OPB's 3rd attempt got super pi 1m 8.906s and new pifast wr

[Langley]

In my opinion... tons of fan boy are there... only this.
OPB and now? The next run be careful and marks another .4/.5s with the same freq/multi/timings

[kemo]

In my opinion... tons of fan boy are there... only this.
OPB and now? The next run be careful and marks another .4/.5s with the same freq/multi/timings

for how long you have been waiting to say this
________
Chevelle

[Ghost]

In my opinion... tons of fan boy are there... only this.
OPB and now? The next run be careful and marks another .4/.5s with the same freq/multi/timings

langley ...my friends....stop please....

è inutile continuare...abbiamo capito ke so de coccio ^_^

[funkflix]

Really ridiculous what some guys are posting here, and how they do it..

Great time OPB and thx a lot for sharing ur tweaks! I'm a bit confused now, because it is really much, but i think i will find a way to understand all of this stuff!

[lawrywild]

In my opinion... tons of fan boy are there... only this.
OPB and now? The next run be careful and marks another .4/.5s with the same freq/multi/timings

Grow up.

[kemo]

langley ...my friends....stop please....

? inutile continuare...abbiamo capito ke so de coccio ^_^

it is a good thing that you know ?it is useless to continue ?
________
hashish

[Dumo]

In my opinion... tons of fan boy are there... only this.
OPB and now? The next run be careful and marks another .4/.5s with the same freq/multi/timings

Hey Langley Just run your own 8 seconds SDuper PI and stop crapping this thread ok?
Its benching and probly sooner or later OPB or someone else (probly you) will passed 7 seconds, so if anyone doesn't have any positive things to say then take a hike.

@Langley....Your 43 count posts mostly started whit congrats or bleh Post your Super Pi run.
It is not healthy!!

[Ghost]

it is a good thing that you know “it is useless to continue “

yes....but...can you translate the second part? no i think....^_^. Stop please...i'm bored.

[andL64]

the World Is Yours

[kemo]

yes....but...can you translate the second part? no i think....^_^. Stop please...i'm bored.

well I can translate a few words from the second part but it doesn?t make any sense and I don?t care
________
Honda Zest history

[skaktuss]

if someone makes some good job in the name of his recently gone friend and others call it bs... damn you guys have no respect at all!

[gillll]

just out of curiosity if the system had used lets say the gigabyte iram as os/pf partition would the results be better ?

[VictorWang]

I think, cpucore is very very hot over 5.4Ghz, even under LN2 cooling.
doesn't matter if you turns off C1E or CPU internal thermal control in bios or not.
Intel cpu always protects itself very well.
maybe that's one reason why OPB's 5.45G pi_1m faster than coolaler's 5.5G.
and higher fsb/different mobo/higher ram_clk and better ram_sub_timings.
besides,conroe runs slower if you disable one core, 'coz there is always other tasks in the OS.

beside, OPB is the master in OS optimize.
but 88MB mem usage before pi running is not something special.
mine=33MB as bellow, normal system start and with memset running :


still Congratz OPB's great results.
wish I have time to play with u guys... but ..........

[VictorWang]

and here is the System service opt thread on my blog:
http://vic.expreview.com/read.php?34&part=2

[Onepagebook]

Dang, I hate you victor..you deserve as XL no doubt.

[thunderstruck!]

Dang, I hate you victor..you deserve as XL no doubt.

I'll second that. Nice work Victor.

[K404]

and here is the System service opt thread on my blog:
http://vic.expreview.com/read.php?34&part=2

Damn I wish that was in English

K

[Onepagebook]

Damn I wish that was in English

K

That part is what he taught me..honestly



OT:
He will do the translation if you find any girl sing for him.

[Maxxpower]

ok...I've seen the validation...so..the wr is your opb...excuse me for same reply...


sorry for my bad english

maxx

[Enoc]

congrats OBP...

[Onepagebook]

ok...I've seen the validation...so..the wr is your opb...excuse me for same reply...


sorry for my bad english

maxx

No problem at all, I hope everyone just no such bad feeling.
Forget about everything and if you guys need my hands to help,
I will definitely do my best.

Thank you for your understanding.

OPB

[v_parrello]

I red this thread with a lot of interest.
First of all OPB, congratulation man, you are best in the world for SuperPI 1M and one of the best overclockers in the world . Anything to say about your world record, only an insane curiosity to know how you made this incredible tweaking (I know thath you will not share your secrets with everybody, but after a lot of hard work this is the right thing to do).


My great curiosity has been provoked from a previous post that I red in this thread. I want to stress and to underline that my following argumemtation is only because I would like to know more about these arguments, and I am shure that the score posted in the previous post is true and is not a fake (congratulation also to this man, coolaler, that is at the moment the best overcloker in the world ), but there is something that I can't figure out how it is possible. Before the argument opened from this thread I had not never watched to this score with attention, and after the tiscussion started in this thread I watched with more attention to the screenshoot of previous superpi 1M WR, that is contained in the following screenshot (I underlined what I can not figure out ):


I try to explain better what I cant figure out and I am shure that there is an explanation related to the tweaking used to make the previous 9.0s world record. The last two run have been made with the following times:
loop 19 in 0.642 seconds
last loop in 0.140 seconds
In understand ( I don't know if I understand right) that the curent version of SuperPI 1M makes calculation of 1 Million decimal digits of pi calculation using Gauss-Legendre algorithm. This algoritm as second order convergent nature. Then if you want to calculate up to n digits, iteration count of the order log2 n is sufficient. E.g. 19 times for 1 million decimal digits.
To calculate superpI 1M it is necessary an initial run for value setting, 19 loops are necessary to calculate the digit with the desidered accuracy and the last loop to close the calculation with pi value.
Now if there is a tweaking to minimize the last run of SuperPI 1M this will be very exciting and I think that your record soon would be exceeded from Coolaler .
What I mean is that the first run and the last run of SuperPI are different from other loops in the middle, in every score I have seen the time necessary to do the first run is very short, for example: 0.125S in your WR, it is the same in S.O.T. score and in coolaler score. Instead the last run was done in 0.375s in your score, 0.390s in S.O.T score and 0.140s in coolaler score (incredible).
In the followings a chart with the run of the best three score in the world for SuperPI 1M to better understand what I mean:

from previous table it is easy to understand that the last run time of coolaler it is very impressive.
At the end is 0.140s time a result of a tweaking or is this just casuality?
Please tell me if my argumentations are totally wrong, or if is something true explain me some about this tweaking for last run.

Tanks in advance and again congratulation for your new world record.

P.S.: sorry to everybody fom my little english

[K404]

Thats a really useful table! Seems like Coolaler only loses out compared to SOT right at the end, despite his excellent finishing time (loop 19-> finish) OPB is ahead from the get go, suggesting an improved set of bench tweaks.

*Maybe* a Maxmem/page file tweak separates SOT and Coolaler?

-> you`re English is fine!

[mursaat]

If you add OPB's "loop19" and "calculate pi" it adds 0.781.
If you add coolalers "loop19" and "calculate pi" it's 0.782.

Something between the cpu, caches and memory made the loop19 to be slower and the calculate to be faster. Don't ask me why because my best 1M time is with an Opty165@3ghz

[cpulloverclock]

I tested some tweaks and for an 1M :

- "Diagnostic Startup" is worse than "Normal Startup" , you must remove some processes (userinit.exe, svchost.exe, mdm.exe, explorer.exe, spoolsv.exe...).

- "Prefetcher"=1 is the best.

- Change "Processor Scheduling" to "Programs" and change "Memory Usage" to "Programs" is better than "Background Services" and "System Cache" (but I dont know if that will cause some problems for the highest frequency for a bench and screen).

- In "Performance Options", just "Use visual styles on windows and buttons" and maybe "Use drop shadows for icon labels on the desktop".