@Niolator

Quote:

Niolator wrote: Very interesting results. So a SAM at 800 MHz is almost as fast as a G3 at 666 MHz? It would be interesting to see which scores A1XE G4 and Pegasos produce.

Does low latency RAM help SAM? Can you tweak the RAM settings on SAM?

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Leo

O yeha...I know that!
On an AMD AthlonXP 3000+(about 5 years old) the results was similar...
I think that the real bottelneck here was the hd speed(I wonder how it takes on a test made on a ram disk if possible...)
If you can do the ffmpeg test too (I never tryed on the pc that) we can compare that better :)

@Chris
I can understand your point of view(and a bit the bernd_afa too)...however plz try to shift the discussion in private...
Different points of view was natural...no avery ppl can have same views of the others...how a borign world will be if yes...

_________________
Simone"Tuxedo"Monsignori, Perugia, ITALY.

@Leo

8 seconds on OSX or Aros?

@thread

Need to test DVD to SFS0 copying with the 200M file.
For example when I copy AF79 CD to SFS0, the speed is only 1M per second.
(DMA is on)

Perhaps it's faster with larger files.

Last edited by KimmoK on 17-Nov-2009 at 02:22 PM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@Tuxedo

Here are my results for WinUAE box of my brother:

Computer: Core 2 Quad Q6600-2400, 8 GB RAM, GForce 8600GT 256 MB
WinUAE 1.5.0 (68040/FPU, JIT, 128 MB RAM, 16 MB GFX RAM)


1 - video conversion test:
ffmpeg-svnr19793-m68k (68040fpu)

14 min 59 s


2 - lame 3.98-2 test:
lame-3.98.2d-m68k (68040fpu)

7.Applications:Utils/lame-3.98.2d-m68k/bin> lame040fpu DH3:archiv/WhereDoAllTheJunkiesComeFrom.mp3.wav
LAME 3.98.2 32bits (http://www.mp3dev.org/)
Using polyphase lowpass filter, transition band: 16537 Hz - 17071 Hz
Encoding DH3:archiv/WhereDoAllTheJunkiesComeFrom.mp3.wav
to DH3:archiv/WhereDoAllTheJunkiesComeFrom.mp3.wav.mp3
Encoding as 44.1 kHz j-stereo MPEG-1 Layer III (11x) 128 kbps qval=3
Frame | CPU time/estim | REAL time/estim | play/CPU | ETA
9467/9467 (100%)| 1:13/ 1:13| 1:03/ 1:03| 3.3501x| 0:00
-------------------------------------------------------------------------------
kbps LR MS % long switch short %
128.0 6.7 93.3 88.6 6.9 4.5
Writing LAME Tag...done
ReplayGain: -5.4dB


4 - Quake test:
AmigaQuake 2.30, 800x600, software render

Demo1 40.6 FPS
Demo2 40.3 FPS
Demo3 40.0 FPS


Conclusion: Emulated CPU is probably little bit slower than 440EP - I think the 680x0 versions of tested applications are more optimalized (thus better result for ffmpeg)
Quake test shows power of modern GFX cards (much faster than on A1!)

@KimmoK: XP

_________________
http://www.warpdesign.fr/

@KimmoK

Quote:

KimmoK wrote: @DAX "Why start benchmarks at the very eve of a major update?" Because it is fun! And because we want to prepare to analyze improvements when 4.1.1 arrives.

_________________
Pegasos2 G3 running AOS 4.1 and MorphOS 2.0
Amikit user, tinkering with Icaros VM (AROS)
EFIKA owner
Amiga 1200

@ChrisH

Quote:

ChrisH wrote: @bernd_afa Well, by completely changing topics twice, you just proved that you are a nasty troll, who attacks whatever they think they can, rather than actually being interested in being helpful. I am not interested in arguing with trolls, nor helping them by posting benchmarks (if it does not serve your purpose then you will ignore it, but if it helps you nasty purpose then you will use it). (Edit: And yes, I could answer your implied criticism(s) - but what would be the point? You would just change topics again.)

What am I missing here? Whats wrong with benchmarks pre and post update for comparison? SAM's production offering is beta software, whatever benchmarks one gets is a true baseline of what a particular machine running that code can do at this time. And any benchmarks can be run after the update is out. And if there is a perceived troll hit the report button. Trying to bait bernd_afa into an escalation in public is not good. Again if I'm missing something please let me (or any mod) know via PM. We want concerns addressed, but this taking stuff into your own hands stuff is not the way.

_________________
Pegasos2 G3 running AOS 4.1 and MorphOS 2.0
Amikit user, tinkering with Icaros VM (AROS)
EFIKA owner
Amiga 1200

@bernd_afa

Quote:

bernd_afa wrote: @ChrisH >Well, by completely changing topics twice, you just proved that you are a nasty troll, >who attacks whatever they think they can, rather than actually being interested in >being helpful. I have done a abuse report, that you call me a troll about this is really bad.

I'm going to repeat a previous suggestion I have made to folks. When you make a report don't escalate things by announcing in public that you've done it. Its only going to make things worse. If you want to vent further after making the report vent to staff in a PM.

I would also request of you to disengage from Chris_H for a while. Your manner of writing has made a number of folks feel you are here to just stir up trouble. If you really are not here for that you should try harder to communicate the same ideas in a way that better shows you are seeking to truly be constructive.

Last edited by fairlanefastback on 17-Nov-2009 at 05:43 PM.
Last edited by fairlanefastback on 17-Nov-2009 at 05:43 PM.
Last edited by fairlanefastback on 17-Nov-2009 at 05:43 PM.

_________________
Pegasos2 G3 running AOS 4.1 and MorphOS 2.0
Amikit user, tinkering with Icaros VM (AROS)
EFIKA owner
Amiga 1200

@Pavlor
>Conclusion: Emulated CPU is probably little bit slower than 440EP - I think the >680x0 versions of tested applications are more optimalized (thus better result for >ffmpeg)

your results are really slow, can you try the fftbenchmark with your winuae ?
what results you get here.

also if you find a 68020 FPU version, this version is fastest on winuae.also dont use more accurate FPU.this setting is not need.

But anyway PPC seem fast in mp3 encoding or have maybe other lame (that use 32 bit floats instead of 64 bit floats or some tweaking)

what speed you get with native X86 lame ?

for winuae there is too winuaelame here, it use native wndows lame and work multithread.

but i should find out wy lame and ffmpeg is not faster on winuae
My AMD64 3000+ (real 1,8 GHZ)get with native X86 lame only 7* speed.but that reach a 1 GHZ G4 PPC.

with winuae i get only 2,9*

Last edited by bernd_afa on 17-Nov-2009 at 07:27 PM.
Last edited by bernd_afa on 17-Nov-2009 at 07:27 PM.
Last edited by bernd_afa on 17-Nov-2009 at 07:26 PM.

@bernd_afa

Quote:

your results are really slow, can you try the fftbenchmark with your winuae ?

7.Ram Disk:> fftdemo
Speed test for FFT + iFFT: (float)
time needed 3839ms for 413696 samples, => 1.2217845916748x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer)
time needed 4504ms for 413696 samples, => 1.04139232635498x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 1736ms for 413696 samples, => 2.70186114311218x speed @44100Hz/stereo

It is really slow in comparison with your numbers.

@pavlor

this show your winuae settings are very wrong and slow.so benchmarks are always good to show if all run ok with a system.

Can you send me please your winuae config ?
I send you address with PM.

winuae and core duo get this values.so i think your system must be faster.

----------------
winuae Core Duo 1,8 GHZ Notebook

Speed test for FFT + iFFT: (float)
time needed 1149ms for 413696 samples, => 4.0821852684021x speed
@44100Hz/stereo
(4136 ms at 500 MHZ)
Speed test for FFT + iFFT: (integer)
time needed 1275ms for 413696 samples, => 3.67876935005188x speed
@44100Hz/stereo
(4590 ms at 500 MHZ)
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 604ms for 413696 samples, => 7.76561403274536x speed
@44100Hz/stereo
(2174 ms at 500 MHZ)
------------------

Your values are too slower as a Netbook with 1,6 GHZ Atom 270.And this netbooks are really slow.Its better to pay 50Eur more and get one with a Pentium M or Celeron mobile M (Banias or Dothan) or Core CPU.in inet can also read that this ATOM CPU is 2-3* slower at same clockrate with native X86 benches

http://en.wikipedia.org/wiki/Pentium_M

I get in the past values from that and notice they have same clockrate/performance as my AMD64.Bit if somebody have such a Pentium M or Celeron M Notebook he can please post fftspeed values here

http://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=29964&forum=25&start=0&viewmode=flat&order=0

Atom N270 1,6 GHZ Netbook speedmode/ fastest = shorttest time of accu

Speed test for FFT + iFFT: (float)
time needed 3837ms for 413696 samples, => 1.22242140769958x speed @44100Hz/stereo
------- (12278 ms at 500 MHZ)
Speed test for FFT + iFFT: (integer)
time needed 3123ms for 413696 samples, => 1.50189912319183x speed @44100Hz/stereo
-------(9993 ms at 500 MHZ)
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 1654ms for 413696 samples, => 2.83581066131591x speed @44100Hz/stereo
-------(5292 ms at 500 MHZ)

I also look in the inet for more PPC lame results.maybe the 68k lame is not optimal build.

Last edited by bernd_afa on 18-Nov-2009 at 08:58 AM.
Last edited by bernd_afa on 18-Nov-2009 at 08:57 AM.
Last edited by bernd_afa on 18-Nov-2009 at 08:54 AM.
Last edited by bernd_afa on 18-Nov-2009 at 08:50 AM.

@bernd_afa

Quote:

this show your winuae settings are very wrong and slow.so benchmarks are always good to show if all run ok with a system.

You numbers are 6 times faster than mine! I wonder how fast would be my WinUAE box with your WinUAE configuration - faster than G5?

Quote:

Can you send me please your winuae config ?

I will send it on Friday (tested computer is not mine)

@pavlor

>I will send it on Friday (tested computer is not mine)

if you have done only a quick test, make sure winuae run on P96 and sound is set automatic switching.

then paula is only emulate when a program do paula output.paula you need selten because winuae have 16 bit upto 96 khz ahi without paula for fullduplex record and playback

but anyway on 68k should play with lame to compile with diffrent compilers, maybe this can work faster too.but for next test try the 68020 fpu Version this is little faster

Last edited by bernd_afa on 18-Nov-2009 at 03:15 PM.
Last edited by bernd_afa on 18-Nov-2009 at 03:15 PM.

@bernd_afa
From what I see, WinUae is just a performance HOG, any native app on Windows runs circles around it billions of times.
This also means that as soon as the next native (and much more powerful) PPC hardware comes out for Aos4, it will run circles around any emulated environment even if it is a 5Ghz I7.

Bottom line, we are having a ton of fun today (with Sam), and we'll be top of the game tomorrow. Does this makes you exited? Thrilled? Randy?

_________________
SamFlex Complete 800Mhz System + AmigaOS 4.1 Update 4
Amiga 2000 DKB 2MB ChipRam GVP G-Force040 Picasso 2 OS3.9 BB2
AmigaCD 32

@Tuxedo

CGXbench for 68k (WinUAE box of my brother: Core 2 Q6600 2.4 GHz, GForce 8600GT; WinUAE 1.5.0: 68040 JIT, 128 MB Fast, 16 MB GFX, OS3.9; scale instead of switching resolution used).


Simple CyberGraphics Benchmark v1.1 by Thomas Wenzel


Raw transfer speed
------------------
Screen | Register to | FAST RAM to
depth | video RAM | video RAM
-------+-------------+-------------
15 | 272.7 MB/s | 319.2 MB/s
16 | 320.1 MB/s | 320.1 MB/s
24 | 208.0 MB/s | 311.5 MB/s
32 | 295.9 MB/s | 311.3 MB/s


WritePixelArray() 320x240
-------------------------
Screen | Source: LUT8 | Source: ARGB
depth | secs | fps | MB/s | secs | fps | MB/s
-------+-------+-------+------+-------+-------+------
8 | 0.06 | 3174.8 | 232.5 | ----- | ----- | ----
15 | 0.37 | 543.6 | 79.6 | 0.09 | 1031.6 | 151.1
16 | 0.42 | 473.7 | 69.4 | 0.10 | 974.7 | 142.8
24 | 0.60 | 331.5 | 97.1 | 0.07 | 1419.4 | 415.8
32 | 0.38 | 526.1 | 154.1 | 0.05 | 1779.8 | 521.4


ScalePixelArray() 320x240 -> 640x480
------------------------------------
Screen | Source: LUT8 | Source: ARGB
depth | secs | fps | MB/s | secs | fps | MB/s
-------+-------+-------+------+-------+-------+------
8 | ----- | ----- | ---- | ----- | ----- | ----
15 | ----- | ----- | ---- | 1.34 | 74.5 | 10.9
16 | ----- | ----- | ---- | 1.09 | 91.5 | 13.4
24 | ----- | ----- | ---- | 0.78 | 128.4 | 37.6


Conclusion: Generally as fast or faster (2x) than your A1SE with Radeon9250

@DAX
>From what I see, WinUae is just a performance HOG, any native app on Windows >runs circles around it billions of times.

lame encoding or something compression is worst case for winuae.because seem code need lots of compare of values and branch.and every branch winuae need check for chipset events.

As i told i get on my 1,8 GHZ system when encode mp3 with native lame 7* speed.with winuae i get 2.9 speed.so native is only 2.5 faster.

I dont know, maybe lame use on X86 some X86 asm code.

but the good news is that most time consuming stuff need not lots branches, for example decode a compressed thing.

Amithlon is lots faster on code that do lots branches as encoding.maybe somebody post lame bench values from Amithlon.

>Bottom line, we are having a ton of fun today (with Sam), and we'll be top of the >game tomorrow. Does this makes you exited? Thrilled? Randy?

i only believe what can buy, it happen too often in the world that announce future not happen.
maybe this motivate to speedup winuae more so maybe branches can use the virtualizating funcs of newer CPU.

its also possible to change gcc that it create more JIT friendly code.

at last its always possible if something is too slow on 68k to use winuae native code.

but to avoid spend work on doing it, it help to see how fast OS4 or MOS HW.and if winuae is faster with same price HW, i see no motivation to speedup winuae or do winuae native code

Last edited by bernd_afa on 21-Nov-2009 at 10:04 AM.

more benchmark ideas
- some MS-DOS benchmark to show emulation speed
- Basilisk/system7.5.5 benchmark
- AIBB in E-UAE of AOS4

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@Tuxedo

GLQuake (BlitzQuake_68k) on WinUAE box of my brother: Core 2 Q6600 2.4 GHz, GForce 8600GT; WinUAE 1.5.0: 68040 JIT, 128 MB Fast, 16 MB GFX, OS3.9; scale instead of switching resolution used). Wazp3D beta 45.

320x200:
Demo1 - 16.1 FPS
Demo2 - 15.8 FPS
Demo3 - 14.6 FPS

640x480:
Demo1 - 4.9 FPS


Conclusion: CPU only Warp3D implementation is (unsurprisingly) slow .

I'll add more when I get the time, but for now I did a few tests with Quake 1.

320x200 software:
Timedemo1(3 runs): 614.04, 554.39, 599.68
Timedemo2(3 runs): 534.32, 589.20, 553.87
Timedemo3(3 runs): 566.95, 568.71, 508.33

640x480 software
Timedemo1(3 runs): 136.60, 135.22, 138.99
Timedemo2(3 runs): 129.21, 142.79, 131.70
Timedemo3(3 runs): 123.39, 132.32, 161.24

800x600 software
Timedemo1(3 runs): 92.81, 88.42, 91.57
Timedemo2(3 runs): 87.36, 88.80, 82.37
Timedemo3(3 runs): 87.49, 92.20, 86.44

1024x768 software
Timedemo1(3 runs): 58.40, 56.35, 57.16

The above is running on AROS native on a core2duo clocked @3.86ghz, RadeonX550, 2 gig RAM (ddr2@900ish mhz).
I havent done any hardware rendered tests yet as AROS 3d drivers wont be available for another few months. All test were run windowed (desktop res of 1600x1200) as I couldnt work out how to go full screen (or if its even possible with the version of Quake I tested). I'll do some Lame encoding/decoding and video encoding/decoding later tonight or tomorrow.

@Pavlor
>GLQuake (BlitzQuake_68k) on WinUAE box of my brother: Core 2 Q6600 2.4 GHz, >GForce 8600GT; WinUAE 1.5.0: 68040 JIT, 128 MB Fast, 16 MB GFX, OS3.9; scale >instead of switching resolution used). Wazp3D beta 45.

you should not use scale resolution, because it cost lots performance to scale the 320*200.please try without it.

but anyway i try your test too.I get 16 frmaes too on my AMD64 with real 1,8 GHZ.but i have only 256 kb 2. Level Cache, so i thought the core duo with the large caches must be lots faster in software 3d.

doing real 32 bit 3d with CPU seem really performance hungry.so the 8 bit software renderer quake have and give lower quality is important.

Or maybe there is a Problem in wazp, i notice drawing problems and the status bar flicker.

Last edited by bernd_afa on 08-Dec-2009 at 04:02 PM.