Let's reply here as well:

@Tomppeli

Quote:

David Brunet from/for Obligement got 20 secs for WB and 10 + 29 secs for Ambient icons mode and 6 secs for WB and 2 + 15 secs on Ambient. Because he compared Workbench's 6 secs to Ambient's 2 secs he assumed that Workbench loads only file names. But in reality Workbench loads always names, icons plus file types no matter of selected mode (anybody can verify this easily). So correct comparison is 6 secs for Workbench and 15 secs for Ambient. That means AmigaOS4 & Workbench is more than 2 times faster than MoprhOS & Ambient. I don't know if Ambient loads file types so in that case Workbench does more work in the recorded time. So Ambient is quite poor. Also if somebody wants to test loading time of file names only that person can download Filer from os4depot. Filer would perform this job on his machine in an eye blink (calculated compared to my slower system) compared to Ambient's poor 2 secs.

Well, you don't know how ambient works, because it also gets filetype (in both icon and list mode), and displays configured default icon (in both modes too). That said, in list mode, If icon column is disabled, only directory scan is processed, and then opening that directory takes 1s.

Besides, that test is particularly unfavorable to ambient, given the amount of configured filetypes, but it deals with it smartly, unlike workbench :

1. be it in icon or list view, you have full access to your files in respectively 10 or 2 seconds (it also already displays a "default" icon at that stage), which is definitely not the case with workbench, where you have to wait that the whole dir +filetype scan is finished.

2. default ambient mimetype database comes with about 500 filetypes, while there are about 100 in Workbench deficon default prefs. So many more rules are checked, and given the filetype+name of the files in this test, almost all rules had to be checked each time (no pattern hint to reduce it, not recognizable header and so on...). So, considering that in both cases all rules probably had to be checked, ambient also scans filetypes quite faster.

3. With ambient, visible entries are dynamically refreshed first (default icons, thumbnails, whatever). With that, i mean that if one scrolls during the scan, then instead of continuing to scan sequentially, it will resume scan from the new first visible entry.
So from an user POV, even if the mimetype scan isn't finished, the transition state is always very short for the visible entries (0.1s or so at most).

4, i tried Filer OS4 (i assume you refer to that one and not to the original 68k filer which is completely different) with 3000 files, and it was certainly not an eye blink either (2s, and it doesn't process as much stuff as ambient).

For comparison, Magellan opens this dir in about 1s, and Filer (the original 68k program) in less than 0.5s. Neither Magellan nor Filer process filetypes or icon display in that case, obviously.

So, for a fair comparison, you'd actually need to divide ambient filetype database by 5 (which would reduce the total mimetype scan time, obviously). But in any case, Ambient is more reactive, giving useful feedback way faster than workbench.


Quote:

That Obligement article doesn't say was it using DHCP or static IP for network or did the tester add "run" into Startup-sequence in front of "addnetinterface" command. So if we can assume he used DHCP so swithing to static IP or adding that run command into the S-S will take at least 3-4 secs away from the boot time. Also MorphOS uses single boot.img file to load the system and AmigaOS4 uses many small Kickstart files. On modern hard drives and their DMA loading a single file is faster than loading multiple small files. But the end user can't change the configuration of MorphOS because its source code is not available to let user to compile his own version. But anybody can modify human readable Kicklayout file (AmigaOS4). So if there's still a couple of seconds difference in boot times I'm happy to sacrifice those couple of seconds for full configurability.

First, your statement about not being able to change boot.img configuration is a bit wrong. You can filter out any module at will, and then use a disk version of the module if wanted. Also, with MorphOS PowerUP, you can also add external rom modules.

Then, having a synchronous DHCP init is no excuse.
By the way, here, if i let DHCP synchronous on OS4.1, then it takes about 1 minute to boot, while DHCP init is way faster (and async) on MorphOS, but don't ask me why, it may even be my router fault. :)


Quote:

But it's said that Altivec can give 10-30% speed up so taking 10-30% away from AmigaOS4's results makes it quite comparable to those MorphOS results. Altivec might help some codecs more than the others that's possibly why there's so much difference between different codecs.

