Gigabyte TweaKING - Official info thread

[dinos22]

lol i was talking to someone on MSN today about that and he was blushing while saying it might even be the size of that Intel freq tool hahahah
who knows we are clutching at staws but it would be good to find out

[elmor]

Not to be an ass (we had no shot at the top anyhow) but this should be investigated more, we can't have these kind of uncertanties in the live competitions. If I understood correctly there is two screenshots, with one being in my opinion invalid. So let's see the real one?

If something special was done to achieve this it should be confirmed (just like in science, in order for a theory to be confirmed it needs to be repeated by other persons). If it can, then fine and OBR just proved he is a better tweaker than most of us. If not then it's kind of a gray zone cause it should be considered a bug which happens with SuperPi sometimes. If I felt that my SuperPi run was way too good compared to others without any reason I would not feel comfortable winning a competition with it.

[massman]

This was before the event. I'm not only extremely lucky with hardware but can also make great predictions.

I will try to win the lotery tonight ...

[hiwa]

nice pics guys thanks for sharing

i see OBR loops, they are look stable and right,OS or maybe intel tool make it faster
look like suck Dj-F or maybe pwrchip there, as guys told me on msn they was damn bad ics
sam is 32m addict but every error can make pi slow, look his avalible, perfect CW in this run but bad PP.

[SF3D]

Our freq tool was small as well and refreshing slowly, so we minimized it's affect to superpi. So, that can't be reason for good or bad times.

IC's on memory was not that bad. They could run same as most of the totally unbinned BBSE chips.

[Sampsa]

Here is a short summary of times and settings used:

1. Czech: 9 min 3,469 s (19 x 210 MHz, DDR3-2100 MHz x-x-x-x TRFC 74 xT)
2. Belgium: 9 min 6,047 s (18 x 222 MHz, DDR3-2219 MHz 8-9-8-22 TRFC 88 1T)
3. Latvia: 9 min 6,907 s (19 x 210 MHz, DDR3-2104 MHz 8-9-8-24 TRFC 70 1T)
4. Finland: 9 min 7,687 s (18 x 222 MHz, DDR3-2216 MHz, 9-9-9-24 TRFC 88 1T)

Compared to Belgium.. we using 9-9-9-24 vs their 8-9-8-22 gave them 1,7 seconds lead which sounds right.

[Kabauterman]

It might actually be the intel freq tool. Everyone had it running full size and I see OBR is running it smaller size.

We also tryed to run it is smaller Size but it doesn't matter...

[hiwa]

what was Czech tRCD and tRP? 8 or 9 ?

[Pt1t]

Here is a short summary of times and settings used:

1. Czech: 9 min 3,469 s (19 x 210 MHz, DDR3-2100 MHz x-x-x-x TRFC 74 xT)
2. Belgium: 9 min 6,047 s (18 x 222 MHz, DDR3-2219 MHz 8-9-8-22 TRFC 88 1T)
3. Latvia: 9 min 6,907 s (19 x 210 MHz, DDR3-2104 MHz 8-9-8-24 TRFC 70 1T)
4. Finland: 9 min 7,687 s (18 x 222 MHz, DDR3-2216 MHz, 9-9-9-24 TRFC 88 1T)

Compared to Belgium.. we using 9-9-9-24 vs their 8-9-8-22 gave them 1,7 seconds lead which sounds right.

That sounds better like this Sampsa

[Sampsa]

That sounds better like this Sampsa

Lol thanks, you can see that I'm not used to 9 min results with spi32m recently

[Leeghoofd]

what was Czech tRCD and tRP? 8 or 9 ?

If I heard correctly they were not below 9-9-9-24 : which is why this is a mega supper efficient run...

[massman]

9 I think

what was Czech tRCD and tRP? 8 or 9 ?

Settings: 210 BCLK (9-9-9-24-1+ trfc 74) 2100 MHz MEMs, CPU at 3990 MHz time was 9,03.
Our system can pass 32M at 221 BCLK (booted from bios), 2210 MHz Mems at (10-9-9-9-31-1) but score was worst - 9,04 (was stable for 3d01 too)

[Mads321]

Photo's etc. from our danish article here:

http://hwt.dk/literaturedetails.aspx?LiteratureID=14992

- video's will follow shortly

[N3RO]

Flashback:

Fugger's 3DMark score @ GB Competition was explained before?

