fun benchmark

tuxy

hes a fun little benchmark i just found
its called super pi
http://homepage.eircom.net/~spellmandsl/superpi.rar

calculate pi to 1 million places
how long does it take you?

p4(533) 2.4@ 3.17
time = 54 seconds
not too bad i think but i know some people around here will smash that time

KdjaCL

1m 2s

2400Xp@2366

kdjac

Adeptus Titanicus

to 1M places: 40 seconds
As per sig, P3.0 @ 3.6

Ran it twice cos I though I had made a mistake!

Col_Loki

AMD Xp2500+ @ 2480mhz -> 43sec's

Am waitin to see how the AMD64's do in this !!

Mutant_Fruit

1:05 2500@stock speed.

neokenzo

39 seconds.

I think the current fastest superfi is 29 seconds.

Dman_15

51 secs to 1 m places

2700 xp, stock speeds (2.16)

Mutant_Fruit

1M places:

0:40 using a P3.0 @ 3.6
1:05 using a 2500@stock speed
0:43 using a 2500@ 2480mhz

Makes for interesting reading. For these type of apps AMD's seem to be pretty damn good. Despite the pentium clocking at twice the speed of the 2500@stock, it only finishes 30% faster. And a XP2500+@2480mhz (still 1.2ghz slower) is just as fast.

The AMD64 would have a really interesting result judging by this.

hostyle

XP1800+ took 54 secs ...

<edit>
Actually I re-ran it a few times again, cause that just sounds wrong ...

1m 21s
1m 17s
1m 14s

weird ...
</edit>

Adeptus Titanicus

Mutant, just in case neo's slipped by, as his is an AMD 64...

0.39 using a Athlon 64 3200 @ 2.25Ghz

compares nicely to the rest allright given the clock speed.

tuxy

look at neokenzo's sig ( you cant realy see it if yuor using the cloud style) hes using an athlon 64 and gets 39 seconds

but i think athlon's were always best at raw calculations like this
when it comes to aplications its differnt amd better at some intel better at others

still its an interesting test

CombatCow

54s to 1M places - AMD xp2800+ (default )

:dunno:
CombatCow

CivilServant

45 secs on Thoroughbred @ 2.3Ghz

RuggieBear

I ran the test on the computer in the Guest House i'm working in just now.....it took a mind blowing 13min 7sec to do 1m places.:D

It's a compaq Presario with (i think) a 733 celeron with 128Mb. Been trying to convince the owner to get a new computer but he refuses. This thing can barely do email...and surfing the web is a bit of a strugle too (even the Boards.ie avatars cause it hardship)

p

Gerry

Its 1m 35 seconds on this machine, p3 tualatin 1400mhz, 512mb ram.

Woden

1M places, 19 iterations, i just took the number after the last loop

47s on a 3200+ xp

gerry was that the end of the line for p3's that 1.4ghz one? i would be interested to see an equivalent clocked p4 do the test and see the difference

edit for better time 49 down to 47

bizmark

I got 1m 2 seconds on a 2600xp

i had a load of progs open at the time though so ill run the test again later

Gerry

Yeah, its the last revision of the p3. Has very slightly better performance on a clock for clock basis than the coppermine when the caches are the same ( the tualatin celeron has 256k l2), but the tualatin p3-s has 512k , so its basically the same speed as an equivalent clockspeed athlon xp, faster at some things, slower at others ( i.e. heavy fpu stuff ). Very cool running chip, bout 30 watts of heat. Hopefully will be overclocking this chip shortly, 1400mhz is the stock speed.

tuxy

i guess calculating pi to 32 m places would be a good stress test
i think ill try that tonight

L5

1min 02 seconds.
p4c 2.4ghz 512mb

Matt Simis

Originally posted by Adeptus Titanicus
Mutant, just in case neo's slipped by, as his is an AMD 64...

0.39 using a Athlon 64 3200 @ 2.25Ghz

compares nicely to the rest allright given the clock speed.

36 Seconds on a Athlon 64 3000 (512k cache) at 2660MHz.

Compared to the 3200 at 2.25, thats not a great score. Could be due to:

- Half the cache on the 3000+
- Program is very old, possibly doesnt work very well with new CPUs.
- Some other variables (RAM amount and speed?)



Matt

joePC

1 million = 50 seconds

2500+ @ 2.1ghz

Woden

Originally posted by Gerry
Yeah, its the last revision of the p3. Has very slightly better performance on a clock for clock basis than the coppermine when the caches are the same ( the tualatin celeron has 256k l2), but the tualatin p3-s has 512k , so its basically the same speed as an equivalent clockspeed athlon xp, faster at some things, slower at others ( i.e. heavy fpu stuff ). Very cool running chip, bout 30 watts of heat. Hopefully will be overclocking this chip shortly, 1400mhz is the stock speed.

cheers, and where they still on a 100mhz fsb with that or was it 133*10.5?

neokenzo

I took it 3 times and the best time was 39 sec and the slowest is 40 seconds. Mind you the one of the fastest Super Fi I've seen is done in 29 seconds P4 5Ghz (315x16). It depends highly on you fsb. The higher it is the lower the score.

Gerry

Nah, on the p3-s, they use a 133fsb. Thats why I can't overclock it, this board was made in 98 I think, max fsb is 133. Getting a be6-2 which will hopefully do the trick shortly.

neokenzo

Originally posted by Matt Simis
36 Seconds on a Athlon 64 3000 (512k cache) at 2660MHz.

Compared to the 3200 at 2.25, thats not a great score. Could be due to:

- Half the cache on the 3000+
- Program is very old, possibly doesnt work very well with new CPUs.
- Some other variables (RAM amount and speed?)



Matt

I dont think its an old program as I've seen ppl running it on very low scores. Even P4 4Ghz are getting 33-35 seconds. You wont see a much difference with these scores IMO.

Woden

sorry for the off topic in general. so would this mean that the that p3 would out perform an equivalently clocked p4 as i think they started about 1.4ghz and they where 100mhz chips?

could there be an analogy also if the p3 out performed the p4, like say the way the northwood outperforms the prescott at lower clock speeds and is this all due to pipelines?

Stephen

P4m 1.7GHz/512MB PC133 on battery power --> running in low power mode (~ 1.0GHz). Time taken: 04:22
P4m 1.7GHz/512MB PC133 on battery power --> cpu set to run at full speed. Time taken: 02:03
XP2800+/512MB Dual DDR PC3200 (running only at 333 though) at stock speeds. Time Taken: 00:53
Ageing P4 1.6GHz/128MB PC133. Time taken: 02:13

bus77

p4a 2.4 256mb
54 secs

Originally posted by L5
1min 02 seconds.
p4c 2.4ghz 512mb

You must taken that off the readout page that comes up during the test L5. Take your figure off the main program page instead.

joePC

P3 500mhz 192 Ram

4min 22sec

1 million PI

ionapaul

AMD Athlon 64 3000+
43 secs

Adeptus Titanicus

Originally posted by Stephen
P4m 1.7GHz/512MB PC133 on battery power --> running in low power mode (~ 1.0GHz). Time taken: 04:22
P4m 1.7GHz/512MB PC133 on battery power --> cpu set to run at full speed. Time taken: 02:03

To continue the mobile bit:
Centrino 1.7GHz/1024MB PC133 Full power--> Time taken: 01:01
Centrino 1.7GHz/1024MB PC133 @ Stepped down to 600MHz--> Time taken: 01:03!

And for historical interest:
Athlon K7 800MHz/256Mb PC133 --> Time taken: 03:22. The poor auld thing

Stephen

Real evidence that clock speed means nothing, eh?

tuxy

Originally posted by Stephen
Real evidence that clock speed means nothing, eh?

well it means something
but the amount of work it can do per cycle also means something

Adeptus Titanicus

Got curious about the difference the clcok speed would make on my desktop:

P4 3.0 @ 3.0 / 1024 Mb FSB @ 200 --> Time taken 00:49
P4 3.0 @ 3.6 / 1024 Mb FSB @ 240 --> Time taken 00:40

On the Centrino, the FSB stayed the same (100), the stepping just dropped the muliplier from 17 to 6 , so that probably explains the similar scores.

Matt Simis

Originally posted by neokenzo
I dont think its an old program as I've seen ppl running it on very low scores. Even P4 4Ghz are getting 33-35 seconds. You wont see a much difference with these scores IMO.

I know new CPUs can and do run this program and achieve low scores, but the fact remains (afaik) its a really old program.

This is in the (old Windows 3.11 style) help file:

Run environment:
Windows NT, Windows 95.
If you uses Windows 3.1, Win32s is needed for the program run.


And the copyright in "About" is from 1995...



Matt

El_MUERkO

AMD Athlon XP 2500 (Barton) running as a 3200.

47 seconds, I'm running some progs right now, will try later with no extra progs running

VeVeX

I got 48s with 3Ghz 1Gb pc3200

Which is only slightly faster then both of the AMD machines (2500 & 2800) we have.

There are 2 factors which would lead me to belive that this is not an air tight benchmark of current hardware.

While runing the test my P4 was only running at 52% usage. This maybe because of the hyper threading , I dont know .Surely to test a CPU's potential it would need to be pushing more then 52%.

The second is- our 2500 actualy scored better then our 2800. The 2500 only has 256mb of ram as compared to 512 in the 2800. They are running at stock speeds and are both on NF7's with almost identical everything else.

Sir Random

Pi to 1M: 36s

Specs as in sig, except I was using Mushkin Lvl2 @ 2,3,2,5

Super Pi is an excellent guide to your PCs raw power. A fast cpu and tight ram give the best results (as in real world performance)

tuxy

Originally posted by KeithD
I
While runing the test my P4 was only running at 52% usage. This maybe because of the hyper threading , I dont know .Surely to test a CPU's potential it would need to be pushing more then 52%.

thats weird my p4 (non ht) hits 100% as soon as i start the test and stays there untill its finished

Mutant_Fruit

I think this program would give better results. It's pretty recent, and so should be much better optimized than that (seemingly) 10 year old program.

I'll be back with my scores in the above program in about 5 mins.

You'd want to be computing at least 10,000,000 places with this, as its a LOT faster than the other program.

Basically launch it up, press 0 to use the Chudnosky method, 0 again for standard mode, then type in 10000000 for 10 million places, and lastly type in 1024 as the FFT size (fastest mode).

My time is 57.27 seconds on a XP2500+@stock speeds for 10million places.

EDIT: As you can see, its 10x faster than the previous program listed.

rabbitinlights

P4 2.4 @ 3.2
1GB Corsiar Matched pair
Abit IC-7

Super Pi:

1m = 48 Sec

Pifast

10million = 45.7 Sec

Not bad.



Sean

neokenzo

I got about 33 seconds in Pifast.
Do you pick yes or no when it asked on Compressed output?

tuxy

i got 51.14 seconds in pifast
pi to 10million places
wow that program is fast

p4(533) 2.4 @ 3.17

Dman_15

54s to 1M places - AMD xp2800+ (combat cow)
XP2800+/512MB Dual DDR PC3200 (running only at 333 though) at stock speeds. Time Taken: 00:53 (stephen)

This confused me as my xp2700 did it in 51 seconds.
Why is this? surely the 2800 should be faster

neokenzo

Probably they are running other programs when they did the test?

Stephen

45.83 seconds for 10million places on pifast.
Same hardware as before: xp2800+ at stock speeds/512MB PC3200 running at 333mhz.

don't have the laptop booted to try it on that atm.

The only programs I am running in the background are BitTorrent and some light app that monitors my DSL router. And of course a few of the usual windows services.

Sir Random

Jeez, you're just confusing things with different Pi software.
O'clockers have used Super_Pi as a benchmark for years. Not because it's fast, but because it's simple pure code that doesn't favour any hardware.
There are Super_Pi threads on most O'clocking forums already, it's a well respected test.

neokenzo

I agree. Even on extreme overclocking sites they still do the Super Pi test eventhough it is an old program.

Sir Random

Here's the OcUK Super_Pi challenge:
http://forums.overclockers.co.uk/showthread.php?s=&threadid=17197191

Matt Simis,
You should post your time there

hostyle

Are the people with AMD64 CPUs running a 64 bit OS (XP 64, seeing as this appears to be windows only software) ? Wouldn't the program need to be written specifically for a 64 bit OS to get a _true_ benchmark? Or am I talking out of my ass again? AMD64 in 32 bit mode is merely playing with your 64 bit CPU, no?