Super pi 1M @3.6ghz

[v_parrello]

Did you actually look at anyone else's timings lol? The ram timings are actually tighter than most here - he's at 4-4-3-4, the TRFC is 21 as opposed to 30 - 45+ and write to precharge & tWTR are both 9, and TRRD is 1.

I think that sub timings are not enough tighter to explain that difference. Here there are some very good tweaks and a very good SO optimization.

If I have a look only to their screenshot (400 FSB, 3603 MHz, DDR2-600 4-4-3-4) I must say this is impossible, but for sure there are something that we do not see with screenshot that make this to be possible.

[C.h.r.i.s]

With my old P5B dlx i found a tweak by changing at every reboot the maxmem dimension in a special order (somewhere i have a big paper about it),with this method the first run (only the first run) of the 1M was faster about 30ms,but this tweak doesn't work with my old asus commando (13.812s@3608mhz) and with my new P5K dlx .
That's all .

[Gautam]

Strange...though 13.812 is still very very good. (Though is more or less equal in difficulty to 13.843 at 3600)

IMHO though C.h.r.i.s...a few hints here and there are nice but you don't need to share all your Pi secrets. Goodness knows I never do.

[C.h.r.i.s]

Strange...though 13.812 is still very very good. (Though is more or less equal in difficulty to 13.843 at 3600)

IMHO though C.h.r.i.s...a few hints here and there are nice but you don't need to share all your Pi secrets. Goodness knows I never do.

The problem is that from now going ahead when someone will find some new tweaks peoples may not believe and we all knows why...
This is not a good thing
Anyway with the P5K 515x7 5/6 cas5 i'm very slow

EDIT=Sorry for my bad english

[koda]

The problem is that from now going ahead when someone will find some new tweaks peoples may not believe and we all knows why...
This is not a good thing
Anyway with the P5K 515x7 5/6 cas5 i'm very slow

EDIT=Sorry for my bad english

We all know that Pi under Linux is really faster than under Windows.
This fact show us that the hardware is not really the main problem to go faster.

I think that all the tweaks are not yet found ... so C.h.r.i.s, maybe you found something really new to improve your time.

[Enoc]

I think that all the tweaks are not yet found ...

true...and that's the main problem...if a guy comes up...discover a tweak and has a time that no one has...the questioning begins and the accusations fly around, and happens he thinks like Gautam (Does not like to share his tweaks) --> example, then escalates to cheater, guy get bitter and gone or not etc......

let's put an example...can 11:50ms be achievable in 32m pi at 8x500 with P965 chipset ...could it be done? it's difficult to imagine it but does not mean it's not possible ... remember mascaras friend running 4-4-4-9 , maybe a unknown tweak...Gautam saw that post too by mascaras...

-->http://img122.imageshack.us/img122/4774/400032meh9.png

everybody has a different approach to tweaking...

[Gautam]

Hey, I'm always willing to help anyone dedicated...but to be honest, I've never let everything I do out of the bag and I don't think any serious bencher out there has.

I myself have done 11:51.4xx at 400x10 2:3 4-4-4 so the same time at 500x8 does not seem too far-fetched to me...

[Enoc]

Hey, I'm always willing to help anyone dedicated...but to be honest, I've never let everything I do out of the bag and I don't think any serious bencher out there has.

lol..it was an example...

I myself have done 11:51.4xx at 400x10 2:3 4-4-4 so the same time at 500x8 does not seem too far-fetched to me...

try 500x8...

[Gautam]

I never could get that kind of FSB working.

[Stapler]

500x8 at 4-4-4-9 would be much slower than with the same tweaks at 2:3 4-4-4-x, I'd say 3-5s or so from my experience, but I don't doubt mascaras time at all. It seems to me like we ought to be less judgemental when looking at other people's times. Not very many are going to give up their secrets after you start a witch hunt.

[G H Z]

Played for an hour with this, damned if I can figure it out

[ziddey]

damn it i could have hit #4 for the 2mb category but I didn't take proper screenshot when I did it. Too late now. maybe I'll try again next week with p35, but my memory seems to have degradated, even though I never fed it more than 2.12v in all the time I've owned it and had 2 60mm's blowing on top

[Stapler]

Nice one GHZ, that ram aint to shabby.

[C.h.r.i.s]

We all know that Pi under Linux is really faster than under Windows.
This fact show us that the hardware is not really the main problem to go faster.

I think that all the tweaks are not yet found ... so C.h.r.i.s, maybe you found something really new to improve your time.

I thought the same thing,but i was not able to repeat that tweak again

[killingspreez]

hm my OS seems to be too slow...were to get this xplite?

mfg

ah i almost forgot:



mfg²

[JMKS]

Hmm... 2,75Vdimm and 450 2:3 @ i965 (with P35 2:3 it was not that good although it requires only 2,55V [unmodded P5K-Dlx])

450x8 [boot @ 400] 2:3 with Allendale (I really want to test it with Conroe and I think I will shortly ) - the comparision between 2:3 400x9 4-4-4-4; 450x8 4:5 4-4-4-4 and 450 2:3 5-5-5-5 is interesting IMHO (and I will probably do this on Conroe)

And a little tweak for those who think that "1066 strap" is an ideal performance deal (of course it requires a strong NB, in my case it was @ 1,95V, I didn't tried lower ) - [with PMem] FED1403F and "22" instead of "23" in upper left corner is slightly faster . It is some tweak for X6800 xtreme benching I think (for the rest CPUs it will be probably impossible due to high FSB).
Edit: I must test it properly (and I will, probably today) because on my daily OS this is unmeasurable - I will post later if it makes a difference or not (with 4:5 surely it does IIRC but with 1:1 and 2:3 I'm not sure ATM).

15.157s - I had some 15.156 (and no screenshot - too small update ) with 5-5-4-4 instead of 5-5-5-5 but I'm looking for lower time .

[JMKS]

Some comparision, still with Allendales unfortunately but this is optimistic also because with Allendales even with 1M there are some differences in score visible with different memory settings .

With jumson's E4400: [it could be his score if author is going to update the rating]
E4400 @ 400x9=3600MHz, RAM 2:3 = 600 4-4-4-4 [known as golden memory settings I think ] => 15.171s


My improvement:
E6400 @ 450x8=3600MHz [boot @ 400], RAM OCZ Reaper 9200 2:3 = 675 5-4-4-4 @ 2,9V => 15.125s

That score is I think the best I can do until I will find some nice DDR3 .

I hope I will be able to test 450x8 2:3 5-4-4-4 with Conroe shorlty but I'm not sure about it.

[Florent]

Nice results JMKS

Jmax made an very impressive run yesterday :
http://www.jmax-hardware.com/ftpjmax...GP/15.094s.JPG

[JMKS]

Very good run indeed . I can tie it with 451x8 = 3608MHz probably.

Unfortunately for that kind of FSB [510+, not 450 of course] my E6400 requires subzero [and that E4400 will not reach it even with LN2 probably although we will fight for ~5GHz with it soon - 4,71GHz under DI so far].

Nice to see someone still fighting with "slow" Allendales .

[monteboy]

DFI P35-T2R , 2:3 , 1200@CL 4 4 4 4

[Bigchrome]

Scores have been updated.

[monteboy]

and a new update

[dinamite2]

p5k dlx
e6600 l631b022
crucial pc8500
win2k03

[coop]

Best I could do. Cas 5 with 4/5 divider does well.

[Kamel]

think i'm gonna get the 1st place!

CELLSHOCK DDR3 1800
ASUS BLITZ EXTREME