Altivec can help *sometimes*, but it doesn't explain a x2 ratio. For instance, it didn't help *that* much in the "benchmark.avi" video i had used in a previous benchmark months ago (and which is also used in this benchmark).
With altivec, the video was played in 76s, and without, 88s. To be compared to the 151s needed on OS4 (you can now assume it would take 139s if it was altivec-accelerated).

Last edited by Fab on 17-Aug-2009 at 12:29 PM.
Last edited by Fab on 17-Aug-2009 at 12:28 PM.

This whole thread reminds me of the days of Amiga/520ST bickering back and fourth.

Come on guys, the facts are there, but it doesn't make OS4 a bad OS!

_________________
-- --
aka Tony Rocks

@m0lebrain

Hey, we can't bash Atari anymore and bashing Microsoft - hmm - that's no fun. So I'm not sure who is Atari and who is Amiga in our little disagreements here

_________________
X1000|II/G4|440ep|2000/060|2000/040|1000

du hast recht...NOT. hehe we can always bash Atari on an Amiga forum ;) Amiga is Amiga and MOS is Atari :)

_________________
-- --
aka Tony Rocks

Quote:

This is one of the most unprofessionally made tests I've seen.

Your comment is also one of the most unprofessionally made comments I've seen.
You doesn't have read/understand my comments on all tests and/or doesn't know how the test is made.

You have noticed some "unfair" tests on a 58 tests benchmarks ? You perhaps know it's difficult to make exactly the same conditions on two different systems.

I'm a supporter of all Amiga systems. I like and use AmigaOS 3.x, 4.x, AROS and MorphOS. I have made those test in the more fair manner I can do.

It's unfair for which system ? You seems to focalise only on AmigaOS 4.1. You know, there are also unfair tests for MorphOS like :

- SFS (oldest one, default configuration -line sof buffers, etc.- have not optimal values,...).
- Boot time (my MorphOS system have a bigger taskbar, more icons on the desktop, a bigger user-startup and perhaps more things which can slow the boot time).
- Warp3D (it's the official 3D system/driver for AmigaOS 4.1 while MorphOS only use a reimplementation).
- Display of icons : Ambient use a *BIG* database of mime types, so it must seek much information to display the right icon than Workbench (but Kiero or Fab could explain it better).
- Radeon 9250 (it's the card recommanded by Hyperion for AmigaOS 4.1 on Pegasos II. On MorphOS the best card is Radeon 8500/9000PRO, so it's possible that the differences in 3D tests can be even greater for MorphOS with this).


Quote:

Opening drawer: David Brunet from/for Obligement got 20 secs for WB and 10 + 29 secs for Ambient icons mode and 6 secs for WB and 2 + 15 secs on Ambient. Because he compared Workbench's 6 secs to Ambient's 2 secs he assumed that Workbench loads only file names. But in reality Workbench loads always names, icons plus file types no matter of selected mode (anybody can verify this easily). So correct comparison is 6 secs for Workbench and 15 secs for Ambient. That means AmigaOS4 & Workbench is more than 2 times faster than MoprhOS & Ambient. I don't know if Ambient loads file types so in that case Workbench does more work in the recorded time. So Ambient is quite poor. Also if somebody wants to test loading time of file names only that person can download Filer from os4depot. Filer would perform this job on his machine in an eye blink (calculated compared to my slower system) compared to Ambient's poor 2 secs.

The goal of this test is : 1. time to access a file, 2. time to display all icons, 3. memory used.
For a more complete explanation how Ambient is used, see fab's comment some hours ago.

And no, I don't use Filer, only default tools in each system.

Quote:

Lha test: Everybody knows AmigaOS4 uses OFS on Ram disk. I got much better results with JXFS on a hard drive which were comparable to Brunet's results on MorphOS and Ram disk using an estimation that G4-1000MHz is two times faster than my 750CXe-600MHz. RAM is always faster than a hard drive (that's why they're using caches in RAM to speed up disk operations) so that means results of MorphOS are not that good. Somebody should remake Lha test using FFS and SFS on both machines and JXFS on AmigaOS4. (That's one of the reasons why that test made by Brunet was very unprofessional.) Ram disk was more important in 1980's when all Amiga's were floppy based. With nowadays fast hard drives Ram disk is less meaningful. I save usually small ascii text files onto Ram disk so small differences on Ram disk speeds is irrelevant. Also if somebody's testing speeds of unpacking I noticed that used con-handler and size of the Shell window affects results so when doing tests with Lha tester should use quiet (-q) option or ">nil:". (That's another reason why Brunet's test was unprofessional.)

In this test RAM is used because it is more common, not because the system "X" is better the system "Y" in this test. In the next update of my article, I will add another test with SFS partition as output. The con-handler used is the default provided by the system. There is no visible difference with or without "-q" argument (it's a packing test with only one file, so not much is displayed on the console). And for a better comparison, I will use a publicly available file (pack a 38 MB video file is harder than a 38 MB Ascii and "empty" file).

Quote:

Boot time: That Obligement article doesn't say was it using DHCP or static IP for network or did the tester add "run" into Startup-sequence in front of "addnetinterface" command. So if we can assume he used DHCP so swithing to static IP or adding that run command into the S-S will take at least 3-4 secs away from the boot time. Also MorphOS uses single boot.img file to load the system and AmigaOS4 uses many small Kickstart files. On modern hard drives and their DMA loading a single file is faster than loading multiple small files. But the end user can't change the configuration of MorphOS because its source code is not available to let user to compile his own version. But anybody can modify human readable Kicklayout file (AmigaOS4). So if there's still a couple of seconds difference in boot times I'm happy to sacrifice those couple of seconds for full configurability.

About boot time, it's DHCP for both and yes I have added "run" into startup-sequence in front of "addnetinterface". With fixe IP, it almost the same time (21,4 s / 21,9 s).

Your comment on "full configurability" is useless here.

Quote:

Using MPlayer for testing: I hate to say this because I do appreciate Afxgroup's efforts and hard work to give modern important apps to us but I'm forced to say this. So I'm sorry. That MPlayer test tests more porting skills of two individuals rather than anything in two operating systems. April version of MPlayer doesn't have Altivec acceleration enabled. There's older versions which have it enabled on the very same web page. (But they might print more debug info and are older code base.) So somebody have to make this kind of tests either on a G3 machine which doesn't have Altivec to affect results or somebody have to help Afxgroup to enable Altivec for MPlayer. But it's said that Altivec can give 10-30% speed up so taking 10-30% away from AmigaOS4's results makes it quite comparable to those MorphOS results. Altivec might help some codecs more than the others that's possibly why there's so much difference between different codecs.

You are right here. I've used the latest version of MPlayer for AmigaOS 4.x. I didn't know this release was not Altivec optimised. So I will add non-Altivec results too (but even with this, the first results I have are always way better for MorphOS).

My article will be updated in september with more benchmarks. Feel free to contact me ("obligement AT free POINT fr") if you have some ideas on "fair / professional" tests.

Last edited by Daff on 17-Aug-2009 at 08:31 PM.

@Daff

Not wanting to comment on any of the numbers as I don't have either system at the moment.
But regardingMPlayer.. This is a bogus test and very favourble to MorphOS.
The Amiga version is a simple port with some "must do" fixed parts to get it to work.
The MorphOS version as I see it is a more mature port.
And no this has nothing to do with use of Altivec or not.
Most Amiga users uses DVPlayer for general video playback AFAIK.

But I still think MorphOS would win the speed tests.. Many parts are more mature than the
ones in OS4. Ambient, MUICON, Cybergraphics and 3D drivers.

_________________
/Harry

[SOLD] µA1-C - 750GX 800MHz - 512MB - Antec Aria case

Avatar by HNL_DK!

@ Samwel :

You're right. I will add a test with DvPlayer.

@Samwel
Quote:

The MorphOS version as I see it is a more mature port.

So why doesn't anyone port the MorphOS version to AmigaOS4?

Quote:

Most Amiga users uses DVPlayer for general video playback AFAIK.

Just curious, does in nowadays support basic features, such as ogm, and more importantly mkv subtitles?

And regarding mplayer test, if, say, blitting routines of OS4 version are suspected to be lower, you can do the test with -vo null (which disables any video output, afaik, the video is still fully decoded). Audio output can also be disabled in the same way (-ao null)

@Jupp3

You'll have to check with COBRA on what DVPlayer supports or not..
As I said before I have no AmigaOS or MorphOS system at the moment.

_________________
/Harry

[SOLD] µA1-C - 750GX 800MHz - 512MB - Antec Aria case

Avatar by HNL_DK!