SuperPI Contest!!!

[IEC]

Originally posted by: ariafrost

Originally posted by: Shenkoa
Athlon XP 2700+ Tbred
512 MB Corsair PC 3200
80 GB WD800JB

44 Minutes and 52 seconds. Crap I know!

Looks like I have some competition here... that's basically my specs except I use XMS... (*goes to download SuperPI*)

*** CONFIG ***
Athlon XP 2700+ Tbred @ 2312MHz (171FSB, 13.5X multi)
ABIT NF7-S Rev.2 Mobo w/Tictac D26 Manta XT BIOS
2x256MB Corsair XMS PC2700 LLPT
Radeon 9700P
80GB WD800JB SE

*** RESULTS ***

Meh. 39m 58s.

 

[zakee00]

bump

 

[deadseasquirrel]

Ugh. Can't get a 32M to run even at stock! I quit prime at 22 hours. It runs all the other SuperPI tests. I've looped 3dmark for 9 hours. And this was all overclocked to 2.5ghz.

But it just doesn't want to complete a 32M run even at stock. Ah well, I wasn't gonna win any contests anyway... it's just annoying now.

edit: LOL it just died on me on the 24th iteration! Last one shows "29m 36s Loop 23 finished". So I suppose I could assume a 31m result. But, dang it, I want it to complete!!

 

[Whizzz]

If you want to see my computer!

I still have some work to do but this case is temporary anyway. I'm getting a 1 Gig kit of G.Skill pc 4400 as my Corsair Twinx 2048 pc 3200 C2PL just isn't cutting the mustard right now.

http://ca.pg.photos.yahoo.com/ph/thelle...ph/thellendfamily@rogers.com/my_photos

AMD 25m 05s with FX55@3.15 MHz, currently ORB 3DMark 05 @ number 05 with 10,451, man that's a lot of fives!

PS: I will be number one in the contest after I get my RAM next week.

Whizzz

 

[sman83]

will run it when i go to bed tonight and post mah results

 

[superkdogg]

Originally posted by: TekDemon

Originally posted by: superkdogg
Is this time o.k.?

$345 for Mobo/CPU/Memory

Wow you got an AV8 to run a 300Mhz FSB?
I'm debating whether to get that or the Gigabyte K8NS, and was worried that the AV8 couldn't quite pump the higher FSBs...

I appear to be pretty lucky in this. Even people with similar rigs who have PM'd me couldn't match my results with the same settings. Oh well

 

[highlandsun]

OK, 31:48 for my A64 3000+ at 260x9
screenie

 

[superkdogg]

Originally posted by: DCypher
Well, thanks for the reply, but didn't really help, the question was whether SuperPi liked tighter timings better, or high RAM frequency.

Read Here
About 1/4 of the way down is 1M scores. They can be extrapolated for your answer.

 

[chinkgai]

aww crap, after some tweaking i've managed to beat my original time of 29:09 and got 28:18

but thats not even 5th place anymore

looks like i gotta tweak some more...

one thing i noticed about superpi (for a64) anyways...the dual channel of the 939's actually make quite a difference...

 

[Duvie]

I found that tweaking cas timing was key....cas 2 is great, but often times to run tight cas 2,2,2, or 2,3,3, timings you have to be between pc3000 to pc3500 levels only....So if you can get cas 2.5 to run at pc3800-pc4000 levels you may actually do better cause the bandwidth will be much higher....

I was testing 9x296 for 2664mhz....I was at cas 2.5,3,3,6 and though I had found my ultimate stability and score...

time: 28:12 seconds.....

Then I downgraded my bios from 1.51beta to 1.41mod (seeing a comment about it being the best for TCCD ocing....

Time: 28:05 bandwidth in the 7200-7100 range

Now while I couldn't get 2.53,3,5 stable before this time it was.....with the 1.41mod bios that is...

http://home.comcast.net/~duvie23/A64-superpi-9296-2664-166-2.5335.JPG

Time: 27:54 bandwidth hasn't changed.....


Then I said well that is the lowest timings I can go...no 2.5,2,2,x will work....So I am think to myself I am at a 166 divider now...I know I can't do 200 but It would be nice to have the memory go faster...In comes a nice app called A64 tweaker and the very nice 183 divider!!!! Now I noticed the app didn't like a whole lot of tweaking the already pretty low cas trcd, and others....

I switched to 183 which should raise memory speed....it did but obviously 2.5,3,3,5 may not be stable...I actually found no go on 5....6 finished but didn't seem that fast and memtest confirmed errors....7 finished no errors...

http://home.comcast.net/~duvie23/A64-superpi-9296-2664-183-2.5337.JPG

This is where conventional wisdom goes out the door I guess...I actually tested 2.5,3,3,8 with all other settings the same...

http://home.comcast.net/~duvie23/A64-superpi-9296-2664-183-2.5338-141mod.JPG

record time 27:28.....I tried 9 and got better then 7 still but not better then 8....

 

[chinkgai]

my mobo cpu is dfi lanparty nf3 and 3400+

here are my settings at 2750, where i got my 28:18:

275x10
9:10 divider so ram is at 250fsb @ 2-2-8-2 1t (2x512)
htt multiplier 3x

i compared my score to zebo's and he has a lower clock speed, slower cas, but at a higher htt and he still wins at 27:20

i tried 299x9 (2691mhz) with a 5/6 divider so ram is again at ~250 fsb @ 2-2-8-2 and actually got in the low 29's....basically i tried it at this setting to try to match zebo's htt and ram, but beating him in both to see if i can even get close to his time of 27:20

like i said, i failed in the second attempt with very similar settings, so i had to conclude that the 939's dual channel was the culprit in this benchmark

EDIT: one other attempt i tried was at 265x10 with ram at 1:1 timings at 2.5-3-8-3 and i got 29:09

 

[Duvie]

Well definitely the bandwidth is the factor here...I pull 7500's at about 533ddr cas 2.5,3,3,8 and you probably only see what 5000???

Bandwidth may not look like a lot in zebos testing but that is because he only tested 1mb...Tha fact is it has a factor...seems timings have a bit more of a factor, but cioupled it make the best tweak...

 

[Icepick]

A64 3500+ (90nm) at 2.500 GHz (250x10)
RAM is OCZ EL at 3-3-3-10 at 208MHz

Super PI: 30:44
Super PI (AMD patch): 30:10

 

[Zebo]

Originally posted by: Duvie
I found that tweaking cas timing was key....cas 2 is great, but often times to run tight cas 2,2,2, or 2,3,3, timings you have to be between pc3000 to pc3500 levels only....So if you can get cas 2.5 to run at pc3800-pc4000 levels you may actually do better cause the bandwidth will be much higher....

I was testing 9x296 for 2664mhz....I was at cas 2.5,3,3,6 and though I had found my ultimate stability and score...

time: 28:12 seconds.....

Then I downgraded my bios from 1.51beta to 1.41mod (seeing a comment about it being the best for TCCD ocing....

Time: 28:05 bandwidth in the 7200-7100 range

Now while I couldn't get 2.53,3,5 stable before this time it was.....with the 1.41mod bios that is...

http://home.comcast.net/~duvie23/A64-superpi-9296-2664-166-2.5335.JPG

Time: 27:54 bandwidth hasn't changed.....


Then I said well that is the lowest timings I can go...no 2.5,2,2,x will work....So I am think to myself I am at a 166 divider now...I know I can't do 200 but It would be nice to have the memory go faster...In comes a nice app called A64 tweaker and the very nice 183 divider!!!! Now I noticed the app didn't like a whole lot of tweaking the already pretty low cas trcd, and others....

I switched to 183 which should raise memory speed....it did but obviously 2.5,3,3,5 may not be stable...I actually found no go on 5....6 finished but didn't seem that fast and memtest confirmed errors....7 finished no errors...

http://home.comcast.net/~duvie23/A64-superpi-9296-2664-183-2.5337.JPG

This is where conventional wisdom goes out the door I guess...I actually tested 2.5,3,3,8 with all other settings the same...

http://home.comcast.net/~duvie23/A64-superpi-9296-2664-183-2.5338-141mod.JPG

record time 27:28.....I tried 9 and got better then 7 still but not better then 8....

Ideal setup for this test, or any test, would be BH-5 @1.5-2-2 1T ~250+mhz.. One can do this with a64 tweaker... if not run 2-2-2~250. Of course need mobo that can put out the Vdimm like DFI's or use OCZ booster. (PSA: do not use booster in EPox will fry board)

 

[Zebo]

Originally posted by: chinkgai
my mobo cpu is dfi lanparty nf3 and 3400+

here are my settings at 2750, where i got my 28:18:

275x10
9:10 divider so ram is at 250fsb @ 2-2-8-2 1t (2x512)
htt multiplier 3x

i compared my score to zebo's and he has a lower clock speed, slower cas, but at a higher htt and he still wins at 27:20

i tried 299x9 (2691mhz) with a 5/6 divider so ram is again at ~250 fsb @ 2-2-8-2 and actually got in the low 29's....basically i tried it at this setting to try to match zebo's htt and ram, but beating him in both to see if i can even get close to his time of 27:20

like i said, i failed in the second attempt with very similar settings, so i had to conclude that the 939's dual channel was the culprit in this benchmark

EDIT: one other attempt i tried was at 265x10 with ram at 1:1 timings at 2.5-3-8-3 and i got 29:09

Winny is a different chip! has optimizations NC does'nt have..not to mention dual channel.

 

[Zebo]

nm

 

[chinkgai]

Originally posted by: Zebo

Winny is a different chip! has optimizations NC does'nt have..not to mention dual channel.

bah. im kinda jealous

what do those winnie optimizations help?

 

[highlandsun]

I saved 27 seconds using Ar Soft RAMdisk, 128MB, dropped my time to 29:12. Given the size of the temp files, I'm a bit surprised it wasn't any faster. Most likely the RAMdisk software isn't optimized for high speed block memory copies.

 

[nealh]

A64@2550, 4x HTT
Neo2 Plat
ram 166 pc3200 kingston valuram 3-3-3-7 1t (212mhz)

with av and spysweeper in background

Super Pi AMD patch 32M=29min 19sec
(2550 at ram133 2.5-3-3-10 1t 170mhz pc2700 KVR got 32min 57sec non patched)

 

[mwmorph]

does superpi run in windows 3.1? i think i might be able to dig up a 486dx2 66mhz with 16mb ram to challenge the slowest record.

 

[chinkgai]

Originally posted by: nealh
A64@2550, 4x HTT
Neo2 Plat
ram 166 pc3200 kingston valuram 3-3-3-7 1t (212mhz)

with av and spysweeper in background

Super Pi AMD patch 32M=29min 19sec
(2550 at ram133 2.5-3-3-10 1t 170mhz pc2700 KVR got 32min 57sec non patched)

thats actually really fast at those settings. got screenie?

 

[John]

Loop 24 finished @ 27:51 & PI value Output @ 28:40

3200+ @ 2.6GHz

 

[GuitarDaddy]

Here's my run with the SSE2 patch.

My CPU gives out just over 2.6gz no matter what I do, don't think I can catch Zebo and Duvie at 2.65 but I'll keep tryin

28min 03sec

 

[Duvie]

Originally posted by: GuitarDaddy
Here's my run with the SSE2 patch.

My CPU gives out just over 2.6gz no matter what I do, don't think I can catch Zebo and Duvie at 2.65 but I'll keep tryin

28min 03sec

lets say I have a bit left if I need it!!!

sub 27min!!!! BUt I dont run it there...I run this speed and that is why I show it....

 

[thriemus]

I just started a run on my Athlon 64 box, Will post results shortly

.... Here we go. I used the SSE2 patched version from the link in the first post.

Loop 24 finished in 29:02, Calculated in 30:03

Before Hit OK
After Hit OK

I believe memory is my major bottleneck on this one. When I spec'd the system I decided that would rather have 2 Gig of ok RAM than 1 Gig of Fast RAM. As soon as I can get 2 x 1 Gig sticks of fast RAM running at 1T I will get them. Holding out for good DDR 500 at crazy speeds then I will transfer this RAM to my linux proxy/router/snort box.

Till then though I am still more than happy with my setup