Of course tweaking will do the job ... :woot:Quote:
Originally Posted by SoF
Printable View
Of course tweaking will do the job ... :woot:Quote:
Originally Posted by SoF
I really hope so, but unless he gets a new board, unfortunatly we can only guess ;)Quote:
Originally Posted by Stelios
No, it's because it's a frickin logarithmic process. If you have something that takes 20 seconds, then 10 seconds is twice as fast. And then from 10 seconds, 5 seconds is twice as fast. And from 5, 2.5 seconds. And so on.Quote:
Originally Posted by metro.cl
I made a little graph that helps illustrates that Conroe is scaling almost perfectly. I took the inverse of the time (time^-1) and compared it to the clock speed. This shows the relationship in a linear light. (You could also graph it on a logarithmic scale and have it be linear, but I thought this was easier to interpret).
http://img125.imageshack.us/img125/5323/conroe2nk.jpg
Sorry for how ugly and unrefined it is. Doesn't compare to iteration's beauties. :)
The only bump in the data is the 4.1ghz speed, and I have no idea why. I threw that out and used linear regression to make a formula (r = .983, very close correlation). Works out to be (time^-1) = .0195138735(ClockSpeedinGhz). So, to figure out when we will hit 10 seconds:
.1 = .0195138735x. X = 5.125Ghz.
So, we should expect to reach 10s (assuming no further tweaking) at 5.125Ghz, +/-10Mhz.
Either way, Conroe is scaling nearly perfectly. It has begun to level out slightly at the top, but I think this may have something to do with the fact that TAM is using much increased multipliers over Coolarer's, so the FSB isn't as fast. But it's mostly irrelevant.
good job explaining and predicting
9.99s spi is goin down:banana:
very nicely done ethernal
Indeed very nicely done, the completion time is indeed a logarithmic process. And your idea to plot 1/time gave me a great idea. Instead we can just track what I'll refer to as DPS (digits per second, lol). For example when you get 100,000DPS that is a 10 second SuperPi_1M run. And that, of course, will also be a linear process.
Check it out:
Er, that is probably a bit more elegant of a solution, eh? :) Nice graphs.
I have a feeling it will be higher than 5.125Ghz looking at the values above. More like 5.25Ghz as a rough guess.
Regards
Andy
nice graphs there
don't forget ram speed. At what ram speed are you using as your constant to figure out those times?
any news on that new mobo?
Nice work there guys. I had been thinking that 9.xxx wouldn't be seen until higher clocks were seen over 5ghz.
Very nice graphics :clap:.
Impressive! :clap::clap::clap::clap::clap::clap::clap::clap: :clap::clap::clap::clap:
Is there any possibility to see some 3dk01 or even AM3? :p:
WOOW !! Superb faster
5 ghz is :slobber: , i think in one week we see one digit pi :)
@SoF-san, and all.Quote:
Originally Posted by SoF
TAM, got new mobo's already.
Thoes are 303 and 304.
303 was moded already, for xtreme cooling. :D
Nice, I really cant wait for the results :p:
:toast: let the ocing beginn!
I'm conroe ready & waiting :worship:
awesome
So where is 5Ghz ? :cool:
c'mon
Com on!
Gimme 5! :D
Nice work, Ethernal and Iterations.
TAM, can't wait to see 5.2GHz ;)
Can't wait for 5.3 :p: