@pavlor

Quote:

68k JIT in OS4/MOS must be really good - two times faster than JIT in WinUAE (according to your results and estimated CPU performance of Core i7 and 7445).

You must be reading different results from what's actually there. I have these values an in all those they are fairly greater then those done by any PPC machine. They're pretty easy to do I get the i7 Mhz and divide it by the competing machine and get:

How faster it is Mhz wise relative to the competing machine:

2.6x macmini
4.9x aone
5.9x sam
26x 604
78x Classic

Then I go on dividing the tests times against those of i7 so I get how much more they've taken to perform:

macmini 6.5 | 3.2 | 2. 7
aone 13.4 | 8.4 | 8.2
sam 19.7 | 13.9 | 16.7
604 62.3 | 48.5 | 47.4
Classic 109.7|90.9|108.2

Finally I check the relation between the Mhz and that of the tests.
2.5x | 1.2x | 1.0x | macmini
2.7x | 1.7x | 1.7x aone
3.4x | 2.4x | 2.7x sam
2.4x | 1.9x | 1.8x 604
1.4x | 1.2x | 1.4x Classic

So actually no, it's not as if this values favor PPC hardware

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@pixie

Quote:

You must be reading different results from what's actually there.

Again, on comparable hardware (eg. PIII vs G3), OS4 JIT results will be (more than 2 times) better than WinUAE JIT results.

Core i7 (2.6 GHz) is nearly 20 times faster than 440EP, thus WinUAE JIT has much better result, but that doesn´t mean that WinUAE JIT is faster than OS4 JIT, it runs only on much faster hardware.

You can complete your chart with a Pegasos2/G4

MorphOS:

time needed 3022ms for 413696 samples, => 1.55209481716156x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer)
time needed 1626ms for 413696 samples, => 2.88464379310607x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 619ms for 413696 samples, => 7.57743263244628x speed @44100Hz/stereo

OS4:

Speed test for FFT + iFFT: (float)
time needed 3183ms for 413696 samples, => 1.47358810901641x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer)
time needed 2235ms for 413696 samples, => 2.09862685203552x speed @44100Hz/stereo
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 961ms for 413696 samples, => 4.88078117370605x speed @44100Hz/stereo

There are better tests to test JIT performance, though.

Last edited by Fab on 15-Nov-2009 at 12:21 AM.
Last edited by Fab on 15-Nov-2009 at 12:06 AM.

@pavlor

Quote:

Core i7 (2.6 GHz) is nearly 20 times faster than 440EP, thus WinUAE JIT has much better result, but that doesn´t mean that WinUAE JIT is faster than OS4 JIT, it runs only on much faster hardware.

The ratio of 3.9Ghz is 5.9x how do you get 20x of a slower i7? When comparing you should compare what can be compared, so you compare 440EP with one core of i7, so that 24 to 48 times the speed of a sam in theoretic raw power can only be accounted by the speed of the core running WinUAE. In the end you'll see it trounces 440EP, at least with the results we have here.

It always be hard to say if Jit is faster due to the harware or not, what it can be compared is speed and times, and i7 with one core trashes all competition but MacMini and Classic, and even on this cases it is between 1x and 2x values. There's not a test where it is slower then the competition.

Last edited by pixie on 15-Nov-2009 at 12:30 AM.
Last edited by pixie on 15-Nov-2009 at 12:26 AM.

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@pixie

Quote:

The ratio of 3.9Ghz is 5.9x how do you get 20x of a slower i7?

According to many benchmarks (eg. EEMBC etc.) 440EP 666 MHz is as fast as 750CXe 600 MHz.

You can compare Spec CPU results for comparison

750CXe 600 MHz scores (estimated by IBM):
25.6 SpecInt95 (= cca 230 SpecInt2000 = cca 1.7 SpecInt2006)
16.3 SpecFp95
Source

Core i7-920 2.6 GHz scores:
28.6 SpecInt2006 Source
29.3 SpecFp2006 Source

Thus
Core i7 2.6 GHz is 16.8 times faster than 440EP 666 MHz in integer operations.

@pavlor
>Core i7 2.6 GHz is 16.8 times faster than 440EP 666 MHz in integer operations.

so we need a 9,9 GHZ 440EP.

when use specfp SAM is more than 20 times slower.then we need a 13 GHZ PPC.

