22-May-2024 05:49 GMT.
UNDER CONSTRUCTION
Anonymous, there are 80 items in your selection [1 - 50] [51 - 80]
[Web] AmigaOS4 Petunia benchmarksANN.lu
Posted on 18-Aug-2003 11:11 GMT by Peter Gordon80 comments
View flat
View list
The petunia website has been updated with benchmarks from a recent version. Also, it is claimed that "This version is already converted to AmigaOS4, and running on the native PowerPC system".

Petunia homepage
AmigaOS4 Petunia benchmarks : Comment 1 of 80ANN.lu
Posted by Hagge on 18-Aug-2003 09:25 GMT
so 180MHz 604e is 1-1.5x as fast as a 50MHz 060.

Anyone care to run http://amigos.amiga.hu/rachy/PetuniaTests.lha on their MorphOS 1.4 setup and give us some numbers?
AmigaOS4 Petunia benchmarks : Comment 2 of 80ANN.lu
Posted by Amon_Re on 18-Aug-2003 09:32 GMT
In reply to Comment 1 (Hagge):
Also running on the classic Amiga's?

Or the newer Pegasos systems? If the latter, i'm sure there are already published results out there.

However, i think it's a silly test aslong as the final product(s) aren't released yet.

Cheers
AmigaOS4 Petunia benchmarks : Comment 3 of 80ANN.lu
Posted by Neko on 18-Aug-2003 09:56 GMT
What's the point? Running 1.3

julia_fpu crashes HARD by accessing a strange memory location.

I couldn't get Mandel to output any times or frames values? It draws fast
enough, though, but what's the story with this 500x200 window? Are we living
in the dark ages of 2:1 aspect screenmodes? :)

c2ptest also outputs nothing in terms of benchmarks.

Demoeffect doesn't draw anything (contrary to "a bubbling picture effect (pixel displacement) on a 320x240 CGX screen from Stephen Fellner") and gives
a HIGHLY different result depending on what the phase of the moon is that
day. Maximum was 235.73, minimum I got was 148.94, and sometimes it would get
192.75 +- 2.0 fps more often than not.

Whatever happens, they only draw for a second anyway, which is a stupid way
to test the speed. Why not run it for 4 seconds and take the 2nd and 3rd second
as an average of the real speed? Then your graphics card doesn't factor into
it and it's an ultimate test of the EMULATION, and not how fast your video card
can fiddle with it's RAMDAC.

Atop that, they are all more tests of how fast WriteChunkyPixels is, and not
anything else. It would have been better to use, perhaps, WritePixelArray()
which would stress memory out during frames, stress the emulation out for the
calculations, and finally and simply do ONE OS call to display, instead of
calling redundant functions.

One of the easiest optimisations to make in a program is not to call functions
more than you have to. The calling conventions of most operating systems take
time!

I still say something like RC5-64 would be a better test of the emulation, or
something that is entirely CPU-bound, and then we can compare it with a more
complex test like graphics or sound or whatever.

Neko
AmigaOS4 Petunia benchmarks : Comment 4 of 80ANN.lu
Posted by Ben Hermans/Hyperion on 18-Aug-2003 10:20 GMT
In reply to Comment 3 (Neko):
The problem is that you need to watch out not to pollute the benchmarks by measuring the performance of the underlying OS components.

I do agree however that comparing a 600 Mhz machine equipped with a infinitely faster memory architecture and on die L2 cache with the crippled L2 less Cyberstorm PPC is completely pointless.

Guess what? The 600 Mhz will be a hell of a lot faster.
AmigaOS4 Petunia benchmarks : Comment 5 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 10:21 GMT
"converted, and running"

So I take it's not integrated?

What's taking so long?
AmigaOS4 Petunia benchmarks : Comment 6 of 80ANN.lu
Posted by tarbos on 18-Aug-2003 10:26 GMT
In reply to Comment 3 (Neko):
>Maximum was 235.73, minimum I got was 148.94 Maximum here was about 248fps and minimum just below 18fps. :-o
AmigaOS4 Petunia benchmarks : Comment 7 of 80ANN.lu
Posted by José on 18-Aug-2003 10:33 GMT
"Changing to 64bit PowerPC processors requires a bit rework of some parts, but it is not a huge problem."

Nice!!
AmigaOS4 Petunia benchmarks : Comment 8 of 80ANN.lu
Posted by tarbos on 18-Aug-2003 10:36 GMT
In reply to Comment 4 (Ben Hermans/Hyperion):
>I do agree however that comparing a 600 Mhz machine equipped with a infinitely>faster memory architecture and on die L2 cache with the crippled L2 less>Cyberstorm PPC is completely pointless. Too bad the Escena Brainstormer didn't work out for you, eh?As for the faster memory architecture...the 200MHz 604e CSPPC beats Pegasos inSysSpeed PPC Writew and Copyw test.Now I'd like to ask again - will we see some AmigaOne G3 memory write numbersfrom lmbench running under Linux or will you wait for a better Articia beforemaking them public?
AmigaOS4 Petunia benchmarks : Comment 9 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 10:54 GMT
In reply to Comment 1 (Hagge):
Tested on Pegasos 1 600MHz and MorphOS 1.4 JIT:

mandel 0.70 secs
julia fpu 2.87 secs
c2ptest 0.74 sec
demoeffect 260.06 fps
Voxelspace the fps reading is way too wobbly to read proper values, it shows 2xx and 1xx fps.

So Petunia is running on AmigaONE now? Where are the results?

G4@800 should easily beat Pegasos with just 600MHz G3... Right?
AmigaOS4 Petunia benchmarks : Comment 10 of 80ANN.lu
Posted by Phill on 18-Aug-2003 10:54 GMT
In reply to Comment 5 (Anonymous):
> "converted, and running"
> So I take it's not integrated?

If he says it has been converted to work with AmigaOS 4 & is running, then what exactly do you want before you think it's "integrated".

> What's taking so long?

Emulation can be a bitch.

Phill
AmigaOS4 Petunia benchmarks : Comment 11 of 80ANN.lu
Posted by tarbos on 18-Aug-2003 11:17 GMT
In reply to Comment 9 (Anonymous):
>Voxelspace the fps reading is way too wobbly to read proper values, it shows 2xx and 1xx fps. Between 23x and 28x fps here in 320x240.
AmigaOS4 Petunia benchmarks : Comment 12 of 80ANN.lu
Posted by Alfred Schwarz on 18-Aug-2003 11:21 GMT
In reply to Comment 11 (tarbos):
> Between 23x and 28x fps here in 320x240.

On CSPPC? (*that* would be interesting)

Ciao, Alfred
AmigaOS4 Petunia benchmarks : Comment 13 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 11:30 GMT
In reply to Comment 9 (Anonymous):
No. 200mhz is not a big difference. G3 =G4 in speed without Altivec(more or less). Petunia is assembler only and don't has Altivec optimizations.

Hell you probably know that anyway...
AmigaOS4 Petunia benchmarks : Comment 14 of 80ANN.lu
Posted by Kjetil on 18-Aug-2003 11:31 GMT
In reply to Comment 9 (Anonymous):
/* So Petunia is running on AmigaONE now? Where are the results? */

No OS4.0 is not running on AmigaOne is only works on CyberStorm for now, the Amiga One version is dependent on new SNAP graphic drivers and IO port devices that needs to be written, than main goal is get it working on Amiga4000 as end user product, then the AmigaOne and after that, what ever that is qualified to be used.

So fare you have not made the wrong decision when you decided to go for MorphOS, Guess some where just before or right just after new year 2004.
AmigaOS4 Petunia benchmarks : Comment 15 of 80ANN.lu
Posted by David Scheibler on 18-Aug-2003 11:32 GMT
In reply to Comment 12 (Alfred Schwarz):
There is an old version of the benchmarks results available using an old
version of Trance running on BlizzPPC 175MHz. Maybe someone will provide newer
values.
AmigaOS4 Petunia benchmarks : Comment 16 of 80ANN.lu
Posted by Alfred Schwarz on 18-Aug-2003 11:34 GMT
In reply to Comment 13 (Anonymous):
> 200mhz is not a big difference.

Of course not... 600MHz to 800MHz are just 33% more, really not much...

Ciao, Alfred
AmigaOS4 Petunia benchmarks : Comment 17 of 80ANN.lu
Posted by tarbos on 18-Aug-2003 11:42 GMT
And just for fun - SysSpeed showed 3399 MIPS once (and ~3200 MIPS regularly) withTrance emulation while it does only 300 MFLOPS. :)
AmigaOS4 Petunia benchmarks : Comment 18 of 80ANN.lu
Posted by Ben Hermans/Hyperion on 18-Aug-2003 12:02 GMT
In reply to Comment 8 (tarbos):
>Now I'd like to ask again - will we see some AmigaOne G3 memory write numbers
>from lmbench running under Linux or will you wait for a better Articia before
>making them public?

Considering the fact that there are hundreds of people out there using their AmigaOne's every day, you can hardly consider this info to be confidential right?

I'm sure one of these people can help you with that.

It seems you are asking the wrong people.

I think we once measured between 500 megabytes and 1 gigabytes per second depending on whether it was writing or reading.
AmigaOS4 Petunia benchmarks : Comment 19 of 80ANN.lu
Posted by BrianK on 18-Aug-2003 12:55 GMT
In reply to Comment 9 (Anonymous):
G4@800 should easily beat Pegasos with just 600MHz G3... Right?
--------

It 'depends'....
1) 800 is faster then 600 but not by much.
2) G3=G4 are pretty much, yes I know they're not, equivalent in speed. The G4 has the altivec...
3) AmigaOS4 vs PegasOS are two different OSes. They are doing things differently. Thus, if AmigaOS4 is less efficent then PegaOS then it may not be much improved.
4) Hardware is a bit different, which board handles tasks more efficently and how does that play a role?
5) Altivec? Will AmigaOS4.x implement Altivec calls and how do these calls play into the emulation. Appears for right now they really won't be used...

So, there are some other factors to consider. I'm comfortable with believing the emulation will be as fast as the 600Mhz G3 but at 800Mhz will probably be faster.
AmigaOS4 Petunia benchmarks : Comment 20 of 80ANN.lu
Posted by Amon_Re on 18-Aug-2003 12:56 GMT
In reply to Comment 8 (tarbos):
Ah bugger off ;)

Why don't you ask that question to someone you know who has an A1, or at the A1 list(s)?

Bitching to Ben about hardware he doesn't make... how sad

Cheers
AmigaOS4 Petunia benchmarks : Comment 21 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 12:59 GMT
In reply to Comment 18 (Ben Hermans/Hyperion):
Um, no. I think you're mistaken about that Ben.
AmigaOS4 Petunia benchmarks : Comment 22 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 13:08 GMT
In reply to Comment 18 (Ben Hermans/Hyperion):
>Considering the fact that there are hundreds of people out there using their>AmigaOne's every day, you can hardly consider this info to be confidential right? Hundreds of AmigaOne owners...reading ANN...noone replys - why? :-> >It seems you are asking the wrong people. Well... >I think we once measured between 500 megabytes and 1 gigabytes per second>depending on whether it was writing or reading. It seems I was definitely asking the wrong people - if I wanted to hear storiesI would have asked my grandma. :o)Come on you hundreds of AmigaOne/Linux users, do me the favour and post yourG3 memory writes results in this thread. Olegil, what about you? I asked inApril afair, is it really a wasted time effort to provide some hard numbers?
AmigaOS4 Petunia benchmarks : Comment 23 of 80ANN.lu
Posted by itix on 18-Aug-2003 13:10 GMT
In reply to Comment 1 (Hagge):
I benchmarked Trance using Petunia tests long time ago and results were pretty good. But I had to remove stats from the web because JIT was not finished yet. Maybe I could publish results now but only if I get permission from MorphOS team.
AmigaOS4 Petunia benchmarks : Comment 24 of 80ANN.lu
Posted by itix on 18-Aug-2003 13:14 GMT
In reply to Comment 3 (Neko):
You must be very silly. I have benchmarked Petunia tests (not julia_fpu) on my BPPC running MorphOS. Some 68k utilities were needed but it is not too hard if you use your brains.
AmigaOS4 Petunia benchmarks : Comment 25 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 13:14 GMT
In reply to Comment 16 (Alfred Schwarz):
>Of course not... 600MHz to 800MHz are just 33% more, really not much...

But it's a G4 vs a G3. I've heard that, the G4's altivec aside, the G3 is in fact the better CPU (at the same clock). Is that true?
AmigaOS4 Petunia benchmarks : Comment 26 of 80ANN.lu
Posted by itix on 18-Aug-2003 13:15 GMT
In reply to Comment 3 (Neko):
Oh by the way, I must still admit those tests are not very good. They run only very short time.
AmigaOS4 Petunia benchmarks : Comment 27 of 80ANN.lu
Posted by tarbos on 18-Aug-2003 13:27 GMT
In reply to Comment 25 (Anonymous):
At least it has twice the Bogomips of G4. ;)
AmigaOS4 Petunia benchmarks : Comment 28 of 80ANN.lu
Posted by Amon_Re on 18-Aug-2003 13:36 GMT
In reply to Comment 22 (Anonymous):
Aren't there any results of memtests on that reviewsite?

Check the archives, or ask someone/somewhere else. This isn't Ben's role.

Cheers
AmigaOS4 Petunia benchmarks : Comment 29 of 80ANN.lu
Posted by BrianK on 18-Aug-2003 13:41 GMT
In reply to Comment 25 (Anonymous):
I believe the G3 can execute more instructions per tick them the G4. So, a G3 600Mhz vs G4 600Mhz the G3 might be a bit quicker on non-altivec items.
AmigaOS4 Petunia benchmarks : Comment 30 of 80ANN.lu
Posted by BrianK on 18-Aug-2003 14:11 GMT
In reply to Comment 29 (BrianK):
I'll correct myself... Well at least according to
http://ece.gmu.edu/courses/ECE543/presentationsF01/hartong.pdf

500Mhz G4 Cube: SpecINT:23.7, SpecFP:22.7
500Mhz G3 Ibook: SpecINT:21.2, SpecFP:12.3
AmigaOS4 Petunia benchmarks : Comment 31 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 14:21 GMT
In reply to Comment 12 (Alfred Schwarz):
So sorry...it was actually the PowerPC routine on G3-600, d'oh! 8-|Emulated it's between 57 and 2xx (wrong timing?) with the average probably just below 70fps.
AmigaOS4 Petunia benchmarks : Comment 32 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 14:22 GMT
In reply to Comment 30 (BrianK):
Comparable cache and bus speeds?
AmigaOS4 Petunia benchmarks : Comment 33 of 80ANN.lu
Posted by Don Cox on 18-Aug-2003 14:38 GMT
In reply to Comment 14 (Kjetil):
"So fare you have not made the wrong decision when you decided to go for MorphOS, Guess some where just before or right just after new year 2004."

So far, I made the right decision when I decided to go for Amithlon. I have had a nice fast, stable Amiga with plenty of RAM for nearly two years.

Looks like another 12 months before there will be something to compete with it.
AmigaOS4 Petunia benchmarks : Comment 34 of 80ANN.lu
Posted by tarbos on 18-Aug-2003 14:50 GMT
In reply to Comment 31 (Anonymous):
Now this is kind of funny, I tried the cgfx voxeldemo before (it said TURBOGFXinside the demo) and now I try the P96 one (says CYBERGFX inside) and after afew seconds the JIT kicks in and accelerates it from fifty-odd to ~250fps, whichis faster than the PPC settings with ~200fps. :D
AmigaOS4 Petunia benchmarks : Comment 35 of 80ANN.lu
Posted by Piru on 18-Aug-2003 15:07 GMT
In reply to Comment 1 (Hagge):
MorphOS 1.4 JIT on 603/240 system is several times faster than the Petunia on 604/180.

c2ptest is over 3.5 times faster, mandel over 2.5, demoeffect over 1.5 times faster.

Remember that CSPPC has twice as fast memory bus, too. The Petunia page describes the A4000 CSPPC test machine to run OS 4.0, so there is no more WarpOS overhead in the results(?).

So does this mean Trance is much faster, or did I miss something?
AmigaOS4 Petunia benchmarks : Comment 36 of 80ANN.lu
Posted by Kelly Samel on 18-Aug-2003 15:20 GMT
In reply to Don Cox: I actually own an Amithlon system (1Ghz) and a
Pegasos and the Pegasos is clearly a nicer system in almost every
respect. Amithlon is an impressive Amiga OS 3.9 based system but
MorphOS v1.4 just improves on the whole concept. It is more stable,
runs more software, has a native PPC OS and apps, plus all existing
Amiga 68k/PPC apps. (including fast JIT for 68k). It looks nicer and
performs generally smoother. :)
AmigaOS4 Petunia benchmarks : Comment 37 of 80ANN.lu
Posted by Nate Downes on 18-Aug-2003 16:18 GMT
In reply to Comment 14 (Kjetil):
That is not what was asked. He asked if Petunia were running, not AOS4.

ExecSG already runs on the A1, so it is more than possible that petunia is also running on there.
AmigaOS4 Petunia benchmarks : Comment 38 of 80ANN.lu
Posted by Nate Downes on 18-Aug-2003 16:19 GMT
In reply to Comment 19 (BrianK):
What is this OS you speak of, PegasOS? I run MorphOS on my Pegasos motherboard, personally.
AmigaOS4 Petunia benchmarks : Comment 39 of 80ANN.lu
Posted by Don Cox on 18-Aug-2003 16:38 GMT
In reply to Comment 36 (Kelly Samel):
" It is more stable,
runs more software,"

My Amithlon setup is somewhat faster than yours, which may make a difference. It is extremely smooth and stable.

I am interested to hear that MorphOS runs more software. Can you tell me which programs run on MorphOS but not on Amithlon? Could this be because CGFX is better than P96? Does MorphOS have "fake native" mode for programs that will not be mode promoted?
AmigaOS4 Petunia benchmarks : Comment 40 of 80ANN.lu
Posted by Nicolas Sallin on 18-Aug-2003 16:52 GMT
In reply to Comment 39 (Don Cox):
MorphOS's Intuition always open a screen, even if modeid is invalid.
AmigaOS4 Petunia benchmarks : Comment 41 of 80ANN.lu
Posted by Remco Komduur on 18-Aug-2003 16:55 GMT
In reply to Comment 22 (Anonymous):
That's because they don't come here! I even barely come here with all the noise.

Instead of comming here, we can do something usefull instead of reading mostly a huge amount of unintelligent crap.
AmigaOS4 Petunia benchmarks : Comment 42 of 80ANN.lu
Posted by Lando on 18-Aug-2003 16:57 GMT
In reply to Comment 39 (Don Cox):
>I am interested to hear that MorphOS runs more software

It also runs WarpUP, PowerUP programs, whereas Amithlon is restricted to 68k.
AmigaOS4 Petunia benchmarks : Comment 43 of 80ANN.lu
Posted by ehaines on 18-Aug-2003 17:34 GMT
In reply to Comment 25 (Anonymous):
> But it's a G4 vs a G3. I've heard that, the G4's altivec aside, the G3 is in
> fact the better CPU (at the same clock). Is that true?

Well, there's not really a "G3" CPU (or a "G4" for that matter). For "G3",
there's the 750, 750cxe, 750FX, 750GX (maybe not out yet?), possibly some more.
It's true that I'd rather have a 750GX than a 7440, Altivec or not....
AmigaOS4 Petunia benchmarks : Comment 44 of 80ANN.lu
Posted by ehaines on 18-Aug-2003 17:37 GMT
In reply to Comment 22 (Anonymous):
> Come on you hundreds of AmigaOne/Linux users, do me the favour and post your
> G3 memory writes results in this thread. Olegil, what about you? I asked in
> April afair, is it really a wasted time effort to provide some hard numbers?

I offered before, more than once, as long as someone sends me a binary. I'm
not going to waste time fiddling about with compiling stuff for this. But
nobody did, so I guess it's not that important.
AmigaOS4 Petunia benchmarks : Comment 45 of 80ANN.lu
Posted by Kjetil on 18-Aug-2003 17:59 GMT
In reply to Comment 37 (Nate Downes):
You intuition.library and gfx.library layers.library to get a workbench up and running, unless the benchmark only do number crunching and used printf() to display the data, you can not dump the results in side gfx window etc, so it's extremely unlike unless you rewritten benchmark to only display numbers in text format,

and the result will be bit hi when compared with a normal system due to system resources missing that will normally take up some time frames now and then, the results will not give in good indication to performance in normal system, how ever it might give a better indication the looking at MorphOS bench mark results and and estimate performance on AmigaOne by this results.
AmigaOS4 Petunia benchmarks : Comment 46 of 80ANN.lu
Posted by Lando on 18-Aug-2003 18:28 GMT
In reply to Comment 45 (Kjetil):
Think you could say all that again in English? :)
AmigaOS4 Petunia benchmarks : Comment 47 of 80ANN.lu
Posted by corpse on 18-Aug-2003 18:30 GMT
In reply to Comment 35 (Piru):
"So does this mean Trance is much faster, or did I miss something?"

too much #debian for you ;).

-- 9/10 trained monkeys say they'd take a tcp/ip stack over 24bit icons anyday ;)
AmigaOS4 Petunia benchmarks : Comment 48 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 18:32 GMT
In reply to Comment 32 (Anonymous):
Comparable Cache and bus speeds?

See the posting you can find that info at the included URL...http://ece.gmu.edu/courses/ECE543/presentationsF01/hartong.pdf . There's lots of info of how they tested along with the systems tested.

Also, here's the info on the systems.
500Mhz G4 Apple Cube:
500Mhz G3 Apple Ibook:
Bus Speeds and cache can be found at Apple's Website. Go under Support and look for hardware.
AmigaOS4 Petunia benchmarks : Comment 49 of 80ANN.lu
Posted by Anonymous on 18-Aug-2003 18:39 GMT
In reply to Comment 35 (Piru):
>The Petunia page describes the A4000 CSPPC test machine to run OS 4.0, so there >is no more WarpOS overhead in the results(?).

No WOS overhead but probably interpretive 68k emulation for at least graphics+p96 + the (much smaller ofcourse) system emulation overhead (ppc cpu traps etc). The page states that the emulator runs under 4.0 as an application, but is not integrated in exec.

But the 604/180 results are probably still 3.9, as I doubt they will be exactly the same under 4.0 ;)
AmigaOS4 Petunia benchmarks : Comment 50 of 80ANN.lu
Posted by Piru on 18-Aug-2003 18:59 GMT
In reply to Comment 49 (Anonymous):
> No WOS overhead but probably interpretive 68k emulation for at least
> graphics+p96 + the (much smaller ofcourse) system emulation overhead (ppc cpu traps etc).

Mandel has total 206 OS calls, of which 200 are graphics calls. There is one call per 500 calculated mandelbrot pixels.

c2ptest has 6 OS calls, all in setup/cleanup. demoeffect has 22 OS calls, all in setup/cleanup.

Somehow I doubt the OS calls are slowing things down here.
Anonymous, there are 80 items in your selection [1 - 50] [51 - 80]
Back to Top