@nexus

Quote:

Why is the 800Mhz Sam flex only half as fast as the 667Mhz Sam ep?

I expect that the 9250 in the Flex only has a 64-bit memory bus. The M9 or Radeon 9000 has a core clock of 250Mhz and memory bandwidth of 6.4GB p/s and a 64-bit 9250 has a core clock of 240Mhz and memory bandwidth of 3.2GB p/s.

@Rob

Yes as pointed out in AmigaNews.it thread 9250 is slower than M9 on SAMep. If you put a radeon 9000 on SAM flex you get better result i suppose.

_________________
ikir

@ChrisH

It might cause slowness as a consequence of heavy debug output spam, when illegal accesses occur, especially if you use serial debug. But these illegal access didn't always happen, especially not in first level, IIRC.

@Nexus

It's just it's a runtime patch, and it hasn't been compiled for OS4. But someone could do the same thing for OS4. It shouldn't be hard to find these illegal accesses in the code.

Last edited by Fab on 10-Apr-2010 at 01:43 PM.
Last edited by Fab on 10-Apr-2010 at 01:42 PM.

I've found Smokin' Guns to work without freezing - OpenArena and Urban Terror freeze ioquake3 regularly.

_________________
: AmigaOneXE (unmod.) 750FX/512 MB +stuff & AmigaOS 4.(0|1)
: A1200/68060&96MB/SCSI/EM1200-Voodoo3 & OS 3.5
: A500/1MB
: Pegasos (ff) 512 MB & MorphOS
Praise seitan.

@ikir

Radeon 9250 is not as slow to get such a result. Lower memory bandwitch should affect only the highest resoultions. On PIII 600 with much weaker GeForce 256 the game is rendered faster...

_________________

@all

Has anyone collected cube -benchmark results with AOS4.1.u1?

http://os4depot.net/index.php?function=showfile&file=utility/benchmark/cubebm.lha

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

@Radov

Quote:

Radov wrote: @ikir Radeon 9250 is not as slow to get such a result. Lower memory bandwitch should affect only the highest resoultions. On PIII 600 with much weaker GeForce 256 the game is rendered faster...

Sure i was just comparing Sam440ep 667Mhz to Sam flex 800Mhz, Sam440 win but i think it is the graphics card which did the difference. No other expanations: flex has more ram and cpu power.

@Kimmok

Cube for OS4? wow didn't know that.... going to try
EDIT: ah Cube benchmark! Not the game. We can try this bench too.

Last edited by ikir on 10-Apr-2010 at 07:00 PM.
Last edited by ikir on 10-Apr-2010 at 06:59 PM.

_________________
ikir

@ikir

It's not THAT Cube Ikir...

Varthall

_________________
AmigaOne XE - AmigaOS 4.1 - Freescale 7457 1GHz - 1GB ram

@Varthall

yeah lol Sorry

_________________
ikir

@Fab Quote:

It's just it's a runtime patch, and it hasn't been compiled for OS4. But someone could do the same thing for OS4. It shouldn't be hard to find these illegal accesses in the code.

Hah, "It shouldn't be hard", says an expert low-level PPC coder . Frankly I wouldn't have a clue about how one would go about doing what you suggest...

The author of the patch seems to be Mark Olsen. Dunno what his "politics" are, but I'd guess he wouldn't be too keen on releasing his code so that it could be used for OS4

@all
BTW, I found this about the patch: Quote:

The Amiga port of Wipeout 2097 has some debug code enabled, which works by writing some data to a fixed address in chip memory. This address is in no way allocated or reserved, so on Amigas, Wipeout 2097 trashes a part of the chip memory area. On the Pegasos, however, there is no chip memory, so these debug writes cause an exception which leads to a debug output dump for each write. These exceptions slow down the game to a level where it is unplayable.

Last edited by ChrisH on 11-Apr-2010 at 09:48 AM.
Last edited by ChrisH on 11-Apr-2010 at 09:41 AM.

_________________
Author of the PortablE programming language.
It is pitch black. You are likely to be eaten by a grue...

@ikir & @Fab

Hi : - )

@ all
here all the tests (AmigaOS, MorphOS, OSX) :
http://amiga.ikirsector.it/forum/viewtopic.php?f=40&t=13185

and at the end of the page, you can see the test in OSX not using Ioquake (but using Quake III Arena X 1.3.2 Universal and Quake III Arena X 1.3.2c with PowerMac MDD Dual G4 con ATI Radeon 9000 Pro ).

My idea is that MorphOS2.x is working very good with 3D GL ... while not so good AmigaOS4.x.

Best regards.

Last edited by divina on 11-Apr-2010 at 12:27 PM.

@divina

Would be nice if it could be tested on Mac mini G4, for a fair comparison.

That MOS /OS4 is lots slower as Mac OS on small resolutions i think is because MOS or OS4 use no Hardware transform and lightning.
all modern OS have that and is a must have when want run modern games with many Polygons.

on higher resolutions then a slower transform and lightning doesnt matter much, because draw slow down.

That OS4 is on higher resolutions lots slower, i think on MOS is the driver async written, it maybe call a polygon draw and during this time there are some more polygons calc.

to verify if i am right, try run quake 3 at 320*200 i guess then OS4 and MOS get near same speed.

It seems safe to predict that with full OpenGL support for AOS4, x1000 should come close to 200fps @640x480 ... Perhaps more, unless Quake3 fails to take any advantage of r700 series GPUs.

But with current AOS4 miniGL we might not even get 100fps? ...


Btw. does Quake3 run on classicAmigaPPC setups?

Last edited by KimmoK on 11-Apr-2010 at 06:40 PM.

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

@ikir

Quote:

Sam440 win but i think it is the graphics card which did the difference

Ok, but my point still is that R9250's lower performance doesn't justify such a bad results.

To compare, these are my results for SamFlex800 + R7500 (64MB), taken with OS4 hardware composition turned off to preserve some V-MEM:
640x480 - 13.2 fps
800x600 - 11.1 fps
1024x768 - 6.7 fps
1280x1024 - 3.2 fps

NOTE: it's weird, but my machine easily runs out of VMEM. Only for a while higher resolutions are rendered correctly, getting massive graphical corruptions and slowdowns after the few seconds. Frame rates for the resolutions of 1024x768 and 1280x1024 should be (IMO) 50% better. Still very low, but... :)

NOTE2: With hardware composition turned on and in-game the highest graphical quality set - I get corruptions even on the menu screen and a resolution of 640x480...

--UPDATE 1--
Because of problems with VMEM I tried to lower Quake III needs and decided to change one of the settings:
SETUP - GAME OPTIONS - HIGH QUALITY SKY - off
additionally, I decided that it may be wise to set also:
SETUP - GAME OPTIONS - SYNC EVERY FRAME - off.

These changes got me a positive result, as I was able to test 1024x768 with only minor corruptions at the end. Unfortunately 1280x1024 still corrupts (and goes to the 1-2 frame rate) from the beginning....
My new results are:
640x480 - 15.3 fps
800x600 - 12.8 fps
1024x768 - 9.1 fps
1280x1024 - 3.2 fps

NOTE3: When the corruption occurs VMEM is not freed correctly anymore and warm reeboot causes system to hang.

--UPDATE 2--
Ok. Another attempt to save some VMEM. This time I've dropped workbench screen resolution from 1680x1050x32 (46MB of free VMEM after boot) to 640x480x32 (over 57MB of free VMEM after boot). It was a good choice, I was finally able to do all tests without graphical glitches.

So, my best results for the SamFlex800 + Radeon 7500 (64MB) are:
640x480 - 15.7 fps
800x600 - 13.1 fps
1024x768 - 9.7 fps
1280x1024 - 8.3 fps (increase from 3.2 fps!)

NOTE4: about NOTE3 - it seems that all reboots, in a QuakeIII context, will make the system to stop. It doesn't matter if there are problems with VMEM or not...

Last edited by Radov on 11-Apr-2010 at 10:25 PM.
Last edited by Radov on 11-Apr-2010 at 09:08 PM.
Last edited by Radov on 11-Apr-2010 at 09:06 PM.
Last edited by Radov on 11-Apr-2010 at 08:58 PM.
Last edited by Radov on 11-Apr-2010 at 08:58 PM.
Last edited by Radov on 11-Apr-2010 at 08:58 PM.
Last edited by Radov on 11-Apr-2010 at 08:38 PM.
Last edited by Radov on 11-Apr-2010 at 08:22 PM.
Last edited by Radov on 11-Apr-2010 at 07:14 PM.
Last edited by Radov on 11-Apr-2010 at 07:13 PM.
Last edited by Radov on 11-Apr-2010 at 07:10 PM.

_________________

Ok, since the mac mini 1.42GHz with original Quake3 1.32c client test was missing, i did it myself:

OSX 10.4
High Quality preset:
640x480: 77.3
800x600: 71.1
1024x768: 52.9
1280x1024: 29.6

MorphOS 2.4
High Quality preset:
640x480: 73.8
800x600: 58.2
1024x768: 48.8
1280x1024: 36.3

So, the difference is quite smaller, when you don't compare oranges with apples. :)
MorphOS is even faster in 1280x1024. :)

Last edited by Fab on 11-Apr-2010 at 07:43 PM.
Last edited by Fab on 11-Apr-2010 at 07:42 PM.

@Fab

Hi,
I have not more Mac Mini, I wait MorphOS for PowerMac G4 MDD : - ) : - ) : - )
As said before, I love MorphOS, It's the best.

regards


Quote:

Fab wrote: @divina Would be nice if it could be tested on Mac mini G4, for a fair comparison.

@ChrisH

Quote:

Quake 3 is obviously the most extreme case, pushing OS4's (Mini)OpenGL to the (performance) limit. With something less demanding I think the difference will be less obvious (or even non-existant), e.g. Wipeout 2097.

Differences are also high with WipEout 2097, GLQuake or Quake 2.
My comparative still valid with AmigaOS 4.1 Update 1 for 3D stuff : http://obligement.free.fr/articles/amigaos41_vs_morphos23.php

I have same results with AmigaOS 4.1 Update 1.

@ikir

Thanx for the great benchmark post. I started thinking about getting new Amiga finally for OS4, but 1280x1024 13,3 fps seems a bit too low for EUR600 motherboard [IIRC]? I guess it is up to 3D drivers or something... I will continue reading Amiga posts from time to time :)

_________________
nykk | deetronic.rs | youtube.com/djnykk | gfx.river | mamavolibebu.com

@Daff Quote:

Differences are also high with WipEout 2097, GLQuake or Quake 2.

Hmmm, OK, looks like there is some intrinsic bottleneck/overhead in OS4's 3D system then. I look forward to the day we see it replaced, as then things should really fly on my Sam440 !

_________________
Author of the PortablE programming language.
It is pitch black. You are likely to be eaten by a grue...