@pixie.

i told in the text that a PPC JIT must be faster on a PPC CPU that is as fast as X86.
because X86 need byteswap(the Bench need lots byteswap), winuae need chipset emulate, and X86 have less register.

that X86 is still lots faster is because its lots more powerfull, not only the clockrate is much higher, also the performance /clockrate is much higher.

BTW: you have efika image, if you own a efika, please do the test.Or somebody else with efika can do the Test.

With efika then we have all HW that can run 68k Softrware that support a JIT.

@Fab

thanks for your values.
Sure there are better tests, but we know that if a testing cost too much time, user dont like to spend some minutes with that.

this test cost not more than 30 sec so chances are better that more do it.

The test can also continue, we all know users are happy when Java was announced for OS4.

But the same guys say that they dont want use 68k programs or code in amiblitz because its not native PPC.
so when OS4 maybe get Java we can do same Test in OS4 Java and see that Java speed is slower as amiblitz code on winuae.

On PPC 68k code must be faster as Java because this need no byteswap.

Last edited by bernd_afa on 15-Nov-2009 at 01:27 PM.
Last edited by bernd_afa on 15-Nov-2009 at 09:20 AM.
Last edited by bernd_afa on 15-Nov-2009 at 09:18 AM.

@pavlor

One core?

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@pixie

Quote:

One core?

Yes.

@bernd_afa
Quote:

when use specfp SAM is more than 20 times slower.then we need a 6,6 GHZ PPC.

Theoretical 68K emulation tests are pointless and we don't need a 7Ghz 440EP.

What we need is a new system that is fast enough in performing what it needs to perform.
If the software I use gives me fast results, I don't care what processor or programming trickery is used underneath.

As long as the next Amiga HW runs software in a very fast and responsive way we will be more than happy.

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

@DAX

Quote:

Theoretical 68K emulation tests are pointless and we don't need a 7Ghz 440EP.

That was me who wrote that Core i7 is 20 times faster than 440EP - this estimate was based on native benchmarks, not 68k JIT emulation results.

@pavlor

So the benchmarks regarding i7 speed are done so that they show just one core... please.

Quote:

CPU(s) enabled: 4 cores, 1 chip, 4 cores/chip, 2 threads/core

Last edited by pixie on 15-Nov-2009 at 11:40 AM.

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@pavlor
Yes I know, my point about this thread, is that trying to establish speed comparisons between different HWs based on Jit 68K emulation is just a small piece of the puzzle and as an overall benchmark is totally pointless.

Also I get the impression (not from you of course) that someone here is trying to subtly convey that we should abandon PPC since we can emulate in WIN-Uae "way faster" thanks to 4Ghz x86 processors.

Again it would seem a pointless assumption. All we need is to run emulated Classics at 1:1 speed (and we gladly wait for new PPC HW for that).

Last edited by DAX on 15-Nov-2009 at 12:11 PM.

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

@pixie

Quote:

So the benchmarks regarding i7 speed are done so that they show just one core... please.

I will cite SPEC2006 FAQ (note the phrases in bold):
Quote:

Q15. What is the difference between a "rate" and a "speed" metric? There are several different ways to measure computer performance. One way is to measure how fast the computer completes a single task; this is a speed measure. Another way is to measure how many tasks a computer can accomplish in a certain amount of time; this is called a throughput, capacity or rate measure. •The SPEC speed metrics (e.g., SPECint2006) are used for comparing the ability of a computer to complete single tasks. •The SPEC rate metrics (e.g., SPECint_rate2006) measure the throughput or rate of a machine carrying out a number of tasks. For the rate metrics, multiple copies of the benchmarks are run simultaneously. Typically, the number of copies is the same as the number of CPUs on the machine, but this is not a requirement. For example, it would be perfectly acceptable to run 63 copies of the benchmarks on a 64-CPU machine (thereby leaving one CPU free to handle system overhead). Note: a speed run which uses a parallelizing compiler to distribute one copy of a benchmark over multiple CPUs is still a speed run, and uses the speed metrics. You can identify such runs by the field "Auto Parallel".

@pavlor

so we have: SPECint2006 28.6 for i7 and SPECint95 25.6 for 750CXe, what does this mean? Why you present two different units which can't be directly compared at all?

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@pixie

Quote:

so we have: SPECint2006 28.6 for i7 and SPECint95 25.6 for 750CXe, what does this mean? Why you present two different units which can't be directly compared at all?

