Gigabyte TweaKING - Official info thread

[Leeghoofd]

Congratz to the winners and to Gigabyte for a flawless event

Thanks all for the nice comments BUT

there is no overal champion lol.... it's only a king ( and if you will see the pictures a queen lol ) per round...


How it went for team belgium :

Preliminary test : First bootup and tests with the UD6 (F3 bios) high PCI-E possible, high Bclock but we had to flash to F5D... lost some of the bclock and PCI-E freq... but decided to not backflash.

Pi32mb run was run at 220 Block bootup ( 222 didn't boot, so we used setfsb to up, looking for the last 40Mhz ) 1.42Vcore, 1.66VTT, rams at 2220 C8-9-8-22 1.9volts ( we didn't have these rams for pretests so we were on unknown grounds)

We were aiming for 9 min 5 secs ( pretests at home with 210bclock rams at 2100C8-9-8-22 ( OCZ Blades ) we got 9 min 9 flat. After we found a working Bclock ( upping the VTT in the red zone did the trick ) we could start the tweaking... Sadly for us we found out too late that the Kingstons did even 7-9-7-22 at 1.86Vdimm. Our last run ( completed out of time ) was already 1.3 Secs faster (2/3 down the test) than our best time... so we estimated it to end at 9min 4secs or something... Still not enough to beat OBR's team score

Round 2: we kept running 222Bclock, upped PCI-E to 115Mhz and started GPU testing the card was stable at 800GPU, Shaders had to be kept at 1600mhz otherwise immediate freezes would occur. We did our first run, took the screenshot and tried to retest Car Low... but the System crashed each time...

Looking for a solution we lowered GPU clocks : to no avail, Bclock 220 then, no go for a rerun. Loosening timings, nothing really helped... Due to the numerous BSOD's, freezes the Windows OS was getting sort of borked and in the end not even the Intel frequency tool would maximise... Really annoying as we didn't have the time to do a restore via the ghost program...

I had heaps of fun, was nice talking and drinkingwith you guys

till next time

[Marco André]

Any pictures to share with us?

[Leeghoofd]

Our 3rd team member ( dutch press guy ) will do a review on the event, I'll ask Danny for some piccies as Massman's camera had some issues.

But I think trouffman will beat me to it

[dinos22]

it would be nice to hear from winners then

i am really confused how OBR and hey managed to beat the competition by such huge margins..........

also i am a bit surprised that the superpi times were a bit out......they should have been around 9m flat mark if you guys did in fact have 1100 on RAM
can someone tell us what submitings you guys were running, how did you prep the OS....

[BenchZowner]

I think some people forgot to setup D3D Overrider

[proxis]

also i am a bit surprised that the superpi times were a bit out......they should have been around 9m flat mark if you guys did in fact have 1100 on RAM
can someone tell us what submitings you guys were running, how did you prep the OS....

dont forget that given os'es were jam-packed with drivers and you had to have an intel frequency tool and cpu tweaker open at all times (even during a 32m and 01 runs)

[dinos22]

dont forget that given os'es were jam-packed with drivers and you had to have an intel frequency tool and cpu tweaker open at all times (even during a 32m and 01 runs)

in know that of course heh
they did have some time to prep the OSes

[Monstru]

it would be nice to hear from winners then

i am really confused how OBR and hey managed to beat the competition by such huge margins..........

also i am a bit surprised that the superpi times were a bit out......they should have been around 9m flat mark if you guys did in fact have 1100 on RAM
can someone tell us what submitings you guys were running, how did you prep the OS....

Well, I do not know about SuperPi, but I do know why the guys from Turkey won with such a big margin

They were running 222, 2220 8-9-8, and they really tweaked the 2k1 runs, so the results is almost perfect. We were running almost same specs with them (222, 2220, 8-9-7-20, tRRD 3) and we were having some....nice fps (ask hey and tosuner_mc, that saw what we were pulling there), but we did 3 horrible mistakes:

1 - we played it all or nothing, and in the end it was nothing.
2 - we didn't get the hint when the tests would random fail after 4-5 runs, and we would not take the BCLK 1-2 MHZ lower like Belgium did. We were such egomaniac stupid bastards that when we have seen hey's fps and our fps we were 100% sure that we will win the 2k1 with 82k. We were so sure about that, that we didn't even realise our problem was with BCLK not VGA, so we wasted all the time trying to find the sweetspot for the VGA. We could boot at 222, we could run 32M with tighter sub-latencies then 2k1, we would pass any test if run separately, we would pass lobbies, nature and CH, but we would ALWAYS fail at the beggining og DH or CL if run in "tweaked" order!!!!!!
3 - we DID NOT try the oldest fkin trick in the book. We didn't try letting 2-3 minutes cool down breaks between runs.


