PDA

View Full Version : Super PI benchmark for powerbooks, G4 beats Centrino!?


dodoei
Oct 30, 2005, 12:30 PM
notebookereview.com had this review that compared Super PI benchmarks among 15" PB and several intel machines. PB 15" was the fastest and faster than IBM T43 1.86G:

http://www.notebookreview.com/default.asp?newsID=2400

This makes me feel that PB is still a great value and can last a couple more years at least. It also makes me want to switch now and I'm very interested in 12" PB for its portability. I'm wondering if anyone else in this board tried Super PI benchmark for their PBs (any model), could you post the results? SuperPI for Mac can be downloaded at:

http://ocing3.free.fr/FTP/overclocking/super_pi/Kanada_lab/exec_Mac_OSX/

cheers!

contoursvt
Oct 30, 2005, 01:08 PM
Not to sound like I'm downplaying the speed of the superpi results but I'm not sure if such benchmarks really mean much. Would be a more realistic thing to test photoshop or cinebench or 3dmax or something that people actually use. I guess this is a valid test if you had to buy a machine to calculate pi :) I have a centrino laptop as well as a xeon machine and in superpi, the centrino is pretty close to the xeon in performance but it absolutely gets slaughtered in real world applications....

Demon Hunter
Oct 30, 2005, 01:23 PM
Fun. What parameter do I use?

------ Started super_pi run : Sun Oct 30 13:21:35 CST 2005
Parameter(%i) to super_pi is missing. Parameter value ?

crazzyeddie
Oct 30, 2005, 02:59 PM
I used 21, which gave me calculation of 2097152 decimal digits.

1Ghz TiBook = 3 minutes 28 seconds.

mikeyredk
Oct 30, 2005, 03:34 PM
Used 21
Rev B 20" iMac
182.403 Sec (reduced power)
108.541 Sec (high power)

:mad: why is the time to long

nospleen
Oct 30, 2005, 04:04 PM
Used 21
Automatic= 101.255
Highest= 100.796

1.9 17" imac, 512 ram.

Toreador93
Oct 30, 2005, 04:29 PM
Thought I'd throw in an AMD processor for some contrast.

SuperPi from the given link for windows; 2M digits (2097152), there was no "21" option.
AMD64 3000+ (1.8Ghz overclocked to 2.16Ghz, 512mb)

1st run: 1m 37s (97s)
2nd run: 1m 39s (99s)

Very close to the 1.9Ghz iMac


I just tried it on my roommate's 1.42Ghz mini, 512mb:

1st run: 2m 57s (177.33s)

There are a few programs running in the background, but I didn't want to close them, so his time might be faster. I'll have him try it when he wakes up.

dubbz
Oct 30, 2005, 04:38 PM
1.33GHz 12" PB RevC

Used 21

269.481 seconds (Reduced)
178.981 seconds (Automatic)
177.874 seconds (Highest)

Athlon 64 X2 4400+

88 seconds

When launching two instances of SuperPI

#1 91 seconds
#2 90 seconds

Too bad it's not SMP aware :p


It's just a silly benchmark anyway that shouldn't be taken seriously (unless you're creating a machine just for PI calculations..)

zap2
Oct 30, 2005, 04:59 PM
wow im going to see what i get!

3.314 Sec used 20

iMac g5 1.8ghz 1GB ram

Demon Hunter
Oct 30, 2005, 06:35 PM
1.67 G4 (17")

162.813s (used 21)

How did the reviewer get 99s with the same chip! :confused:

g0gie
Oct 30, 2005, 06:49 PM
of course the g4 would beat out the x86 cpu's on eaclulating pi- poewrpc computers are well known for thie floating point processing power. Thats part of the reason why its so populat in the science/mathmatics subjects . in other tasks tho.. the g4 is seriosuly outdated :X

PCMacUser
Nov 1, 2005, 07:38 AM
Thought I'd throw in an AMD processor for some contrast.

SuperPi from the given link for windows; 2M digits (2097152), there was no "21" option.
AMD64 3000+ (1.8Ghz overclocked to 2.16Ghz, 512mb)

1st run: 1m 37s (97s)
2nd run: 1m 39s (99s)

Here's some others to add to the compilation:

Dell D400 (old 1.7Ghz Pentium M, 256mb RAM) - 2m 20s
Dell D410 (new 1.86Ghz Pentium M, 256mb RAM) - 1m 41s
Dell SX280 (old 3.00Ghz Pentium 4, 1Gb RAM) - 1m 41s

Abstract
Nov 1, 2005, 09:06 AM
1.67 G4 (17")

162.813s (used 21)

How did the reviewer get 99s with the same chip! :confused:

Because this test is utterly pointless because of the type of task it's trying to do.

Don't general usage feel and real world stats mean anything? Open up a program and lets watch even a "slow" Centrino SPANK a G4.

Mikael
Nov 1, 2005, 12:45 PM
Has anyone confirmed that the program is a direct port? If it's not, it won't be a paricularly good program to compare architectures with. That said, here are the times for the computers at home:

ASUS W3N (Pentium-M Dothan 1.7GHz): 1m54s (114s)

AMD Sempron @ 2.2GHz (256kB L2): 1m31s (91s)

These CPUs seem rather close in terms of efficiency.

Demon Hunter
Nov 2, 2005, 01:43 PM
Because this test is utterly pointless because of the type of task it's trying to do.

Don't general usage feel and real world stats mean anything? Open up a program and lets watch even a "slow" Centrino SPANK a G4.

You mean you don't use your CPU to calculate pi all day? Damn hippies. :D

TheMonarch
Nov 2, 2005, 01:47 PM
How much faster is a top-of-the-line centrino VS a 1.67 G4?

I keep seeing people raving about the centrino and how it spanks a G4. Really? How badly? :confused:

Demon Hunter
Nov 2, 2005, 02:08 PM
How much faster is a top-of-the-line centrino VS a 1.67 G4?

I keep seeing people raving about the centrino and how it spanks a G4. Really? How badly? :confused:

Yeah, I don't get that either. Centrino is and always has been a piece of shitaki. Pentium M is a different story.

TheMonarch
Nov 2, 2005, 02:12 PM
Yeah, I don't get that either. Centrino is and always has been a piece of shitaki. Pentium M is a different story.


Isn't a centrino a Pentium M? :confused: :confused: :confused:

Uh-oh :o ... Shows what I know about intel... So yeah, how badly does Pentium M (Best one) spank a G4?


25% faster?
30% ?
50% ?

What? I can't be _That_ bad. Right?

dubbz
Nov 2, 2005, 02:16 PM
Isn't a centrino a Pentium M?

Centrino is just a marketing name for some specific combination of Intel tech (wirless, chipset, cpu). Pentium M is one of the parts.

So you're kind of right.

Flowbee
Nov 2, 2005, 02:24 PM
Isn't a centrino a Pentium M?

I think he may have been confusing Celeron with Centrino.

TheMonarch
Nov 2, 2005, 02:28 PM
I think he may have been confusing Celeron with Centrino.


I know what a celeron is :)


Its like the iBook of CPUs... Crippled :p

Demon Hunter
Nov 3, 2005, 02:28 AM
I think he may have been confusing Celeron with Centrino.

Ahh that's it! Damn Intel and their stupid names. :p

superbovine
Nov 3, 2005, 02:47 AM
Has anyone confirmed that the program is a direct port? If it's not, it won't be a paricularly good program to compare architectures with. That said, here are the times for the computers at home:

ASUS W3N (Pentium-M Dothan 1.7GHz): 1m54s (114s)

AMD Sempron @ 2.2GHz (256kB L2): 1m31s (91s)

These CPUs seem rather close in terms of efficiency.

I think the actually test is in question. The software itself wouldn't use all the power of thea archtecture anyways. it sort of pointless. Basically, a RISC arch will be faster in calculating Pi anyways. it isn't a surprise.

generik
Nov 3, 2005, 03:15 AM
Wow.. i never thought this day would come.

But having used a 1.6Ghz Centrino before, it seems like the PB does fair somewhat better in Photoshop! Even with 1/2 the ram!

Flowbee
Nov 3, 2005, 03:20 AM
I know what a celeron is :)

I was referring to dferrara.

gekko513
Nov 3, 2005, 03:26 AM
It's just a silly benchmark anyway that shouldn't be taken seriously (unless you're creating a machine just for PI calculations..)
To this and other comments that downplay the significance of the benchmark. I agree that it doesn't test how good or fast the processor is in general, but no benchmark does, even the ones who are made to test several aspects does not give the full answer to the performance.

This benchmark is just as significant as the other Photoshop benchmark that is going on in another thread. Both test only one particular problem.

What both benchmarks tell is how the processor performs for that problem and for similar problems. More benchmarks helps in drawing a fuller picture of the potential of the different processors. No single benchmark is sufficient just by itself.

Edit: And for some reason, overclockers are fond of using this superPi benchmark when they measure the length.... uh test the speed of their computers.

gekko513
Nov 3, 2005, 03:38 AM
2m 7s (127s) on Athlon 2700+ (2M digits)

Mikael
Nov 3, 2005, 05:46 AM
I think the actually test is in question. The software itself wouldn't use all the power of thea archtecture anyways. it sort of pointless.
Excuse me? Would you care to elaborate? Why couldn't it make full use of the architecture?

Basically, a RISC arch will be faster in calculating Pi anyways. it isn't a surprise.
You're making the classic mac enthusiast mistake, thinking that RISC CPUs are faster than x86 just because they're RISC. What you forget is that x86 CPUs haven't used CISC cores for a very long time. Every AMD and Intel CPU sold in the past ten years has utilised a RISC-like core with a CISC frontend and decoding stage. Basically, x86 CPUs provide the compact code of CISC with the fast, pipelined execution of modern RISC CPUs.

revisionA
Nov 3, 2005, 09:42 AM
The true PC problem is windows... so much so that Vista is based on a linux core and designed off the same principles then ... of OSX. (After wasting years on Longhorn... or Longsnout!!!)

Biters!!

$

csubear
Nov 3, 2005, 09:53 AM
The true PC problem is windows... so much so that Vista is based on a linux core and designed off the same principles then ... of OSX. (After wasting years on Longhorn... or Longsnout!!!)

Biters!!

$

While that nots to ture.... there is a famous quote that applies (imo) very well to vista.


Those who do not understand UNIX are doomed to poorly re-implement it.

link92
Nov 3, 2005, 08:34 PM
80.476 Sec. Dual 2.5GHz G5 (I can't help but wonder what if it was SMP aware).

Am I right in thinking it's partly to do with PPC's advantage on floating point calculations?

Melkor
Nov 3, 2005, 09:01 PM
That review made me feel a whole lot better about the powerbook i ordered :D

However, i didnt realise the battery life is so short. 2.5-3 hours seems very little to me. Thats the only bad thing though so i cant complain really.

contoursvt
Nov 3, 2005, 09:24 PM
On my dual CPU Xeon 3Ghz (which I assume SuperPi only uses one cpu) I got a score of 1:40 seconds

On my Asus Centrino 1.6Ghz laptop, I got a score of 2:12 seconds

mwpeters8182
Nov 3, 2005, 09:49 PM
I ran this on my Desktop (Sempron 3100+), and I got 2m18s, while keeping pretty much everything else running.

Not that I find pi all day with my computer, but I throw a lot of information around on a regular basis, so pure number crunching speed is important.

