Click Here
home features news forums classifieds faqs links search
6071 members 
Amiga Q&A /  Free for All /  Emulation /  Gaming / (Latest Posts)
Login

Nickname

Password

Lost Password?

Don't have an account yet?
Register now!

Support Amigaworld.net
Your support is needed and is appreciated as Amigaworld.net is primarily dependent upon the support of its users.
Donate

Menu
Main sections
» Home
» Features
» News
» Forums
» Classifieds
» Links
» Downloads
Extras
» OS4 Zone
» IRC Network
» AmigaWorld Radio
» Newsfeed
» Top Members
» Amiga Dealers
Information
» About Us
» FAQs
» Advertise
» Polls
» Terms of Service
» Search

IRC Channel
Server: irc.amigaworld.net
Ports: 1024,5555, 6665-6669
SSL port: 6697
Channel: #Amigaworld
Channel Policy and Guidelines

Who's Online
30 crawler(s) on-line.
 71 guest(s) on-line.
 0 member(s) on-line.



You are an anonymous user.
Register Now!
 MagicSN:  11 mins ago
 AmigaPapst:  23 mins ago
 pixie:  46 mins ago
 agami:  49 mins ago
 Vidar:  1 hr 47 mins ago
 OlafS25:  1 hr 50 mins ago
 Musashi5150:  2 hrs 6 mins ago
 jPV:  2 hrs 26 mins ago
 amig_os:  3 hrs 45 mins ago
 kolla:  4 hrs 39 mins ago

/  Forum Index
   /  Amiga General Chat
      /  How good or bad was the AGA chipset in 1992/1993.
Register To Post

Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 Next Page )
Poll : How good or bad was the AGA chipset in 1992/1993.
10p Excellent (Best at 2D/3D, colors, and resolution, frame rate etc.)
5p Good / better than most computer.
0p Barely hanging in there.
-5p Below average / slow but usable
-10p useless / horrible
 
PosterThread
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 8:10:40
#221 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3646
From: Germany

@Gunnar

Quote:

Gunnar wrote:

Quote:
Sure: with a HORRIBLE and SLOW patch to access / set the additional color registers sizes.


I think its a clever solution.

Only for someone which likes to add a quick and dirty patch to the project which requires a very small implementation cost.
Quote:
And work greats and is not slow if you load many colors.
Only if you want to change a single colors like with copper bar then yes its a bit awkward to use.
On SUPER-AGA you can do a 24bit color load with a single 32bit Copper MOVE.
This makes it more convenient.

In fact this should have been made.
Quote:
Quote:
With another HORRIBLE and SLOW patch to access / set the new color registers.


I think its a clever solution. And work greats and is not slow if you load many colors.
Only if you want to change a single colors then its awkward.

Same as above.

BTW, AGA already had internally those 256 x 32-bit CLUT registers. It should have been enough to expose / map them at $DFF400 (for example). And leaving the existing ones catching the writes and automatically converting them to the corresponding 32-bit value.

Plus the 32-bit Copper, of course. Or, even better, a CLUT DMA, as I've already reported around 10 years ago on Olaf's forum.
Quote:
Quote:
Yes, but not a quantum leap, since the Blitter remained the same: 16 bit and 7Mhz.


Actually the Amiga Blitter runs at 3.5 MHz.

Gunnar, Gunnar, Gunnar...

https://amigadev.elowar.com/read/ADCD_2.1/Hardware_Manual_guide/node012A.html

The system clock speed for NTSC Amigas is 7.16 megahertz (PAL Amigas 7.09
megahertz). The clock for the blitter is the system clock.

Quote:
I found the 64bit sprites very cool. One could use them very nicely for a background playfield. The show sprite twice per row feature was handy too.

Actually this is one of the few useful case.

Otherwise 64 pixels sizes are good only when you have much larger displays: hires (640) or, even better, super-hires (1280) horizontal resolutions.

At lores it's much better to have sprites with 8 or 16 pixels horizontally.
Quote:
An OCS game with 5 planes (32 color) had about 37% DMA available for the Blitter.

No, it had more. You made a mistake on the calculation because you made a wrong assumption.