Regarding SuperPI, we were just stupid. We wasted most of the time trying to run 235 BCLK. When we realised we cannot do it, we wasted more time tightening the latencies on the ram. After that we had a run (9.15), after that we had another run with copy wazza (shaving 5 seconds, so 9.10.3 smth), and when we should have bring in the heavy system tweaks and shave 3-4more seconds, out time was up.

So, from my point of view, even if usually we have bad luck in this sort of competitions, this time it had NOTHING to do with luck. We were just stupid and inefficient, so we deserved to achieve nothing.


But enough about us. I REALLY want to congratulate Czech and Turkish for their result, they really played a GREAT game. They were far in front if everybody and they did an extraordinary work! I know how hey and tosuner won in 2k1, I do not know yet what OBR used in Pi32M, but I am really curious , because it worked great.

Also, BIG congrats to the Belgium team, you guys really played a very intelligent game, and you deserve all the credit for that. Also, my respect for Norway and their 2k1 performance, and for Latvia and Estonia for their PI32M performance.

All my respect and my congratulations for everybody in this competition, everybody worked hard and played nice. Also, BIG congrats to Gigabyte, this was an extraordinary event and I am proud that we took part of this.

Thank you for this great experience everybody!!!

And do not forget, the party is in 7081, Romania's room (at least we didn't suck at that )

[dinos22]

shame man
was really cheering on for you on the IRC heheheh

[Monstru]

That's it man, turn the knife around, I deserve that

But thank you for the support bro, I know you believed in us!

[dinos22]

i think what a good OC event needs is the following

1 hour morning test so everything runs ok bla bla bla and then function (during which everyone could prep the setups and occasionally get interrupted) and 5 hours for the competition for say two rounds with official timetable...say 3d first, superpi second but you are allowed to go back and rebench the first round and submit any time you like.........this is what MOA was supposed to have but they kept changing their minds....that was just a perfect format imo


in this timeframe you have time to check the components, prep them for subzero, prep the OS and also quickly speedbin your board/RAM/GFX..........

i reckon baring bad luck this would be very close to a perfect kinda comp format where it would be luck and skill spread more evenly and make things even closer/more exciting

[Gautam]

I hope they do more contests like this around the world. I'd love an all air contest...not even I'd be able to kill hardware then!

[yotomeczek]

Congratz to ALL OCers and to Gigabyte for a flawless event
It wasn't perfect tweaking competition - but still very good (for example not all CPUs hit 222 BCLK and so on ...)
Paris city of love ... and overclockres doing what they love best
You ware GREAT

[K404]

I'd love an all air contest...not even I'd be able to kill hardware then!

Good work, Captain Jinx

[SF3D]

Hi all!

First I have to apologize, that one of my comments to solid have made such a huge mess in here. Sorry!

I will soon send e-mail to bulldog, solid and puunakki, cause they are the ones who have been misleaded in the beginning. I have nothing to do with selections and I heard the whole story in Paris, so I can tell more about that to these guys. It is nothing, that needs to be talked in here and there is only few misunderstandings in there. Or, so I believe.

I have never had problems in Finnish forums, but of course you cannot please everyone. So it is totally understandable, that some people in Finland might feel abandoned sometimes. I hope those guys would remember that you have to do a lot of work to get where you are. But that is it, no more this off topic here. If someone have something to ask or say, there is always possibility to send PM's to me. I usually answer to all of them.


Then some talk about competition itself:
Competition was very interesting and different. Level of it was really high, even we were so limited by frequencys.
Too bad we lost our 01, because of some weird D3D overrider issue. Luckily we found out the issue with elmors help after the competition, so it is not that frustrating.