MP

PCMacUser
Nov 7, 2005, 04:16 AM
Here's some others to add to the compilation:

Dell D400 (old 1.7Ghz Pentium M, 256mb RAM) - 2m 20s
Dell D410 (new 1.86Ghz Pentium M, 256mb RAM) - 1m 41s
Dell SX280 (old 3.00Ghz Pentium 4, 1Gb RAM) - 1m 41s
And my new 12" iBook (using 21 as the parameter)...

3m 17s

Okay not the fastest machine this century, but hey, it's cheap!

oingoboingo
Nov 7, 2005, 05:34 AM
Barefeats.com ran a Cinebench2003 test a while ago comparing the 1.33GHz 17" PowerBook and the 1.25GHz 15" PowerBook to 1.3GHz and 1.6GHz Centrino notebooks. The Centrinos thoroughly thrashed the PowerBooks, with the 1.6GHz Centrino completing the test in 124 seconds, the 1.3GHz Centrino in 150 seconds, the 1.33GHz 17" PowerBook in 214 seconds, and the 1.25GHz 15" PowerBook in 227 seconds.

I think we should all be looking forward to new Pentium M based iBooks and PowerBooks.

http://barefeats.com/al15b.html

Sromur Cam
Nov 14, 2005, 11:36 AM
For those of us that like specs in the RAW.

ABOUT THIS MAC:
Hardware Overview:

Machine Name: PowerBook G4 17"
Machine Model: PowerBook5,1
CPU Type: PowerPC G4 (3.3)
Number Of CPUs: 1
CPU Speed: 1 GHz
L2 Cache (per CPU): 256 KB
L3 Cache (per CPU): 1 MB
Memory: 1 GB
Bus Speed: 167 MHz
Boot ROM Version: 4.6.2f1

SUPERPI TEST RUN:
Last login: Mon Nov 14 08:54:38 on ttyp2
/Applications/super_pi/pi; exit
Welcome to Darwin!
[G4PB:~] sc% /Applications/super_pi/pi; exit
Version 2.0 of the super_pi for Mac OS/X
Fortran source program was translated into C program with version 19981204 of
f2c, then generated C source program was optimized manually.
gcc-3.2.2 with compile option of "-O3 -ffast-math -finline-limit=1000" was used
for the compilation.
------ Started super_pi run : Mon Nov 14 08:59:13 PST 2005
Parameter(%i) to super_pi is missing. Parameter value ? 21
Start of PI calculation up to 2097152 decimal digits
End of initialization. Time= 2.902 Sec.
I= 1 L= 0 Time= 8.822 Sec.
I= 2 L= 0 Time= 10.109 Sec.
I= 3 L= 1 Time= 10.103 Sec.
I= 4 L= 2 Time= 10.105 Sec.
I= 5 L= 5 Time= 10.111 Sec.
I= 6 L= 10 Time= 10.096 Sec.
I= 7 L= 21 Time= 10.105 Sec.
I= 8 L= 43 Time= 10.087 Sec.
I= 9 L= 87 Time= 10.124 Sec.
I=10 L= 174 Time= 10.097 Sec.
I=11 L= 349 Time= 10.091 Sec.
I=12 L= 698 Time= 10.094 Sec.
I=13 L= 1396 Time= 10.096 Sec.
I=14 L= 2794 Time= 10.064 Sec.
I=15 L= 5588 Time= 10.055 Sec.
I=16 L= 11176 Time= 10.038 Sec.
I=17 L= 22353 Time= 10.004 Sec.
I=18 L= 44707 Time= 9.912 Sec.
I=19 L= 89415 Time= 9.663 Sec.
I=20 L= 178831 Time= 9.045 Sec.
End of main loop
End of calculation. Time= 209.377 Sec.
End of data output. Time= 0.839 Sec.
Total calculation(I/O) time= 210.216( 24.587) Sec.
------ Ended super_pi run : Mon Nov 14 09:03:07 PST 2005
logout
[Process completed]