Here's the reference: https://amigadev.elowar.com/read/ADCD_2.1/Hardware_Manual_guide/node012B.html

The calculation isn't that easy / straightforward and requires multiple things to be taking into account to get a realistic / real-world value.
Quote:
An AGA game with 8 planes (256 color) had about 75% DMA available for the Blitter. (fmode 3)

See above: it's not the case.
Quote:
You need to blit more for 8planes but this was possible as you had more free slot.
So this worked fine.
The extra available DMA slots allowed you to do the same game with the Blitter but now in 256 colors instead 32.

128 colors is more realistic.

However and assuming that your numbers are correct (and they are not), from 32 to 256 colors it means that the Blitter "improved" (SIC!) by 60%.

Now compare it to the 300% improvement of the display controller when it got 64-bit fetches...
Quote:
I think AGA was really an improvement over OCS.

It's for sure. However it was very modest. After 7 (SEVEN) years...
Quote:
But I also think that not many games used AGA to the maximum

Possible, like not many games pushed OCS/ECS to the maximum (me neither. but for practical reasons that I've explained on the other thread).

@bhabbott

Quote:

bhabbott wrote:
@Gunnar

Quote:

Gunnar wrote:

An OCS game with 5 planes (32 color) had about 37% DMA available for the Blitter.
An AGA game with 8 planes (256 color) had about 75% DMA available for the Blitter. (fmode 3)
You need to blit more for 8planes but this was possible as you had more free slot.
So this worked fine.
The extra available DMA slots allowed you to do the same game with the Blitter but now in 256 colors instead 32.

Thanks for putting numbers on it to show that even though the blitter was still 'only' 16 bit, AGA could significantly improve its performance. I knew intuitively that this was true but didn't know just how large the difference could be.

Not the case: see above.
Quote:
It's a pity that this fact was not made clear to users when AGA was released, as it might have changed their minds about Commodore's decision not to make the blitter 32 bit (which probably was impossible anyway with the technology and resources they had).

A 32-bit Blitter was needed and was feasible. Even a 64-bit one, considered that the fetch was extended to 64-bit

However if Commodore smart engineers didn't wanted to improve the Blitter this way, they had a very simple way to improve its speed: increase its clock frequency. Providing an equivalent (and different!) fmode register for selecting the frequency to keep the backward-compatibility.

I think that from 7 to 28Mhz in 7 (SEVEN) years should have been possible. If you take into account how much the chips frequencies increased over that long period.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 8:18:20
#222 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3646
From: Germany

@bhabbott

Quote:

bhabbott wrote:
@Hypex

Not quite. CIAs didn't need to be speeded up for HD floppies. Paula MFM decoding could easily have been speeded, but more DMA slots would have to be made available - with bigger buffers for it. Still doable, but not quite as simple

You don't need extra time slots: "just" add a fmode like registers for controlling the fetch width for disk accesses.

So you can fetch 32 or even 64-bit data with the same available slots.

Exactly like that was done with the bitplanes and sprites.
Quote:
and possibly breaking compatibility.

Only for supporting 2.88MB disks: this would have required an extra register for extending the disk length.
Quote:
What good have 16 bit have done? A stock A1200 doesn't have the CPU power to decode MP3's or the memory to store large samples. If you want CD quality music then use a CDROM drive. Amiga MODs were always 8 bit - and sounded great. 8 track MODs generally don't add much. If you really wanted more then cards were available, but few people bought them because - unlike PCs - every Amiga has a 'sound card' built in.

Nevertheless, the existing 4 x audio channels could have improved exactly in the same way of bitplane and controllers.

This would have brought 16-bit PCM and up to 56Khz frequencies.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 8:34:48
#223 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3646
From: Germany

@bhabbott

Quote:

bhabbott wrote:

The original Amiga (aka everything OCS with a 68000) was designed so that the chipset could access the RAM without blocking much of the CPU access.

In theory yes, and if you run the o.s.. For games usually it was much better to let the Blitter use all the available memory slots.
Quote:
Quote:
But the topic is AGA,

AGA continued the philosophy of OCS, offloading graphics operations from the CPU to free it up for other things. When you put Fast RAM in an A1200 this really shows up, but even with just Chip RAM it's noticeable. Everything runs smoother because the blitter and CPU are working in parallel.

With the o.s.?

Because for games it wasn't a game changer: there was a small boost only when copying some rectangular regions at 32-bit boundaries.

 Status: Offline
Profile     Report this post  
Hammer 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 8:56:22
#224 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5268
From: Australia

@matthey

Quote:
What Amiga was most desirable in 1992 for the base model...
A) Amiga 600 68000@7MHz ECS 2MiB $499 (Amiga to C64 rebirth CBM thought we wanted)
B) Amiga 1200 68020@14MHz AGA 2MiB $599 (actual best seller)
C) Amiga ? 68030@21MHz AA+ 4MiB $649 (potential AA+ spec and price)
D) Amiga ? 68040@28MHz AAA 8MiB $1199 (32 bit AAA no VRAM)

There's very little IPC difference between 68030 and 68020.

AGA install base building should have started from Xmas 1991.

C) 1992, A1200 with 68EC020-25 (25 Mhz) and apply minor overclock to 28 Mhz like today's PC GPU OC AIB vendors. This SKU includes Fast RAM. This SKU competes against fast 386DX-33-based gaming PCs.

386DX-25 runs Wing Commander VGA pretty well and this is the entry machine for free Doom shareware taster.

Organize time exclusive game bundle deal that uses 68EC020 @ 28 Mhz.

Amiga's out-of-the-box Wing Commander bundle is an embarrassment with stock Amiga 1200 since the performance was crap.

Nintendo's Wing Commander SNES release was supported by an out-of-the-box SuperFX CPU accelerator in the game cartridge.

Amiga may have a separate Wing Commander + 68EC020 @ 28 Mhz accelerator bundle SKU for stock Amiga 1200.

D) 1993, A1200 with 68LC040-25 and Fast RAM. This SKU competes against Apple's Quadra 605 with 68LC040-25 with a $1000 USD price tag and comparable priced 486SX-25/486SX-33-based gaming PCs.

These "official" accelerated A1200 SKUs close the gap between entry-level A1200 and A4000/030 (68EC030/68882).

Doom-type games don't use FPU.

My approach is to mitigate user base bleed to the PC and SNES platforms and buy enough time for Amiga Hombre.

Commodore's ECS wasn't competitive against Amiga add-on graphics cards with PC SVGA chipsets e.g.
EGS 28/24 = Cirrus Logic GD5426
Retina = NCR 77C22E+
Visiona = IMS G300C
Piccolo = Cirrus Logic GD5426
Piccolo SD64 = Cirrus Logic GD5434
Rainbow II = Analog Devices ADV7120
Rainbow III = Inmos G365
Domino = Tseng Labs ET4000
Merlin = Tseng Labs ET4000W32
Picasso II = Cirrus Logic GD5426

Tseng Labs ET4000AX ISA comes in either 32-bit DRAM or fast 16-bit VRAM.

Tseng Labs ET4000AX ISA with 32-bit DRAM smashed Commodore's ECS into the ground.


_________________
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)

 Status: Offline
Profile     Report this post  
Hammer 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 9:17:45
#225 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5268
From: Australia

@cdimauro

Quote:

Again? You already provided that link and I've already replied you. Why continue to bring complete non-sense?

Here's what the video reports:
Machine: Amiga 1200, 68020, 64 MB Fast RAM (emulated)

And looking at two videos with the real machine:
Machine: Amiga 4000T, 68040/25 MHz, 12 MB Fast RAM

I've no problem producing something similar with such massive specs. Even by using MAME on such machines and getting the original Final Fight running...


Apple released Quadra 605 with 68LC040-25 for $1000 USD in October 1993.


https://www.youtube.com/watch?v=A39vwT2w5po
Real Amiga 4000T, 68040 @ 25 MHz, 12 MB Fast RAM was running Amiga Sprite Engine "Final Fight" at 60 fps.

Cost-reduced hardware with 30 fps version is good enough.

From Amiga Sprite Engine developer
A 68EC020 with 8 meg Fast RAM is enough to run it (in fs-uae at least). But note that this is an old .exe that's the same as the one used for the “emulated” video, except there's no audio. An up-to-date version of this demo would use less RAM.


Commodore's A3640 wasn't the fastest 040 accelerator due to 030 bus translation.

Last edited by Hammer on 09-Oct-2022 at 09:22 AM.

_________________
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)

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 9:20:35
#226 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3646
From: Germany

@Hammer

Quote:

Hammer wrote:
@cdimauro

Quote:

Again? You already provided that link and I've already replied you. Why continue to bring complete non-sense?

Here's what the video reports:
Machine: Amiga 1200, 68020, 64 MB Fast RAM (emulated)

And looking at two videos with the real machine:
Machine: Amiga 4000T, 68040/25 MHz, 12 MB Fast RAM

I've no problem producing something similar with such massive specs. Even by using MAME on such machines and getting the original Final Fight running...


Apple released Quadra 605 with 68LC040-25 for $1000 USD in October 1993.

Irrelevant. Usual padding...
Quote:
https://www.youtube.com/watch?v=A39vwT2w5po
Real Amiga 4000T, 68040 @ 25 MHz, 12 MB Fast RAM was running Amiga Sprite Engine "Final Fight" at 60 fps.

A cost-reduced 30 fps version is good enough.

Let's see the needed hardware.
Quote:
From the developer

A 68EC020 with 8 meg Fast RAM is enough to run it (in fs-uae at least). But note that this is an old .exe that's the same as the one used for the “emulated” video, except there's no audio. An up-to-date version of this demo would use less RAM.


Do you understand that FS-UAE is an emulator, and fast ram on it works much faster than a real one?

 Status: Offline
Profile     Report this post  
Hammer 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 9:26:01
#227 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5268
From: Australia

@cdimauro

Do you understand Commodore's A3640 card wasn't the fast 040 accelerator since it's coupled with mediocre 25 Mhz 030 bus translation and mediocre 25 Mhz Fast RAM?

Last edited by Hammer on 09-Oct-2022 at 09:28 AM.
Last edited by Hammer on 09-Oct-2022 at 09:27 AM.

_________________
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)

 Status: Offline
Profile     Report this post  
Karlos 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 9:42:23
#228 ]
Elite Member
Joined: 24-Aug-2003
Posts: 4402
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@Hammer

Commodore couldn't even solder the caps on the 3640 the right way round, either. It wasn't their finest product.

I think it was the Warp Engine that took the same 040, ran it at 28MHz with local fast memory on board that utterly ran rings around the 3640 in benchmarks at the time.

The AGA machines had a knack of hobbling the CPU performance at launch.

Last edited by Karlos on 09-Oct-2022 at 09:44 AM.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 10:01:55
#229 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3646
From: Germany

@Hammer

Quote:

Hammer wrote:
@cdimauro

Do you understand Commodore's A3640 card wasn't the fast 040 accelerator since it's coupled with mediocre 25 Mhz 030 bus translation and mediocre 25 Mhz Fast RAM?

Again: totally irrelevant.

The most sold Amigas where the 500, 600 and 1200.

So THOSE where THE reference for the software houses (the 500 bundled with at least 512kB expansion).

Understood?

 Status: Offline
Profile     Report this post  
kolla 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 10:06:30
#230 ]
Elite Member
Joined: 21-Aug-2003
Posts: 2880
From: Trondheim, Norway

@Karlos

SCSI cards were readily available, and the A4000T came with SCSI onboard. The A4000 was low cost, and went even more low cost with the CR revision with onboard 68EC030 and semi-broken zorro - made for SCALA kiosks.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

 Status: Offline
Profile     Report this post  
Hammer 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 10:17:02
#231 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5268
From: Australia

@Karlos

From https://www.amigareport.com/ar227/p1-8.html
or all of you who think benchmarks are important, here they are
(courtesy of AIBB v6.1 - compared to stock Amiga 4000/040):

EmuTest 1.88 Writepixel 1.47 Sieve 2.87 Dhrystone 1.61
Sort 1.76 EllipseTest 1.12 Matrix 2.25 IMath 1.61
MemTest 4.61 TGTest 1.15 LineTest 1.02 Savage 1.63
FMath 1.61 FMatrix 2.80 Beachball 1.71 InstTest 2.41
Flops 1.60 TranTest 2.22 FTrace 1.69 CplxTest 1.71

Sysinfo (version 3.23) reports 29.89 MIPS and 7.58 MFlops.

Most startling in the above benchmarks is the memory speed of the
Warp Engine. With the 60ns 16MB SIMM module, the Warp Engine's RAM speed is
over 4.5 times faster than a stock 4000

...
Real world performance is the only benchmark in my book, however. I
am running a 16 colour Workbench that feels like a 4 colour Workbench.
Icons and windows just fly onto the screen. Response from the Amiga is
instantaneous. Click on the close gadget of a window, and it's gone before
you can blink. Screens open faster and programs load much quicker. Boot-up
time was reduced by almost 10 seconds over a stock 4000 with an IDE hard
drive. I can run the Emplant Macintosh emulator in 256 colours with
absolutely no slowdown. PageStream 2.2 screen updates are so much faster
with the Warp Engine 4040, it's a dream to use.

------------

Warp Engine's RAM speed is over 4.5 times faster than a stock 4000.


_________________
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)

 Status: Offline
Profile     Report this post  
Gunnar 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 10:30:20
#232 ]
Regular Member
Joined: 25-Sep-2022
Posts: 477
From: Unknown

@cdimauro

Quote:
Quote:

Actually the Amiga Blitter runs at 3.5 MHz.


Gunnar, Gunnar, Gunnar...
[/quote]

My friend Cesare Di Mauro,

The AMIGA has many clocks, 28Mhz, 14MHz, 7Mhz, 3.5Mhz
Even with the 28MHz clock effectively the Amiga DMA system runs on 3.5 MHz. The Blitter does all its "work" on these 3.5 clock event. The Blitter is part of AGNUS Amiga chip. AGNUS has all clocks connected.

Quote:

The system clock speed for NTSC Amigas is 7.16 megahertz (PAL Amigas 7.09
megahertz). The clock for the blitter is the system clock.



The Blitter needs for each operation:
8 clock at 28.0 MHz
4 clock at 14.0 MHz
2 clock at 7.0 MHz
1 clock at 3.5 MHz


Quote:

Quote:
An OCS game with 5 planes (32 color) had about 37% DMA available for the Blitter.

No, it had more. You made a mistake on the calculation because you made a wrong assumption.

Here's the reference: https://amigadev.elowar.com/read/ADCD_2.1/Hardware_Manual_guide/node012B.html

The calculation isn't that easy / straightforward and requires multiple things to be taking into account to get a realistic / real-world value.


The DMA calculation depends on many factors,
how big is the Copperlist, does the game use Overscan?
There is never only one correct result - its game depending.
As ballpark result my numbers are good.




The whole discussion what in theory could have been done to make an even better on AGA is mote.
Please not forget that "AGA was a stop gap fix". As the real planned solution was not ready.
And as a stop gap solution AGA was a real improvement.

Look at this like you invite your mother in law for a "Thanks Giving Dinner".
Unluckily your Turkey is burned.
You have only 10 minute time to get something on the table - so you make Spaghetti.


Under the circumstance and the time they had to get AGA done. In my opinion AGA was a very good solution. Is easy to put an infinite wishlist of extra features together - but you miss the point.
Extras wishes all take more time.
If they would have a lot more time, then they would not have done Spaghetti in the first place - they would have roasted another Turkey.

 Status: Offline
Profile     Report this post  
Hammer 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 10:34:52
#233 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5268
From: Australia

@cdimauro

Quote:
Again: totally irrelevant.

The most sold Amigas where the 500, 600 and 1200.

Your A500 and A600 remark is irrelevant to this topic.

Why don't you look at AIBB's MemTest benchmark for A1200-NF and A4000/040?


http://amiga.resource.cx/perf/aibb.pl?amiga=1200&testgroup=int&testgroup=float&order=mem&ref=a1200

MemTest
Amiga 1200/NF scored 1.0 (reference)
Amiga 4000/040 with A3640 scored 1.23

VS

A1200's Fast RAM add-ons and 68020 CPU accelerators
DKB 1202 with just stock 68020 @ 14 Mhz and DKB s Fast RAM design scored 2.17
Blizzard 1200 with just stock 68020 @ 14 Mhz and Phase 5's Fast RAM design scored 2.17
Blizzard 1220 with just 68020 @ 28 Mhz and Phase 5's Fast RAM design scored 3.63

ACA 1220 with 68020 @ 20 Mhz with ACA's Fast RAM design scored 3.04
ACA 1220 with 68020 @ 33 Mhz with ACA's Fast RAM design scored 4.22


My WinUAE's A1200 with Fast Memory expansion approximation


When comparing my WinUAE's A1200 with Fast Memory expansion approximation, ACA 1220's Fast RAM implementation is FASTER

Last edited by Hammer on 09-Oct-2022 at 10:51 AM.
Last edited by Hammer on 09-Oct-2022 at 10:50 AM.
Last edited by Hammer on 09-Oct-2022 at 10:48 AM.
Last edited by Hammer on 09-Oct-2022 at 10:41 AM.
Last edited by Hammer on 09-Oct-2022 at 10:36 AM.

_________________
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)

 Status: Offline
Profile     Report this post  
Gunnar 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 10:55:25
#234 ]
Regular Member
Joined: 25-Sep-2022
Posts: 477
From: Unknown

Hello Hammer,

this time no photo of INTEL chips but an emulated AMIGA?
Well done, nice improvement my friend!



Quote:

Why don't you look at AIBB's MemTest benchmark for A1200-NF and A4000/040?


http://amiga.resource.cx/perf/aibb.pl?amiga=1200&testgroup=int&testgroup=float&order=mem&ref=a1200

MemTest
Amiga 1200/NF scored 1.0 (reference)
Amiga 4000/040 with A3640 scored 1.23

VS

A1200's Fast RAM add-ons and 68020 CPU accelerators
DKB 1202 with just stock 68020 @ 14 Mhz and DKB s Fast RAM design scored 2.17
Blizzard 1200 with just stock 68020 @ 14 Mhz and Phase 5's Fast RAM design scored 2.17
Blizzard 1220 with just 68020 @ 28 Mhz and Phase 5's Fast RAM design scored 3.63

ACA 1220 with 68020 @ 20 Mhz with ACA's Fast RAM design scored 3.04
ACA 1220 with 68020 @ 33 Mhz with ACA's Fast RAM design scored 4.22


I agree that memory performance is very important.
BTW
The 68080 accelerators score 99.4 in the AIBB fastmem test to the base of A1200

The 68080 accelerators score 272.4 in the AIBB fastmem test to the base of A600


Yes, memory speed is a big factor for total system speed.
Yes, the 3640 sucks here a lot.
The 3640 was done hastily. The 3640 is a 100% copy as a reference design done by Motorola, made as example to show how a bus translation could be made.

3640 and 3660 limit the CPU a lot.

Last edited by Gunnar on 09-Oct-2022 at 10:58 AM.

 Status: Offline
Profile     Report this post  
Hammer 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 10:58:37
#235 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5268
From: Australia

@Gunnar

Quote:
ook at this like you invite your mother in law for a "Thanks Giving Dinner".
Unluckily your Turkey is burned.
You have only 10 minute time to get something on the table - so you make Spaghetti.


Under the circumstance and the time they had to get AGA done. In my opinion AGA was a very good solution. Is easy to put an infinite wishlist of extra features together - but you miss the point.
Extras wishes all take more time.
If they would have a lot more time, then they would not have done Spaghetti in the first place - they would have roasted another Turkey.



Commodore created two 256-color display capable chipsets i.e. C65 (completed in December 1990) and AGA (completed in March 1991).

Commodore made spaghetti and marinara pasta.

_________________
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)

 Status: Offline
Profile     Report this post  
Gunnar 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 11:00:48
#236 ]
Regular Member
Joined: 25-Sep-2022
Posts: 477
From: Unknown

@Hammer

Quote:

Commodore created two 256-color display capable chipsets i.e. C65 (completed in December 1990) and AGA (completed in March 1991).

Commodore made spaghetti and marinara pasta.


Where these the same developers?

Big companies have many teams working on different solutions in different areas.

 Status: Offline
Profile     Report this post  
kolla 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 11:05:12
#237 ]
Elite Member
Joined: 21-Aug-2003
Posts: 2880
From: Trondheim, Norway

@Hammer

Quote:

Hammer wrote:
@kolla

Quote:

kolla wrote:
@Gunnar

At least the A1200 motherboard has space for a real FPU…

It's nearly pointless for games.


