hmm and no screenshots.. interesting...
fantastic contribution an discussion..



This post has been edited by cstkl1: Feb 1 2008, 11:33 PM

no dude thats 4ghz 1.23v

dude screenies..


This post has been edited by cstkl1: Feb 2 2008, 10:17 AM

many things said.. in the end clawhammer ,,,, talking about the wind affect.. hmm he really must have whispered to his proc to make the
unknown world record 4ghz less than 11sec proc gone with the wind..

Based on my experience such ppl should be shot.

kekek nothing to understand

oh yeah mr ace..

found out some interesting stunts with maximus..

ah.. geee pity the asus maximus... so many ppl misunderstand and ill treated such a lovely mobo..
we have STUPID asus engineers to blame..

y?? cause they did not mention what the 4 settings meant.

Dram Static Control Read
Ai Clock Twister
Transaction Boost.
FSB termination Voltage.

Example 400fsb
Ok as u know the from reading anandtech that strap 400 default tRD is 10-12

gaygaybyte mobo on auto is 12
asus maximus is 9.66 (estimated)!!!!!... thats y so difficult
the ai clock twister is in between... strong will make it around (9.33 estimate)

on auto we get Transaction Boost Disable = Relax 0/Ai Clock Twister Moderate =trd 9.66 ( estimate)

Transaction Boost = Disable - Relax 0 /Ai Cock Twister Light/Dram Static Control Read Disable = TRD 10..

so when we set all of them auto we are running somewhere above 9, in between strong and 10.

This is y..
High bus difficult (use the anandtech calculation of Trd as example ) on auto..

after 11 hours of clocking.. this is my result.. and all i know is DRAM Static Control Gives another 100mb/s on everest... so i guess its another TRD boost thingy

advice..
Leave it at TRD 12 Strap 400.. and go from there..
Ok as u know PLL voltages is the voltage for the PLL chip that controls the bus of the proc.
NB is like command central that handles the everything to do with bandwidth... bus,FSB and mem.
FSB termination is the terminal value..





Hence.. u need to get a specific value for termination.
not over or lower..
too high BSOD..
too low instability.

Also for flatfinger..
ure trd is definately high..
as anandtech stated.. which i verified.. at 450 trd 7 CL4.. u need 1.6v..so gaygaybyte mobo owners..
ure settings is the turbo ram settings..
try 1000mhz CL4 TRD 7/8

---

Added on February 2, 2008, 11:54 pm
ty kaiser ..

This post has been edited by cstkl1: Feb 2 2008, 11:56 PM

hmm that means whatever i said cant apply
i do admit never had a low end proc..
lowest was e6600 ..

actually a bit confused.. i had to forget every chareristic of a C2d 65nm.. as it DOES not apply at all on 45nm..
wasted time trying to conform..

maybe tommorow if the e8500 tak jadi.. then i will get one in lyn for fun..
whats the highest clock in lyn ah??

new posting ready in 30 minutes... interesting stuff..

pity flatfinger selling off his proc..

for those that remembered the earlier settings.. the template and posting.. new one chun wei..
theory put in practise.

also want to say something..

posting is sharing..
talking about ure clocks and saying trust me is same like going to puchong food court and buying dvd
and trusting that guy when he say DVD9 good one.. ( he's been selling for a long time)
and thats contributing..
worst still when u defend heavily without any post and claimed the impossible..and makes an excuse by saying flamer etc....heck can call me the devil for i care..

This post has been edited by cstkl1: Feb 3 2008, 01:22 AM

hmm the problem with value rams is the bin.... its hard to overcome it..

800 CL3's were binned by handpicked raw rams at CL3.
same as y some D9GKX are binned at CL5 1150/1200 but clock poorly on CL4 on lower clocks..
careful ppl i have been deemed having baseless statements..

malaysia is next to singapore.. baseless from me..
so baseless means if u dont know.. hence baseless??

well thats true..
heck there is a classic example here in the thread.. depreciation is rapid.

no dude.. hmm price ratio would be getting 667's binned at CL3 or 800 binned at CL4 preferably D9GMH..

hmm get some good proc coolers..the best
then rams..
then worry about what proc and mobo..

yorksfield i only had it for a few days..
and wished i can retest it again.
cause looking at what anandtech post..

can overcompensate the 400bus with 1000mhz CL4 with low trd's

This post has been edited by cstkl1: Feb 3 2008, 02:09 AM

have tried yet agin..
cause no 4.0 is 1.28v.
gaygaybyte is still plug and play on loose timings..

their x48 is still slow and the worst.. heck even slower then most x38's

This post has been edited by cstkl1: Feb 3 2008, 09:09 AM

hmmm dude.. dont mess with memset other than

write to read delay.. max 3 dont go lower than that
trd ...
act to act delay max 3
read to precharge delay again 3..

advice go on 4 on all these trd..
and lower ure trd's..

then tighten..

hope u have a lot of voltage for ure NB.

oh yeah u need to play around with the GTL.. in a few cases where it was almost a stable voltage..
just need to shift the phase ..

then first..
tighten the trd based on the speed ure rams were rated at preferably 1:1

learn from there..

current issue with E8400..

the FSB termination voltage
When i go lower than Vcore there is a input/ouput thingy that reduces any overshots by the proc according to intel spec.
hence lower vcore was easy..

problem..
NB .. needs higher FSB termination voltage.... and with the proc having so low vcore.. errr..


hmmm.
500fsb 1:1 mem .. still not stable..

a few ppl been asking me whats the difference with OCCT and Orthos
basically this is OCCT

run Orthos priority 7/8 with memtest windows 256mb on 25 minutes..
1st minute for temp monitor.. last 4 minutes for cool down..

pass this and u pass OCCT 30 minutes guaranteed

better to do this way as u will know whether ure it was ure mem or proc that failed u.

example of brute force overclocking aka plug and play.

