@SimplePPC

Well that only reinforces what I said. If a 750MHz Duron running a 3DNow-optimised OpenGL driver with Permedia2 attached via AGP (and assuming that allows the interrupt-driven DMA based texture download that the P2 hardware manuals describe), then I don't think the BlizzardPPC/BVision with all of it's hardware limits (in particular the slow bus and PIO based driver) would be able to achieve any better.

No surprise to see Quake3 performing better on the lowest settings - again it can make use of vertex lighting which eliminates one complete pass.

_________________
Doing stupid things for fun...

@recedent

Hello, thanks for suggestion. however, as I have stated, my results are based on custom CFGs and partially moded Q2 version (better looking skins, new HQ sound effects, new crosshairs).

Although, I am fairly sure that it does NOT, have strong impact (i.e. the results shouldn't differ over 10 fps more or less), you shouldn't take mine as 100% accurate, but more or less as a general direction regarding performance.

If you want, I can re-install a vanilla Q2 setup with Capehill's binary and run the timedemos for you.

BTW, I looked at your website with benchmarks and I confirm the results of Q1 SameEp.

@ChrisH

On with my results.

With 800x600 and high quality settings I get 689 frames, 45.0 seconds: 15.3 fps.
With low quality settings, same resolution I get 689 frames, 28.5 seconds: 24.2 fps.
With 640x800 low quality i get 24.4 fps.
With 320x240 low quality I get 24.1 fps
and with 1024x768 low quality 19 fps.

System is Sam440 667Mhz +512 Ram + M9 / OS4.1 u2, Acube's latest updates installed and mGL 2.5.

* Note that by upgrading from mGL 2.4 to 2.5 I got only 0.5 fps more.

Based on my extensive knowledge over Q2, I find the results very awkward.

1) For example, by amending the resolution from 320x240 to 1024x768 you only get 5 fps more. Something that I 've see for the first time in Q2 in any platform that I 've used it.

2) Furthermore, the same moded copy of Q2 with equal settings of CFGs seems to run only 5% more than on P133Mhz MMX, 64MB Ram, TNT 2 16MB PCI, Win 95 (a PC I got from 1995-6).

3) Same copy, same CFGs run in SamFlex 440 800Mhz+ 1GB Ram + 9250 128MB Ram /OS 4.1 u2, Acube's latest updates, mGL 2.5 give the exact same results while on a PegII it doubles the performance.

4) Still, ~ 60fps at a PegII with a G4, 9250 etc, is low. A 2000s equal PC ( in terms of raw power of hardware) almost triple doubles it.

So there is something going on REALLY bad in Q2 port or in the OS itself. Although, I agree with you that SamEp wasn't built as a gaming machine, it should handle 15 years old pretty much fine. The hardware isn't for sure the bottleneck. FS2Open behaves almost flawless in 800x600.

Please download the "low.cfg" from here -> http://db.tt/1KIfMtc .

Save the cfg at PROGDIR:Baseq2.

Run Q2, bring down console and type (without the quotation marks):
"disconnect"
"exec low.cfg"
"timedemo 1"
"demomap demo01.dm2"

I believe that this includes the most balanced commands for Q2 while favoring performance.

Last edited by Cool_amigaN on 25-Jun-2011 at 09:42 AM.

_________________

however
Pentium 2@350 with Radeon 9200SE 64bit 128MB
catalyst 5.8
Windows XP SP3
128 MB

Quake 2
1024x768= 54,4 fps

AmigaOS4 has worst 3D drivers..

Last edited by Seiya on 25-Jun-2011 at 12:38 PM.
Last edited by Seiya on 25-Jun-2011 at 12:38 PM.

_________________

@Seiya

yeah true hope the new system will fix a bunch of crap!

_________________

@hotrod

Quote:

hotrod wrote: @Cool_amigaN Hm... something must've been done wrong with the 3D-drivers then, I was playing the game fine on my old a4k at 640x480 and it works fine on my amigaone. Actually the last time I tried it it didn't, it was jerky and really werid but it used to run fun. The point is that it should be playable on a SAM.

Try a multiplayer match with people who own a pc and you will see that the FPS aint sufficient.
These games need around 60fps to get rid latency.

@thread

According to some x86 benchmarks.

High CPU usage benchmark called crusher should NEVER fall below 24/25 fps to be called playable.

The tests were ran with 3 different systems, a Pentium II 400, a Celeron 300 A and a K6-2 300.

All systems were using:
Hard Drive: IBM DGVS 09U, ultra wide SCSI SCSI adapter: Adaptec 2940 UW Memory: 128 MB PC100 SDRAM Operating System: Windows 98 Refresh Rates: 85 Hz for all cards in all resolutions except 1152x864 and 1600x1200 75 Hz Quake 2 ver. 3.19 Unreal original release version, plus D3D patch for TNT, ver. 209 for Banshee

Motherboards:
Asus P2B for both Slot1 CPUs Asus P5A for AMD CPU

Drivers:
NVIDIA RIVA TNT: 4.10.01.0036, Chip clock 90 MHz, Memory Clock 112.5 MHz 3Dfx Voodoo2: 4.10.01.0180-2.17 3Dfx Voodoo Banshee: 4.10.01.0090-1.00

advisable to play at the highest screen resolution you can get

pretty useless built in benchmark demos 'demo1.dm2' or 'demo2.dm2'

Massive1 benchmark - created by 3 Fingers
http://planetquake.gamespy.com/View.php?view=Quake2.Detail&id=866

Crusher benchmark - created by 3 Fingers
lots of explosions and light weapons - calculations done only by cpu, therefore it is an ideal cpu benchmark.
Ideal worst case fps test
Using 'crusher.dm2' shows how bad the frame rate can get.

Crusher fps test -> worst case fps must be >25

http://planetquake.gamespy.com/View.php?view=Quake2.Detail&id=865


MINIMUM framerate (worst case fps must be >24)

Disable VSYNC
disable dynamic lighting (gl_dynamic 0)
enable flash blend (gl_flashblend 1)
switch off 'polyblend' gl_polyblend 0

Quake 2 v 3.19 With 3Dnow! Enabled

Quake 2
W/intel P2 400 mhz, NVIDIA Riva TNT

demo1, massive1, crusher

460x480,  80, 63, 41
800X600,  63, 55, 40
1024x768, 40, 38, 32
1152x864, 32, 31, 27


W/intel Celeron 300A/p2 300 mhz, NVIDIA Riva TNT

demo1, massive1, crusher

460x480,  58, 45, 29
800X600,  53, 43, 29
1024x768, 37, 34, 27
1152x864, 29, 28, 24

(still ok at 24 fps)


W/AMD k6-2 300 mhz, NVIDIA Riva TNT

demo1, massive1, crusher

460x480,  37, 28, 19
800X600,  36, 27, 19
1024x768, 34, 27, 19
1152x864, 29, 25, 18

18-19 on crusher means minor slowdowns in some cases.


3dfx Voodoo 2 benchmarks

W/intel P2 400 mhz, 3dfx voodoo 2

demo1, massive1, crusher

460x480, 93, 71, 44
800X600, 63, 52, 37

460x480  SLI, 105, 75, 45
800X600  SLI,  98, 73, 45
1024x768 SLI,  75, 62, 43


W/intel Celeron 300A/p2 300 mhz, 3dfx voodoo 2

demo1, massive1, crusher

460x480, 71, 53, 33
800X600, 56, 45, 30

460x480  SLI, 73, 53, 33
800X600  SLI, 72, 53, 33
1024x768 SLI, 74, 50, 33


W/AMD k6-2 300 mhz, 3dfx voodoo 2

demo1, massive1, crusher

460x480, 72, 53, 33
800X600, 58, 46, 32

460x480  SLI, 73, 53, 33
800X600  SLI, 73, 53, 33
1024x768 SLI, 66, 50, 33

Last edited by asymetrix on 25-Jun-2011 at 03:09 PM.

_________________

I continue to be surprised: Some people are acting all shocked that WinXP with heavily optimised 3D drivers can out-perform OS4's rather aging MiniGL/Warp3D system. This is not exactly news! (But it is why Gallium3D is now being ported to AmigaOS4 by Hyperion.)