I was joking. The A1200 motherboard does indeed have space for FPU and all you need to do is to solder one on and it works. But my tongue-in-cheek remark was more about all the hoopla back and forth about the Cyclone 3 used on Vampire V2 cards not having space for FPU… REAL Amiga had at least space for it.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

 Status: Offline
Profile     Report this post  
Hammer 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 11:39:13
#238 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5268
From: Australia

@Gunnar

Quote:

Gunnar wrote:
Hello Hammer,

this time no photo of INTEL chips but an emulated AMIGA?
Well done, nice improvement my friend!

I agree that memory performance is very important.
BTW
The 68080 accelerators score 99.4 in the AIBB fastmem test to the base of A1200

The 68080 accelerators score 272.4 in the AIBB fastmem test to the base of A600


Yes, memory speed is a big factor for total system speed.
Yes, the 3640 sucks here a lot.
The 3640 was done hastily. The 3640 is a 100% copy as a reference design done by Motorola, made as example to show how a bus translation could be made.

3640 and 3660 limit the CPU a lot.

Intel? You got the wrong person. You're using many Intel products, not me.

For raster operations, memory performance is very important.

Depending on the frame's compositing complexity, compute power is also important.

A3640 sucked when it used A4000's semi-Fast RAM.
----

For modern example via Xbox One's 32 MB eSRAM with 1080p render overspill


In this example, 32 MB eSRAM wasn't enough for 1080p render, hence spilling into DDR3.

NAV21 has a 128 MB Infinity Cache that is coupled with delta color compression (DCC).

Based on XBO's frame buffer usage vs 32 MB eSRAM, NAV21's 128 MB Infinity Cache with DCC can support 4K (3840x2160) frame buffer. NAV21's raster performance is competitive against NVIDIA's GA102's 384-bit GDDR6X-19000.

NAVI31 has 192 MB Infinity Cache with 384-bit GDDR6-18000 bus. AMD allocates very high memory bandwidth for the frame buffer.

For retro hardware
1997-era hardware, NVIDIA RIVA 128's 128-bit memory @ 100 Mhz has 1.6 GB/s memory bandwidth.

1998-era hardware, NVIDIA TNT's 128-bit memory @ 110 Mhz has 1.76 GB/s memory bandwidth.

From the retro to the modern era, memory bandwidth is very important for raster operations and it's double important for shared memory designs with APU and SoC.

Vampire's high memory bandwidth for the retro hardware market segment is good. I agree with Vampire's highly effective memory bandwidth approach.

PiStorm with Emu68's AIBB MemTest scored 350.06 when compared to A1200-NF

Last edited by Hammer on 09-Oct-2022 at 12:16 PM.
Last edited by Hammer on 09-Oct-2022 at 12:14 PM.
Last edited by Hammer on 09-Oct-2022 at 12:13 PM.
Last edited by Hammer on 09-Oct-2022 at 12:12 PM.
Last edited by Hammer on 09-Oct-2022 at 11:48 AM.
Last edited by Hammer on 09-Oct-2022 at 11:44 AM.

_________________
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)

 Status: Offline
Profile     Report this post  
Hammer 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 11:42:21
#239 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5268
From: Australia

@Gunnar

Quote:

Gunnar wrote:

Where these the same developers?

Big companies have many teams working on different solutions in different areas.

Two teams, hence the wasted resources.


_________________
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)

 Status: Offline
Profile     Report this post  
Hammer 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 9-Oct-2022 11:57:32
#240 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5268
From: Australia

@cdimauro

Quote:

cdimauro wrote:


So THOSE where THE reference for the software houses (the 500 bundled with at least 512kB expansion).

Understood?

Nope, Mehdi Ali told them to f__koff and these developers jumped on the Sony Playstation bandwagon a few months later. The time period was around 1992 to 1993.

Mehdi Ali is a walking disaster for 3rd party developers' relations.

Last edited by Hammer on 09-Oct-2022 at 11:58 AM.

_________________
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)

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 Next Page )

[ home ][ about us ][ privacy ] [ forums ][ classifieds ] [ links ][ news archive ] [ link to us ][ user account ]
Copyright (C) 2000 - 2019 Amigaworld.net.
Amigaworld.net was originally founded by David Doyle