As I recall it, BH5 actually tended to slightly outperform TCCD @ same clocks and visible timings back then. Someone correct me if this is wrong.
Printable View
As I recall it, BH5 actually tended to slightly outperform TCCD @ same clocks and visible timings back then. Someone correct me if this is wrong.
OPB is the only one that is using winXP and the only one with non 2-2-2-x timings, what a bull :slapass:Quote:
Originally Posted by zakelwe
I've made a suggestion to OPB. It involves running at the same speeds as Kyosen's 12.0xx run at 4167mhz with both the X6800 ES and X6800 Retail, also documenting the steps involved. If the retail matches the speed and cant get any further and the ES tops the run, then it could be a cpu abnormality.
BH5 give better results than TCCD at same freq since their timings are tighter
with A64 there is nothing else than 2-2-2- or 1.5-2-2- everything else had a significant dis-effect on performanceQuote:
Originally Posted by Lithan
coolalers s spi times..
5200 , 9.72 SN (July 12)
5300 , 9.64 SN (July 17)
5396 , 9.485 SN (July 18)
5409 , 9.484 SN (July 19)
5485 , 9.204 SN (July 23)
5512 , 9.156 SN (July 28)
5525 , 9.000 SN (Aug 2)
5200-5300 ,, 0.080/100 =0.0008 Sec PER MHZ
5300-5396 ,, 0.155/96= 0.0016 Sec PER MHZ
5396-5409 ,, 0.001/13 = 0.0001 Sec PER MHZ
5409-5485 ,, 0.280/76 = 0.0037 Sec PER MHZ
5485-5512 ,, 0.048/27 = 0.0018 Sec PER MHZ
5525-5512 ,, 0.156 /13 = 0.0120 Sec PER MHZ GAIN..
increasing frequency of lets drop of second/mhz gain because of the architecture ,, so according to above results ,, coolaler should use different tweaks that he didnt use before for at least increases 5409-5485 and 5525-5512,,
statisticly results are like that ,, only differences can be because of new tweaks ,, if they had and used them its ok but if not there is sth wrong..
Quote:
Originally Posted by zbogorgon
It is a miracle of the life :woot:
OPB is a God ?? :eek: :eek: :eek: :eek: :eek:
:banana: :banana: :banana: :banana:
I am sorry, but at this point, every run just doesn't prove anythingQuote:
Originally Posted by JasonDTM
he can film the whole run and everything and it doens't mean anything
right now there are just to many methods to cheat spi run
Quote:
Originally Posted by zbogorgon
Yeah I know that... but as I remember it, we couldn't access all ram timings on most boards back then. So people would test BH5 CH5 TCCD etc all @ the exact same timings you could adjust... and BH5/CH5 always came out significantly ahead of TCCD even when everything was running 2-2-2.
Offtopic, but I think I could match that :) Given that I didnt know a few tweaks that I do now, 25.2 secs is do-able with 3200MHz, A64.Quote:
Originally Posted by afireinside
I pulled 25.015 at 3358, add the extra tweaks and scale down, 25.2 should be easy
Judging by what you've just said, you will not allow OPB to prove his worth? What was the basic intention of this thread then?, if you're going to deny him the right to prove his worth, then this thread should be pulled and never spoken of again.Quote:
Originally Posted by zbogorgon
Zbogorgon did not start this thread. It appears his opinion is that Spi is too easily cheated and that needs to be addressed before it can be taken seriously as a demonstration of skill. I hope he'll correct me if I've misrepresented him there.
I believe vapor simply wants to know why there were three members who's results didn't mesh with the hundred of spi benchers out there, in a benchmark that responds relatively little to tweaking. One admitted he was faking the whole thing. Two other's haven't responded yet. And instead of waiting for their response, it seems the majority of posters have decided to take the stance of "How dare you question something that mathematically makes no sense! My faith is stronger than your science!"
That make already more 1 year that it must proven these scores on the 28GP and CPumark FX55 Claw, the little which it said is impossible :slap:Quote:
Originally Posted by JasonDTM
It simply lowers the frequency to make believe that it is a very good optimizer
It is complicated a to understand ?
If you read both two topics, you would understand why at this point there is no other way for him to proove he is right.Quote:
Originally Posted by JasonDTM
As I see the whole thing, the only way is to come forward with the tweak, so sorry.
I really don't understand why is he preparing to put on line his whole reputation and carrer just for some tweak. If he prooves that his result is result of a tweak, he will be a hero and he will proove all of us that we were wrong.
If I was in his position, It would already be written. And given the time betwen his superpi runs, he wasn't trying that hard to get that tweak
p.s. but only my oppinion doesn't matter, If I were the only one with a doubt, we wouldn't be having this discussion right now, there are many others, with great respect, he has to proove and convince them too
Quote:
Originally Posted by raju
Prove it.
When you have benchers as respectable and well known as OPB and coolaler, and they present results that are statistically impossible to replicate, the results should be reviewed by the people who've SUPPORTED them all this time. XS has supported them constantly and taken their results as true and never made accusations of cheating until this data was brought to the forefront. If they're not judged by a group of their peers, doesn't that give them the right to do whatever they want and never be questioned?
Sometimes when data as outstanding as these are presented, one must try to remove themselves as much as possible from biases toward OPB or coolaler or any of the other Legends around here and look at it mathematically.
Don't give me that "they have tweaks which shaved 4 seconds off their SPi1m time at the same frequency!" BS, either! That is outright impossible. No single or collection of tweaks will knock FOUR SECONDS off your Super Pi time, given the frequency at which you benched has remained constant. It might be attributable to unstable clockspeeds, producing abnormally good results, in which case they're just as guilty for not maintaining that their systems were in fact stable enough to run SuperPi, legitimately. These guys know what numbers they should be getting at 5.5GHz, and when they get something abnormally low, they know something is wrong. And if they post these results as verified and true, then they're only misleading the community that respects them.
I've been doing a crystal mod last week for a VGA card. OC-tools read stock speed but you get lot's of higher performance (~9%) at same clock read by Rivatuner. When I overclock the card I seem get stuck on the same score, but I hit my final clock much earlier (read: lower clocks for the same highscore). Besides the fact that clocks really must have been changed and the software is not possible to read the correct clock (maybe @ boot it reads the BIOS set clock and starts from there on), my display was not able to set 1600*1200 60Hz where with the original crystal I could do this. So changing the crystal does affect the whole part.Quote:
Originally Posted by jmax_oc
Now... if we changed the mainbord's main crystal in a version which is only a few Hz higher, would that affect the whole system, and even the Time general used in Windows? Wouldn't that make it possible to get higher 3D Marks or better superpi times as to it is based on time.
So would this be possible, because, it would be just plain simple if you have soldering skills.
Quote:
Originally Posted by Lithan
Exactly.
It has been boiled down into statistical results, there is no point putting a human face on things and what you want to believe, the only answer now is a statistical reply that can be replicated and shown to be true.
Or else it will always be seen as an anomaly.
Regards
Andy
I had never seen great OPB's & Coolaler's score about 3DMark...
why ? :D
becasue opb sux on 3D lmaoQuote:
Originally Posted by giampa
and coolaler doesnt know how to use solder XD
LOL! So true but I didn't want to say it :clap:Quote:
Originally Posted by guess2098
jmax and wittekakker: Changing PLL's and clock crystals is nothing new and definately not a cheat.
However, should the OS lose its ability to tell time accurately there will be large problems.
As has been stated multiple times in this thread, clarification is the objective. If this is a tweak that they would not like released to the general public then PM'ing it to a trusted individual is not a hard task. Of course keeping the tweak for themselves means it doesnt get a cool name like "The OPB/Coolaler Tweak" though.
Filming a run means nothing aswell ( you can easily have the "extra tweak" loading in the background... you can edit the video [ :D ] , and more ).Quote:
Originally Posted by zbogorgon
IF there's a tweak that gives him the current advantage, the only thing that could prove his claims only by running the calculation with system settings that other members can match ( lets say 5.00GHz ) and post the tweaks that he's using for that run.
A man with balls & self-respect would prove himself in any way at any cost instantly without having a second thought.
Changing quartz to get a FSB WR is a cheat :DQuote:
Originally Posted by STEvil
That's what coolaler did to get 17.4s @ 3600MHz with a conroe 4M
i just got curious, i got to take a look at all the data being used here.
i sorted it out a little differently, i took only those scores made with 5ghz or greater. i only used all data except for OPB and coolaler's scores
the lowest product score i had out of everyone's scores 5ghz+ (lower meaning better), was 50437.
now, how many of OPB's and coolaler's scores are below that
there are 2 of them
for OPB, the WR run...
at 5550 mhz in 8.91seconds for a product of 49450.5 (1000 below )
and Coolaler, the 9.00 run...
and another at 5525 mhz in 9 seconds for a product of 49725 (700 below)
now, for those familiar with box and whisker plots, here are two to look at
everyone except for coolaler and OPB's graph
http://www.xtremesystems.org/forums/...1&d=1166057222
and for the stats people
Q1-50728
Med-51275
Q3-51403
IQR-674
all the data i have, including coolaler and OPB's
http://www.xtremesystems.org/forums/...1&d=1166056345
Q1-50624
Med-51073
Q3-51348
IQR-724
everything said below is concerning the graph including opb's and coolaler's results
now, i dont care about the top of the graph, im only concerned with the lower data, so how much of coolaler and OPB's scores are below the 3rd quartile (value is 51348)
every single bit of it. that means that all 12 of their runs fell into the lower 25%(by looking at the graph). seems a bit odd with so many skilled benchers around.
another thing, if you look at the graph, you will notice a small * at the bottom, that is what is known as an outlier, and it just so happens that this outlier is the supposed "world record".
after looking at all of this, it appears either some supertweak was used or some foul play involved
_______________
if you compare both graphs though, notice that once i cut off coolaler and OPB's data, the graph becomes much more compact on the lower end, and with no outliers
They have not a special tweak... :stick:Quote:
Originally Posted by STEvil