But those who think this is "crap" should try using a niche OS which doesn't even support hardware acceleration of 3D... Personally I think it's good we have something.

@CoolAmigan
Yes, it is very surprising that 320x256 is hardly any faster than 1024x768. No need for extensive Q2 experience to tell that! Obviously there is some resolution-independant bottleneck. I wonder if this also happens on AmigaOne's and Pegasos2's?

Last edited by ChrisH on 25-Jun-2011 at 03:17 PM.

_________________

@ChrisH

Quote:

ChrisH wrote: I continue to be surprised: Some people are acting all shocked that WinXP with heavily optimised 3D drivers can out-perform OS4's rather aging MiniGL/Warp3D system. This is not exactly news! (But it is why Gallium3D is now being ported to AmigaOS4 by Hyperion.) @CoolAmigan Yes, it is very surprising that 320x256 is hardly any faster than 1024x768. No need for extensive Q2 experience to tell that! Obviously there is some resolution-independant bottleneck. I wonder if this also happens on AmigaOne's and Pegasos2's?

The problem is that even old systems had better performance. I played both Quake 3 and UT with pretty decent frame rate on my old amd k6-2 450@500mhz using a nvidia tnt2 32meg agp card.
The performance of the k6-2 is horrible and a friend of mine using a p1 166@220mhz mmx nearly got same crunching times with seti@home.

I kinda expect that a "modern" amiga system at least bets such a ancient low end system, but sadly that is not the case.

@Tomas

When people start comparing drivers fairly, is when i would be inclined to listen.

as ChrisH said, comparing against drivers made by nvidia/ATI themselves is pointless, as they will always be better. the graph by Rose is pointless for this reason.

benchmark against MorphOS, or other open drivers, then you can see what is possible.

_________________

@Karlos

Not really, there is a real issue with Quake 2 ITSELF, for example Quake3 runs FASTER on my BlizzardPPC then Quake2, i dont even get more then 1 FPS on my BlizzardPPC.

Quake2 on SAM440 ep is faster in software renderer then HW renderer. 23 fps vs 17 here..

I think its because its ported from WOS based sourcecode and we overlooked something.

@Tomas

Quote:

I kinda expect that a "modern" amiga system at least bets such a ancient low end system, but sadly that is not the case.

When only defence for bad results is "highly optimized drivers" when we are talking from flower of 90's tech is quite sad. Putting fingers on ears and yelling "EVERYTHINGS FINE" has never been a good strategy. And this is coming from owner of a SAM.

@Derfs

Quote:

Derfs wrote: @Tomas When people start comparing drivers fairly, is when i would be inclined to listen. as ChrisH said, comparing against drivers made by nvidia/ATI themselves is pointless, as they will always be better. the graph by Rose is pointless for this reason. benchmark against MorphOS, or other open drivers, then you can see what is possible.

Yeah.. but i think comparing a computer from 2011 with one from 90s is still relevant. Even our drivers should be better than this... It should simply be able to beat a computer/gpu from the 90s.

Something is just wrong when next generation amiga is hardly any faster with 3d than a classic ppc system.

And the k6-2 system was even a low end system that is significantly slower per mhz than the first generation pentium. The k6-2 even made a intel celeron system look good.

Last edited by Tomas on 25-Jun-2011 at 04:16 PM.

@Rose

EVERTYHING'S FINE, SHUT UP.

;->

_________________
OpenWindows Initiative. Port PS3 hardware to bananas. For free. Join today and receive expired $50 cupon from AI!

@asymetrix

Quote:

asymetrix wrote: @thread According to some x86 benchmarks. High CPU usage benchmark called crusher should NEVER fall below 24/25 fps to be called playable.

Ah, finally someone brought this up! I was reading a similar article on a gaming magazine around... 1996, LOL.

Some can justify as enjoyable 1 fps other 1000. However, the international standard for a game to be called playable is 24/25. There is no need for tech/gaming knowledge just human nature, 23.967 or simply know as 24 fps is the minimum possible because lower than this, the human brain will notice lag of images on screen. For online gaming, 60 fps is the lowest possible due to latency reasons. Maximum should be around 220 btw. Still, for anyone who had played online FPS games, knows that on intense moments, >60 will dramatically affect your ability to dodge or shoot.

That being said, I cannot achieve 24 fps with Q2 on Sam since the timedemo is straggling to get over the limit, meaning that when playing single missions, depending on the scenery you notice constant frame drops, no matter how much you lower down the quality settings.

Therefore, for the above reasons, I would never recommend Q2 on Sam, and would love to have an official explanation for this problem from a valid source. Personally I suspect that the trouble lies on drivers, but someone has to clarify this. Does the hardware has the problem, the binary of the port or the OS? Because, by no means this is a normal behavior.

Last edited by Cool_amigaN on 25-Jun-2011 at 10:57 PM.

_________________

@Tomas
A K6-2 450 is muuuuuch faster than a p1 166. If youve seen things that suggest ptherwise then the k6-2 system had massive bottlenecks elsewhere. Yes a Celeron beat it, but in those days the Celerons were actually great value for money and not too far removed from high end performance (a celeron 300 clocked at 450 actually marginally outperfromed a p3 450 when the p3 450 was the highest spec machine (unless simd extensions are used). Celerons had l2 cache running at ful clock, even p3's at the time only ran l2 cache at 1/2 clock.)). It was only with Coppermine that Celerons became poor options/performers relative to othre options. Mendocino and earlier Celerons were mostly good choices.


@thread

Just out of sheer curiousity I ran the 680x0 version of quake2 on a 1ghz p3 based celeron, running on amithlon with a 2meg matrox g100 card (ie. system whose raw performance is even lower than that in the Sam boards). At 640x480, in software rendering I get between 28 and 32 fps.

This is a machine that is slower than a Sam, and is running an emulated 68k cpu (ergo, roughly 40% of native cpu speed being available for 68k), using an ooooold 2meg non 3d card running on OS3.x.

The performance numbers people are getting for OS4.x Q2 are horrible, no matter which way you look at it. There's obviously something wrong, and this is excluding 3d driver issues.

Last edited by fishy_fis on 27-Jun-2011 at 05:25 AM.

@eliyahu

Quote:

i'm curious if the performance is better with hyperion's version since the optimizations done by peter gordon.

I didn't really make optimisations to the game itself, i'm no GL expert. What I did was take the existing source, compiled it with the current SDK for OS4.1, fixed the HW renderer to actually work, fixed the windowed mode to work properly (and added mouse grab to it), added the AHI settings to the in-game menu, got the videos to play again, fixed several bugs, and packaged it up with a new installer.

Edit: Oh yeah, I ditched the DLL hack and used shared objects instead.

I think there were some other fixes, but I can't remember offhand.

Last edited by xeron on 27-Jun-2011 at 06:28 AM.

_________________
Playstation Network ID: xeron6

@Cool_amigaN

Listen the game works fine, plays great, plenty fast enough, and was easy to install. I am thankful to Hyperion for making it available, thankful to Acube for making Sams, and thankful to Amigakit for selling to me. Your negativity is uncalled for.

Last edited by JayCee on 27-Jun-2011 at 04:04 PM.

_________________
A1000, A500, A600, A1200, CTDV, A2000, A4000 Towered, SAMFLEX 800mhz

@JayCee

Amiga used to be cutting edge for games, now its not even on par with the other platforms. If you can't play games that are modern spec, what's the point in new hardware? I might as well buy a P3 with a Voodoo 3 on ebay.

_________________
Amiga A1200, 3.1 ROMs, Blizzard 1230 MKIV 64MB & FPU, 4GB DoM SSD, Workbench 3.1

@A1200

Quote:

Amiga used to be cutting edge for games,

Amiga stoped being cutting egie in 1992, too slow CPU, graphics where slow compared to PC's, and no CD-ROM. 3D was always better on PC.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Quake 2 is plenty fast enough on my Apple iBook G4 why would I buy a Sam board to play it?

I'm thinking of buying Bioshock for my latest Intel Mac. Is it as good as the hype suggests?

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios