a sudden frames jump to 40FPS! This is where true simulation Nirvana begins! Don't rest until you see some instability, brownouts in your neighborhood, and no longer need to put the heat on in your house in mid-winter!

 

 

1:1 scaling you say? Not according to what you said here, please explain. :smile:

No wonder we are getting Global Warming

 

<_<

1:1 scaling you say? Not according to what you said here, please explain. :smile:

 

Arjen... Noel changed his mind, based on further evidnce.

 

He's honest you see, if he was wrong and now thinks diferantly... he tells you. :wink:

1:1 scaling you say? Not according to what you said here, please explain.

 

Arjen,

 

Either you're pulling my leg now & paying me back, which I don't blame you for, or I need to be more obvious & come clean! Here's another one of my comments from the same thread, and I already apologized for this one. It was all tongue-in-cheek, i.e., facetious, a joke, poking fun at what appears to be, as I say, magical thinking...

 

Ahh, well if it's unstable at 4.8 then the other speed/timing Nirvana comes when you overclock just to 4.6Ghz. You'll most likely get better performance from 4.6 than 4.7Ghz, in fact sometimes big increases come from 'sparking' the in-line sequencing modulation by setting your clock speed to an even number (never mind 4700 Ghz is an even number--many don't know but it's the 100's column digit that determines the even/odd augmentation algorithm spacing generating a strong sparking response w/ the new try-gate based processing units. Try 4.6 and see what you think...

 

Cheers!

:biggrin:

Arjen,

 

 

With regard to your original polite query. Be sure it's worth the degradation increase on your processor. Be sure you've satisfied yourself as to what sort of performance gains come from that extra 100 or 200Mhz since as we know there is a steep curve towards diminishing returns in terms of heat, voltage, and power consumption. Let us know where you end up.

There is no such thing as safe overclocking, however under 1.5V there is more reward than risk.

 

HLJAMES

There is no such thing as safe overclocking, however under 1.5V there is more reward than risk.

 

HLJAMES

 

That's sort of right, Intel max is around 1.5. However, if you happen to have a sub standard CPU cooler, and are running at high temps, CPU degradation is a concern.

That's sort of right, Intel max is around 1.5

 

I think for my dead QX9650 it was 1.45v as the absolute max voltage after which you were assured premature failure, even after returning the CPU clock to its default frequency. A lessor response but of the same quality happens outside the functional limits of the CPU, ie. with voltages over 1.3625 and under 1.45v

 

I'm staying one notch under 1,3625 E

IIRC it's 1.55 for Ivy Bridge. But of course no one would get near that without de-lidding.

 

I'm at 0.030 offset, whch gives me 1.304.

 

Argen,

 

That sloppy loose memory timing. You can get 2FPS in FSX without overclocking!

 

HLJAMES

Argen,

 

That sloppy loose memory timing. You can get 2FPS in FSX without overclocking!

 

HLJAMES

 

+1....but he has quet fast mems for SB , it can even do more on a IB the guys that run

1600mhz loose timings instead of 2400-2666 MHz and thight them a bit.