[attachmentid=394884]

based on the above u will also know if u can pass OCCT 1 hour.. samething but on 55 minutes.

best way to test if ure like surfing and stressing
if not heck do memtest 1024 would do.

* note.. theres 3 versions of OCCT... some uses 256mb and some uses 512mb..
personally feel 256mb on long periods reserved to a program for stressing is enough..
and dont advice to run d9's too long..

This post has been edited by cstkl1: Feb 7 2008, 01:06 PM

higher priority

its just means it uses most of the unused cycles thats all..
most of the time no program runs on priority 7/8 other than ure OS.
but which OS stresses the core that much .. highest was vista.. around 75-80 percent.

cause getting a e8500... well thats the aim.?? and a e2140( curious actually on this proc)

ures i said pity cause true potential unknown..

uik.. yet ure hitting so high vcore..

have u taken a volt meter and check whats the actual voltage

i know mine on vcore is

0.1375v less 85 percent accuracy with loadline on
and 0.1326 less with loadline off.

mem is 0.0025v more 85 percent accuracy.

PLL and Northbridge.. negligible.. changes +- 0.020v.

whats ure voltage on the 5v rail.

on the list..
if it hits this 2 shops that i normally buy.. i will be notified asap.

i was hoping for a extreme core 2 duo version.. kekek.. so heck with the fsb..
just keep raisin multi although . higher multi the scaling gets worse..
wonder is it the same with 45nm..

second reason is the C1 batch is coming out soon. but one thing..
i cannot on my freezone elite on max on this puppy unless i am on 1.5v..
freaking condensation... that day i thought i board had died!!!... all water on the proc heatsink and the freezone tec blocks...

This post has been edited by cstkl1: Feb 7 2008, 07:14 PM

try running orthos + memtest 512mb/256mb. for that one hour..

u will know whether its ure mem..

if its hang.. usually pll
if its machine check error usually NB voltage..

if memtest error usually vdimm not enough or ure ram is fault to begin with.

true potential.. now messing with the rams timings.. trd 8.. thats tight.. very tight..
all u need was to mess with the GTL and fsb termination..
just ran this to show u..
dont give up on that proc batch..
remember voltage reading on different boards cannot be compared..
each boards true value of their vcore is different..
but this batch is good..
another friend also has this batch .. running on a x38 P5E.. about the same as mine now..
it scales very well.


E8400 Q746A332... heck got ppl reach 4800ghz on air in japan etc with the same batch and vid..

also another thing vid is taken from the ACPI table reading on the bios.. each board is different.. vid is based on ure box max vcore recommended.. usually theres a 0.1v min on core temp. error..
different board different readings.. some even change base on overclocks the vid goes lower..
best way to see the vid is the box actual voltage itself.. hence y super chips with low vid sometimes are different on different boards.. cause the voltages vcore was not true..
thats y a lot of extreme overclockers usually look at high fsb proc that scales well. that usually is the same for any board..

so dont give up on the board or proc.. .. theres a way if u really want it..
overclockers are meant to use cheaper boards... true ones anyway..

[attachmentid=395123]

This post has been edited by cstkl1: Feb 8 2008, 12:13 AM

depends on trd
give i go lax like trd 10.. less mch..

this one is 1.57.. it overvolts..
for trd 10 at CL5 its okay at 1.5v...


actually posted above not to show proc stability.. i tested that already up to a few hours..

now getting the rams stable at highest trd ( means lowest no).. is the priority
then will go for 8.5 multi and 9..

damn pissed that this board cannot do trd 7 on 500fsb..
theres a new bios coming.. also cannot use strap 400 and now using strap 333 which is y the bandwidth is so high
and the trd is at 8...

Asus Engineers Testers already said new bios is on the way to correct some things.. one of them is the refresh cycle time



This post has been edited by cstkl1: Feb 7 2008, 09:45 PM

dude its on the screenies..the vcore and vcmh aka northbridge

asus has a few settings for trd 8
this is the second lowest

theres two more stronger... for rams.. much faster

example.. here is on the stop sp1m.. beat my previous.. and it can go lower..
all this on a non tweaked OS.

[attachmentid=395131]


the vcore u see is on idle on the.. cause i disabled the loadline calibration.. on load it will go down 1.264.
will settle the loadline calibration once i determine what vcore..

read edgeofstability.com good read on mch/gtl...



This post has been edited by cstkl1: Feb 7 2008, 09:55 PM

safely i think 1.4 or lower depending on cooling..
not because of heat cause

most e8400 running at 1.119v on stock..
try imagine 1.4/1.119v = 25 percent.. thats like on 65nm e6850 1.3v*1.25= 1.62v...

well until somebody burns their mobo or proc.. we have no idea.. but the theory is this..

the mch on nb depends..
as i said i am running the second lowest . the next two i need big BIG jumps.. on nb voltages..


This post has been edited by cstkl1: Feb 7 2008, 10:47 PM