Good work Gigabyte! Thanks to Kristina, Phillip and all the other GB crew!

[RichBa5tard]

there was no overall category , 32m and 3dmark01 were separate competitions , two champions (czech and turkey) were crowned () etc

trouff and hwbot - please fix : 3rd place in 32m was taken by Jaan and Sam , not Octopus and Sam

Corrected, thank you.

[Hyperhorn]

I hope everybody arrived fine and enjoyed the last days. I went through my photos (>400) and I`ll upload some pictures later then of course. Unfortunately my cold is getting worse, so I really have to rest now. I need a Lido nurse!

@Gigabyte: Thank you very much & special Thx to Kristina!

[Leeghoofd]

it would be nice to hear from winners then

i am really confused how OBR and hey managed to beat the competition by such huge margins..........

also i am a bit surprised that the superpi times were a bit out......they should have been around 9m flat mark if you guys did in fact have 1100 on RAM
can someone tell us what submitings you guys were running, how did you prep the OS....

Once Hicookie puts up the official screenshots all the timings will be reveilled, belgium team maybe messed up by running TRFC too high ( 88 ), our last non official run with TRFC sub 80 nibbled off easily close to 2 secs of our official time... ( inexperience with the ram, I pretested with Hypers and thought we got the 6gb PC1600C8 kit from Kingston : FAIL lol )

Rumour is that OBR and co ran really loose timings (read C10-10-10-...) Maybe this allowed them to get away with a really low RTL value ( we couldn't touch ours in the bios, as that would result in a no post or an infinite loop ) I was nicely surprised by them rams as my previous K rams weren't that performant (read OC potential )

Thx Bernice and Sibren from Gigabyte Holland for having me. Massman to support my snooooooooooooooring during the short sleeps we had in our superexpensive hotel rooms...

Really nice to have met you guys, especially Petri aka the viking giant impressed me ... Looking forward to all the piccies...

Good luck Monica in Paris !!



If there's a next time for me, I'm going for Monstru's beer (again)

[massman]

But I don't understand why they didn't win the 01 round, with Massman's mad P55/01 skills

Oops!

We couldn't really get our tweaking skills out as during the 70 minute competition we were able to run the complete benchmark twice. 80806 and 80860, but with a poor Car Low. If I estimate correctly (judging by the optimal subtest scores we had already), we could have come close to the Turkish first score of 81499, but their second one was a display of absolute skill.

It will be interesting to see the subtest details, because we can all learn from this competition!!

By the way, the reason why we came in second twice was because I was mocking the Polish team for being twice in all competition. As Chaos said: "What comes around, goes around"

[massman]

can someone tell us what submitings you guys were running, how did you prep the OS....

Basic preperpations we used:

- Pagefie to slave partition (512-512)
- LSC + two other reg tweaks (remove pagefile and something I forgot)
- Maxmem to 615
- Manual copy-waza with folder of 600+ MB

- d:/ -> d:/folder (two times
- d:/folder cut to c:/
- c:/folder copy to d:/
- Best ratio was 543/547

- Disable most services (only event log - RPC - shell hardware detection)
- Set theme to olive (doesn't change that much, IMHO)

- LOD=4.6 + D3D
- Run order: CH-N-LH-DH-LL-DL-CL

Hardware configuration:

- Boot at 220, 222 didn't work, and setfsb up to 221.9x18 to have not over 4GHz.
- Enable three cores. 0/1 for 32M and 2 for applications as Intel Frequency Tool and CPU-Tweaker (which had to be running during the 32M run)
- Memory 2:10 1.9V 7-9-7-22, TRFC=88 (77 also worked in last run), TRRD=4 or 5. Of course B2B to 0 and RTL by auto (couldn't be changed because no boot; around 52, I think)
- VGA: 800/1600/1350 in Riva; something like 775/1584/1350 real.

[over@locker886]

Basic preperpations we used:

- Pagefie to slave partition (512-512)
- LSC + two other reg tweaks (remove pagefile and something I forgot)
- Maxmem to 615
- Manual copy-waza with folder of 600+ MB

- d:/ -> d:/folder (two times
- d:/folder cut to c:/
- c:/folder copy to d:/
- Best ratio was 543/547

- Disable most services (only event log - RPC - shell hardware detection)
- Set theme to olive (doesn't change that much, IMHO)

- LOD=4.6 + D3D
- Run order: CH-N-LH-DH-LL-DL-CL

Hardware configuration:

- Boot at 220, 222 didn't work, and setfsb up to 221.9x18 to have not over 4GHz.
- Enable three cores. 0/1 for 32M and 2 for applications as Intel Frequency Tool and CPU-Tweaker (which had to be running during the 32M run)
- Memory 2:10 1.9V 7-9-7-22, TRFC=88 (77 also worked in last run), TRRD=4 or 5. Of course B2B to 0 and RTL by auto (couldn't be changed because no boot; around 52, I think)
- VGA: 800/1600/1350 in Riva; something like 775/1584/1350 real.

Impressive work as always my friend!

Huge congrats to massman and Leeghoofd!

[M.Beier]

Basic preperpations we used:

- Pagefie to slave partition (512-512)
- LSC + two other reg tweaks (remove pagefile and something I forgot)
- Maxmem to 615
- Manual copy-waza with folder of 600+ MB

- d:/ -> d:/folder (two times
- d:/folder cut to c:/
- c:/folder copy to d:/
- Best ratio was 543/547

- Disable most services (only event log - RPC - shell hardware detection)
- Set theme to olive (doesn't change that much, IMHO)

- LOD=4.6 + D3D
- Run order: CH-N-LH-DH-LL-DL-CL

Hardware configuration:

- Boot at 220, 222 didn't work, and setfsb up to 221.9x18 to have not over 4GHz.
- Enable three cores. 0/1 for 32M and 2 for applications as Intel Frequency Tool and CPU-Tweaker (which had to be running during the 32M run)
- Memory 2:10 1.9V 7-9-7-22, TRFC=88 (77 also worked in last run), TRRD=4 or 5. Of course B2B to 0 and RTL by auto (couldn't be changed because no boot; around 52, I think)
- VGA: 800/1600/1350 in Riva; something like 775/1584/1350 real.

Pretty much same... A few changes from us:
LOD 4.8, MAXMEM=666 order; DL-CL-LL-CH-DH-LH-PAUSE-N, 805/1610/1250 (didnt have time for finding gfx ram - rund order we experienced as only stable for our clocks for some reason...)
... surprise, we forgot D3D (yea, live is stressed, especially with bad start)

Automatic CW, 16K filesize 384MB, CDT enabled

PF disabled..
LSC, all other left flat, estimated wasnt worth doing with our very limited time (Gain not worth considering time consumption)
- Removed windows apps (WMP, MSN etc.)
- Boot ini, sorry Gigabyte, but really, EasyTune6 and launch with registry

PS: Next year, really, we need a decent screwdriver, we COULDNT unmount gfx cooler during prep time to change thermalpaste....
Oh and I was suprised no one used the CPU fan for cooling graphics during 01SE, was indeed possible

Once again, really enjoyed meeting you guys, dispite Pieter, really, pierced my heart to loose... I have to get you really drunk at AOCM ;-P

EDIT: 4.8 for nature, think others were 3, cant remember... We couldnt bring laptop, and we experienced some was higher in 4 range, others in 2 range, so just took something in between

[Leeghoofd]

I have to get you really drunk at AOCM ;-P

Where were you when the lights went out in Paris City ?

We were still partying






ps : don't tell my wife

[dinos22]

so what ICs do you guys think they were
i was under the impression they were BBSE

lol 88 TRFC.............i'm sure you could have done low 60s if you tried more :p
what RTL were you on?

settings timings looser will not let you set tighter RTL....it must have been something else that allowed them to run tighter RTL....IF that was the case.....but running cas10 would have killed his efficiency big time no matter what clocks...it's gotta be something else that allowed him

what about other subtimings....yeah let us know what else you set in bios or what you left on auto whatever is easier

u got rid of maxmem for 3d right?

did you guys do any pretesting for run order? unusual order there from both

[A2C]

Great results Massman, have somebody full results? or where finished Slovak editors team?