Conroe-L >>> a.k.a. Intel Celeron 460ES Socket 775 >>> AMD Killlaaaar >>> SS rotary

[dinos22]

Maybe ES chips slower than retail.. B2 vs B1ES..

yes that is what i am saying

sooo what tweaks did you use on Win2003 lol
i find that XP Pro is still faster tweaked in 32M but 1M it seems a touch faster

oh btw the difference between B2 and B1ES is around 20-30 seconds in my testing

this is really odd hey

[over@locker886]

yes that is what i am saying

sooo what tweaks did you use on Win2003 lol
i find that XP Pro is still faster tweaked in 32M but 1M it seems a touch faster

oh btw the difference between B2 and B1ES is around 20-30 seconds in my testing

this is really odd hey

I will tell on msn..

OR Maybe have bigger difference between Celerons 1M L2 (es vs retail), than Conroe-s (B2 vs B1ES), what have 4M L2.

[dinos22]

cache does not help 32M at all (well it's insignificant)

Allendale to Conroe is only 3 second difference in 32M at 3.6Ghz
i only use msn at work heheh

[massman]

I will tell on msn..

OR Maybe have bigger difference between Celerons 1M L2 (es vs retail), than Conroe-s (B2 vs B1ES), what have 4M L2.

CDT reverse? (or how is it called???)

Anyway, I've seen other great benchers do fantastic 32M runs at low speeds just by applying the good tweaks.

[over@locker886]

CDT reverse? (or how is it called???)

Anyway, I've seen other great benchers do fantastic 32M runs at low speeds just by applying the good tweaks.

I not ken this tweak, only I heard. I don't know what is CDT, and how help for spi. I will search info about that..

[massman]

Okay then .... what tweaks did yo use then?

Honestly, you've got a véry fast 32M, congrats man!

[dinos22]

I not ken this tweak, only I heard. I don't know what is CDT, and how help for spi. I will search info about that..

i don't think kevin has actually posted about this but a friend and i think we know what he's going on about

anyways we are yet to find the balance for the 3 stages of Pi

just need time to experiement

[massman]

Actually, he posted the tweak on Coolaler's forums, but as it's in Chinese, I had a very hard time to understand.

As far as I know, it's some sort of better Copywaza.

[dinos22]

wazza done right gains me 8-9 seconds

so how much extra would that gain
i doubt much more

when he posts his runs they don't look THAT much quicker clock for clock if at all to others

there is no magic mushroom for 32M superPi but rather lots of RAM tweaking and finding the right balance of fsb/speeds/latecy/clocks/volts

and a few small tweaks which are well knows will do the trick

[over@locker886]

Okay then .... what tweaks did yo use then?

Honestly, you've got a véry fast 32M, congrats man!

I used some "usual" tweak, maxmem, copy-wazza, ramdisk, real priority, largesystem cache=1, almost all service disabled.

Thank you

i don't think kevin has actually posted about this but a friend and i think we know what he's going on about

anyways we are yet to find the balance for the 3 stages of Pi

just need time to experiement

aha, If I finished mods with Asus P5K and have free time, I will try some superpi round

Actually, he posted the tweak on Coolaler's forums, but as it's in Chinese, I had a very hard time to understand.

As far as I know, it's some sort of better Copywaza.

Hmm, interesting..

[massman]

wazza done right gains me 8-9 seconds

so how much extra would that gain
i doubt much more

when he posts his runs they don't look THAT much quicker clock for clock if at all to others

there is no magic mushroom for 32M superPi but rather lots of RAM tweaking and finding the right balance of fsb/speeds/latecy/clocks/volts

and a few small tweaks which are well knows will do the trick

True, but SuperPi sometimes behaves not like you would expect, which has to do with the amount of memory which is available when you start the 32M bench.

For example, some people have better results by minimizing the benchmark while doing 32M. The minimizing on itself is not the big tweak, it's how the free memory behaves.

Try the following (this is how I've understand the tweak, I'm not speaking chinese, so I could wrong, though it's worth trying):

1) Apply all your tweaks
2) Open Superpi
3) Run 16k twice
4) Click Calculate > 32M > Ok, but wait to start the bench
5) Do the copy-waza tweak
6) When the maximum amount of memory is available, press ok to start the 32M

You may have to retry to catch the good run.

ALL credits go to 狂少

[dinos22]

well that is basically wazza described right there

he is not doing anything special in that case

it is not a tweak unless you can replicate it and be able to estimate the time gain

[redratamd]

Do you have a bad windows ?

On air cooling under noctua .

[dinos22]

Do you have a bad windows ?

On air cooling under noctua .

do some 32M as well please

what mobo??

[a1k]

430 box 16.407 3780mhz http://www.htaccess.free.fr/Celeron%...%2016.407s.PNG

[redratamd]

This is runned with biostar TP35D2-A7.
My cpus are coming from boblemagnifique and Frebrez.
I'm testing actually a 480 ES like newbeetle .

[youngpro]

cdt-reverse is slightly different to copy-waza, it involves understanding the three stages of pi 32mb (static, variable and bonus). The static stage is between about rounds 3-10/11 and variable rounds 10/11-20, now theres nothing you can really do to enhance these stages other than ram timings, system cache and raw mhz. But the final bonus stage (rounds 20-24) can be enhanced by cdt-reverse.

It involves the normal copy-waza method, but then when system cache is at its most optimized (usually between rounds 6-8 of the static stage) you quit out of your 32mb pi mid loop and immediatly start another 32mb pi. Now while task manager reads that system cache is abotu half of where it started, it seems to be more optimized and when you finally get down to the bonus stage in your new 32mb run, the last few loops can fly (you can shave an extra 6-10 seconds on top of copy-waza). It's really tough to get a grip on, currently when I do it perfectly im only shaving another 1-2 seconds on top of copywaza, but certain people are shaving 8+ seconds off their time with it)

[dinos22]

This is runned with biostar TP35D2-A7.
My cpus are coming from boblemagnifique and Frebrez.
I'm testing actually a 480 ES like newbeetle .

yeap we definitely have a problem with these CPUs

[dinos22]

cdt-reverse is slightly different to copy-waza, it involves understanding the three stages of pi 32mb (static, variable and bonus). The static stage is between about rounds 3-10/11 and variable rounds 10/11-20, now theres nothing you can really do to enhance these stages other than ram timings, system cache and raw mhz. But the final bonus stage (rounds 20-24) can be enhanced by cdt-reverse.

It involves the normal copy-waza method, but then when system cache is at its most optimized (usually between rounds 6-8 of the static stage) you quit out of your 32mb pi mid loop and immediatly start another 32mb pi. Now while task manager reads that system cache is abotu half of where it started, it seems to be more optimized and when you finally get down to the bonus stage in your new 32mb run, the last few loops can fly (you can shave an extra 6-10 seconds on top of copy-waza). It's really tough to get a grip on, currently when I do it perfectly im only shaving another 1-2 seconds on top of copywaza, but certain people are shaving 8+ seconds off their time with it)

which certain people are shaving off additional time pro

[dinos22]

430 box 16.407 3780mhz http://www.htaccess.free.fr/Celeron%...%2016.407s.PNG

boooooooooooooo hahahah

[massman]

cdt-reverse is slightly different to copy-waza, it involves understanding the three stages of pi 32mb (static, variable and bonus). The static stage is between about rounds 3-10/11 and variable rounds 10/11-20, now theres nothing you can really do to enhance these stages other than ram timings, system cache and raw mhz. But the final bonus stage (rounds 20-24) can be enhanced by cdt-reverse.

It involves the normal copy-waza method, but then when system cache is at its most optimized (usually between rounds 6-8 of the static stage) you quit out of your 32mb pi mid loop and immediatly start another 32mb pi. Now while task manager reads that system cache is abotu half of where it started, it seems to be more optimized and when you finally get down to the bonus stage in your new 32mb run, the last few loops can fly (you can shave an extra 6-10 seconds on top of copy-waza). It's really tough to get a grip on, currently when I do it perfectly im only shaving another 1-2 seconds on top of copywaza, but certain people are shaving 8+ seconds off their time with it)

Thanks mate, that makes a lot more sense .

[dinos22]

Congrats my friend

Try you close me, my Celeron 32M WR

hahaha i figured out your little tweak there buddy

[redratamd]

http://www.hwbot.org//result.do?resultId=699804
Keep pushing guy

[dinos22]

will beat that easy with single stage

but only problem is this board has a problem with this CPU

it does not want to boot past 356Mhz
very weird lol
356*12 should be enough

[RAW-Raptor22]

Ha, ha, ha......They STILL produce Celeron CPUs?......

They are basically core disabled C2Ds, and are quite good chips. The fastest single cores ever made precisely. Faster than even the AMD FX57...