Chiller - 4848.5 MaxxMemMarks MaxxMem

[Massman]

Latency result looks weird: 3x lower than what other users are reporting. Is this a bug in the program or did you use a tweak or OS difference?

[Chiller]

this is my daily setup and daily vista,

only set other cooling fins on my memory,

 

and set the timings in bios lower, not only the 7-8-7-20 65 1T but other settings also.

 

as you can see, i ran this benchmark several times, and getting higher and higher with lower timings.

So i think no bug.

[Massman]

Hm, okay ... then we need to figure out why the others are scoring so low :-/

[Chiller]

i will run more and more tests with maxmem ok?

try to get as high as possible.

 

i'm going to run it now at stock like the settings are 24/7 a day , maybe it is that i have a Gulftown with more L3 cache or so ?

i don't know, the score is everytime above 4000.

 

and succesfull, no errors or anything.

[Massman]

Gulftown might be an issue. Can you retest the exact same configuration (mobo/mem/os) with a Bloomfield?

[Chiller]

don't have a bloomfield at the moment in the house

[westsider]

gulftown ist the issui

thats right

i think latency calculator deosnt work correct on gulftown

because i benched 1048MHz 7-8-7-20 1T and got just 38ns latency with W3540

there is no way to get 11ns latency with bloomfield

[Chiller]

if i had a Bloomfield, i would run it, but i got alone a Gulftown Q3QP unlocked.

 

The memory is Corsair Dominator GT 2000 cl7 3x 1Gb.

 

and everythime the benchmark is correct en submitted succesfull, so i don't think i got a bug.

everytime the run is between 4300 and 4600, and when i get my memory higher it goes to 4800.

 

so maybe when i cool the memory good, i should get hopefully 5000

[westsider]

dude less cores mean more bandwidth and better memory performance

if you try 2 cores on your result would go beyond 5.5K

i believe you no problem

its just i tried it myselves and i am not bad at Ram/system tweaking

maybe somebody gets some more juice out of it but our results of about ~1600-1700

and you have 5K

you know... there is something weird in software

or you mean gulftown has a magical IMC

nothing personal

[westsider]

hier i have done some test for you

2 cores HT:ON

http://i032.radikal.ru/1001/11/ef9f4bb741f3.jpg

4 Cores HT:ON

http://s61.radikal.ru/i174/1001/ea/2cf4655496bd.jpg

 

they are identical setups

only core count is different

and gulftown manages to get 11ns with 6 cores HT:ON ???

if thats true ,i am glad

but i think somehow maxmem code has issue with latency calculation on gulftown

[Chiller]

ok, i try with HT of, lets see, what that does, maybe 5000 who knows.

and submit the score

[Chiller]

you have rigt, less cores, HT off, is higher score.

now past 5000, trying to get higher.

 

and do more testing with less cores

[Kal-EL]

What is that tec contraption you got mounted?

[Chiller]

i tried with less cores, and jou're right, less cores more score.

 

first HT off, then only two cores on.

now you have more then one result of me you can compare.

 

now the score with only 2 cores are of the chart lol.

 

 

That TEC contraction i mounted is a Freezone Elite TEC Chiller, and it cools very good on Gulfy.

With only 2cores activated, the coolant starts to freeze lol.

going subzero, needed to put toiletpapier on the cpu for no ice forming lol.

 

i only change the stock cpu block with a heatkiller 3.0 for better cooling because the stock cpu block is a bad toy

Edited January 2, 2010 by Chiller

[Bartmasta]

how to do it with less cores?

[westsider]

you are welcome dude

i think this tweak is my copyright

nobody confirmed me uptoday that its so.yo uare the first one

less cores somehow produce huge bandwidth

and it helps on SPI1M and 32M very well

nobody tried it uptoday as far as i look

 

 

 

[westsider]

i tried with less cores, and jou're right, less cores more score.

 

first HT off, then only two cores on.

now you have more then one result of me you can compare.

 

now the score with only 2 cores are of the chart lol.

 

 

That TEC contraction i mounted is a Freezone Elite TEC Chiller, and it cools very good on Gulfy.

With only 2cores activated, the coolant starts to freeze lol.

going subzero, needed to put toiletpapier on the cpu for no ice forming lol.

 

i only change the stock cpu block with a heatkiller 3.0 for better cooling because the stock cpu block is a bad toy

 

u r welcome dude

u should tweak it better dude

you have more speed but my bandwidth is higher

i had 1048mhz 7-8-7-20 1T

what are you subtimings?

Edited January 2, 2010 by westsider

[Chiller]

how to do it with less cores?

 

just disable them in bios

[Massman]

The creator of MaxxMem has confirmed that the score is a bug caused by the application: the hardware was incorrectly recognised which resulted in an impossible latency result.

 

Will be blocked.

[Chiller]

that's something you are fast with he?

 

blocked result?

 

 

i have also contact with creator of Program, testing now internal beta for him.

 

he only says cpu with L3 cache doesn't be recognized properly, not bug

[Massman]

Yes, especially when it's you.

 

From the creator to the hwbot staff:

 

Hello,

 

position 1.

 

http://hwbot.org/rankings/benchmark/maxxmem/rankings

 

is FALSE/wrong! but no cheat.

 

it causes by false detecting of cpu.

 

please correct this, i will try to avoid this in the future.

 

He requests your score to be blocked.

[Chiller]

no cheat, false detection of cpu,

 

 

is not my problem, i ran it, and that's my score,

 

simple

[Massman]

It's indeed not your problem. It's ours. We solved it.

[Chiller]

you solved it ?

 

there's nothing solved, just deleted,

so i can't run and submit score now because it detect's my wrong.

 

what's there about solved ?

[Massman]

*zucht*

 

Why should we allow a bugged result?