[Leeghoofd]

it has been let's call it analysed :p

[OBR]

my timings was +/- this: 9-9-9-24-34-5-5-10-21-7-74-5-21-1-46 (CAS-trcd-trp-tras-trc-trrd-twtr-twr-twtp-twl-trfc-trtp-tfaw-cpc-rtl)
but wait for second screenshot ...

i can post whole tweaking list i have ... for registry, spi tweaker etc etc ...

[massman]

Flashback:

Fugger's 3DMark score @ GB Competition was explained before?

Not in detail. People now refer to it as "LOD MIPMAP".

But that's a different discussion.

The strange thing is that 9min6 is actually not that good at 4GHz, some would even say very bad for 2200CL8. OBR's run would be considered 'normal' if done at home, but it was done at the live event with very bad OS.

I have no idea what happened. For all we know, the 'inexperienced' overclocker could have done something all the 'experienced' forgot. I don't know, I can't find a reason why this one OS would be 'normal' and 21 others 'slow'.

[proxis]

obr, didnt you change them to something like 9-8-8-18 with memset for the last run ?

[dinos22]

Flashback:

Fugger's 3DMark score @ GB Competition was explained before?

what a ridiculous thing to say

there is NOTHING wrong with his 32M SuperPi. All you will do now is encourage people to and moan on this forum like every other competition

i feel that i should have never asked about these scores cause it seems to just pour fuel to fire .... my intention was just to understand how they beat everyone else...not how they pulled of a dodgy score to win

what is up with you guys..............why are you posting in this manner when there is clearly nothing out of the ordinary.....the times are lame from everyone for this frequency pretty much and what they did as massman said would probably be considered a "normal" run

also beware that in my original post i wasnt sure if he ran wazza and now i see that he has in fact used OCXTweaker which i can say does work sometimes much better than manual wazza but its OS dependant and not only that....sometimes it needed an OS in a certain state which i could not replicate consistently to be run properly and didnt bother to find out wtf was up with that.........very odd but if he looked at his Task Manager he would have seen that system cache and available memory don't "behave" the same way as your run of the mill manual wazza would.....

[massman]

also beware that in my original post i wasnt sure if he ran wazza and now i see that he has in fact used OCXTweaker which i can say does work sometimes much better than manual wazza but its OS dependant and not only that....sometimes it needed an OS in a certain state which i could not replicate consistently to be run properly and didnt bother to find out wtf was up with that.........very odd but if he looked at his Task Manager he would have seen that system cache and available memory don't "behave" the same way as your run of the mill manual wazza would.....

I tried both manual and OCX utility in the event. First run was 9.09-something with OCX tweaker: 222BCLK 9-9-9-24 1T. No giant difference for us/our OS.

[Sampsa]

Dinos, calm down I don't think anyone is accusing anyone, just interested and trying to understand where those couple extra seconds were shaved off from. It's best for everyone that things are being discussed and kept open

[Gautam]

Not to be an ass (we had no shot at the top anyhow) but this should be investigated more, we can't have these kind of uncertanties in the live competitions. If I understood correctly there is two screenshots, with one being in my opinion invalid. So let's see the real one?

If something special was done to achieve this it should be confirmed (just like in science, in order for a theory to be confirmed it needs to be repeated by other persons). If it can, then fine and OBR just proved he is a better tweaker than most of us. If not then it's kind of a gray zone cause it should be considered a bug which happens with SuperPi sometimes. If I felt that my SuperPi run was way too good compared to others without any reason I would not feel comfortable winning a competition with it.

this has to be said

P.S. so does this

Dinos, calm down

[Ufon]

Gigabyte TweaKING Paris 09 in fifty pictures:
http://www.extrahardware.cz/cechoslo...aking?page=0,1
(and from Czech and Slovak point of view. You can use Google Translate Czech to English, if you're interested in the descriptions)

[Trouffman]

fifty pictures ? 'only' ?

[Marco André]

Thanks for sharing the pics and some settings
I think all teams did a nice job on OS tweaking because you don't had the time we usually take when benching at home.

Nice event from Gigabyte!