Yes, one can´t directly compare generations of SPEC benchmarks, however, it is possible to estimate some correlation between those generations:

eg. 34-50 SpecInt92 = 1 SpecInt95
9-11 SpecInt95 = 100 SpecInt2000
110-150 SpecInt2000 = 1 SpecInt2006

I use this (rough) recalculation:
1 SpecInt95 = 9 SpecInt2000
135 SpecInt2000 = 1 SpecInt2006

You can compare my estimate with real results:
Pentium III 750 MHz: 330 SpecInt2000 or 36 SpecInt95
750GX 1 GHz: 469 SpecInt2000 or 52 SpecInt95

Pentium 4 670 3.8 GHz: 12.3 SpecInt2006 or 1852 SpecInt2000
Core 2 Duo E6600 2.4GHz: 18.3 SpecInt2006 or 2563 SpecInt2000

@Wanderer
>Thanks for hotlinking to private links.

What do mean with private links , the links you have post in the amiforce Forum and so they are not private.

>What we need is a new system that is fast enough in performing what it needs to >perform.

No, to get new software the system must be at comparable speed with other.only then devs begin write programs when they see the user have the power for it.

see in PC History, there were many years graphiccards sold with 3d features, but most time this was not need but user get 3d for near free.only later come programs that use it.now since all have 3d and pixelshader and Hardware lookup and translation, there is no new program written anymore that work with direct x3

same happen with directx 7 9 10 features.Developer make only soft with the new features if the features can use many users.

So to get exciting Software and new users, the most important point is that the speed is good enough for that software

Last edited by bernd_afa on 15-Nov-2009 at 02:21 PM.
Last edited by bernd_afa on 15-Nov-2009 at 02:21 PM.
Last edited by bernd_afa on 15-Nov-2009 at 02:19 PM.

the result of a blizzard PPC i add in the Intro Post too.see more here

http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=30008&start=60&post_id=525273&order=0&viewmode=flat&pid=0&forum=2#525273

Blizzard PPC 321Mhz 80.333Mhz bus. OS4.0

Speed test for FFT + iFFT: (float)
time needed 17928ms for 413696 samples, => .261626005172729x speed @44100Hz/stereo
(11566ms at 500MHZ)
Speed test for FFT + iFFT: (integer)
time needed 15377ms for 413696 samples, => .30502900481224x speed @44100Hz/stereo
(9920 ms at 500MHZ)
Speed test for FFT + iFFT: (integer handoptimized 68K ASM)
time needed 12774ms for 413696 samples, => .367185741662979x speed @44100Hz/stereo
(8221 ms at 500MHZ)
-------------------------------------------------------------

Last edited by bernd_afa on 13-Dec-2009 at 02:47 PM.

@bernd_afa: >So to get exciting Software and new users, the most important point is that the >speed is good enough for that software

If that is really true, AROS must have a lot more of exciting new software, because it is available for faster hardware.

I think it´s much more important to have a stable OS, a good SDK and all the "basic"-software (good and modern web-browser, office programs etc.) to attract users and developers.

_________________
AmigaOne X5000/40, 2.2 Ghz, 4 GB RAM, Radeon R9 280X, M-Audio Revolution 5.1, 240 GB SSD

@AP

>If that is really true, AROS must have a lot more of exciting new software, because >it is available for faster hardware.

The disadvantage of AROS is that it is not promote as future there is no marketing on it, not much money invest and many bash against it and devs do not support much.
But what do you think the Hardware of AROS and the marketing of OS4, does this not attrakt more users ?

Or lets say a 4 Core PPC 3 GHZ System für 500 Eur ?
A notebook with 1,6 GHZ dualcore PPC for 500 Eur ?

A nettop for 250 Eur ?

I think so you can attract more users.
ANd if a rivival of amiga really want realize there must be invest money in a system that can offer fas speed and low price.

but now we have

AROS with his fast hardware, but less money and marketing and less developer support
OS4 with slow Hardware some more marketing and money and developer support and future announces.

so best is good Hardware price and power and much marketing and money.

that can get out of AROS with more money more, you can see on MOS.MOS use AROS.

Last edited by bernd_afa on 13-Dec-2009 at 07:44 PM.

@AP

Quote:

I think it´s much more important to have a stable OS, a good SDK and all the "basic"-software (good and modern web-browser, office programs etc.) to attract users and developers.

So you don't think that a cheap yet powerful machine is important into this kind of reasoning?

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga