Commodore Free 35/Commodore_Free_Issue_35_2009_Commodore_Computer_Club.d64/natami part 1

Text File | 2023-02-26 | 12KB | 455 lines

*************************************
               NATAMI
*************************************
   http://www.natami.net/qa.htm

The NatAmi Project

This hardware project is dedicated to
the still innovative system
architecture of the Commodore AMIGA
computer.

The basic concept is straight: Get
the original Amiga design up-to-date.

The NatAmi approach is to rebuild the
original system
- without emulation
- without abstraction layers
- without legacy drawbacks

This means the system needs to be
built from scratch. It will not use
standard PC components when their
adaption or implementation causes
workarounds which limit performance,
functionality or usability. The
priority is to build a very simple
system design which is predictive,
easy to use and fast responding like
the Amiga was/is.

The original creators put great
effort and many thoughts into how to
keep things simple. The NatAmi is
intended to keep this guideline as
the foundation of its system.

This concept is the project name:
Native Amiga

The NatAmi does not revive or copy an
A500/A4000. It will be a succeeding
Amiga model not done by emulation -
it is done by reconstruction.
Up-to-date and good performing
components are used to retain the
efficiency. It will rely on the
original philosophy - but remove old
limitations.

The NatAmi is an AMIGA compatible
machine, allowing you to natively run
original Commodore Amiga software

Natami Design Layout

In a consumer version the NatAmi is
intended to be a very affordable
Amiga successor.

An Amiga that is powerful enough to
be useful for today's typical tasks.
The major long-term intentions for
NatAmi are

- being an Amiga compatible design
- have the original AmigaOS, binary
  OS replacements and Amiga
  applications running
- enable playing new and classic
  Amiga games
- expand features to 24bit truecolor
  gfx and 16bit audio on custom chip
  register level
- implement USB, Ethernet and
  flashdisk support
- provide immediate system on/off and
  hibernation
- provide enough resources and power
  to surf the internet and watch
  online videos
- add features to be able to playback
  DivX, Xvid and DVD

We think that in this area the
AmigaOS still has some major
advantages. The original AmigaOS is
very efficient. The combination of
this very efficient OS and the usage
of the powerful SuperAGA, will result
in a very swift computer.
The Natami will not be a competitor
to a Wintel Desktop machine, nor will
it be a PlayStation 3 killer.

NatAmi Roadmap

Our road-map currently is:

MC68060 CPU board
NatAmi developer board
NatAmi retail board

The Natami dev-board has two main
purposes.
- Allowing early developers to
  develop/port software to Amiga.
- Allowing more people to verifying
  the SuperAGA chipset to be error
  free.

When the dev-boards are out we will
focus on three things:
- Designing the cost reduced Natami
  version.
- Developing our Next Generation 68K
  Core
- We will add new features to the
  SuperAGA chipset

NATAMITechnical FAQ

COMMODORE FREE
The natami team were very keen for an
interview, however they were rather
busy so for the moment I just took
the FAQ with their permission, this
answers 90% of my questions. Is it
just me or are there just too many
exciting projects, I hope this one is
successful and I would be very keen
to get my hands on one of these
machines. I love the Classic Amiga
but (now I am going to get more hate
mail) I am not to keen on where the
New Amiga is currently going!

CPU

Q. Is the Natami team still
evaluating the ColdFire CPU or don't
you expect any performance gains by
using a V5-core?

The 400 MHz ColdFire V5 is certainly
a very powerful CPU. To be able to
get the ColdFire we would need to
"bake" chips. Baking chips is not a
target that we can reach without help
as require a major investment. If we
"bake" our own chips we would of
course have the option to bake our
own 68K_CPU design. If we bake our
own "softcore" into a chip then we
could reach very high clock rates.
Our softcore would then provide
better compatibility than the
ColdFire.

The ColdFire is 68k compatible, but
to get maximum performance out of the
ColdFire for AmigaOS, some patching
of Kickstart and Workbench will be
mandatory. By going for our own
N68070 design we will save any
patching and time.

Q. Do you think you can break 1ghz
with ColdFire or 68k you are talking
about?

No, the 1 GHz barrier will not be
broken by these chips. The New 68K
chips that we are developing will be
in the range of 100-166 MHz inside
the FPGA.

Q. Which is better? A real 68060 CPU
or the new N68070?

The 68060 which runs at a max of
90-100 MHz is the only one in
production and Freescale can stop
production at any time. 68060 is
Expensive and has no memory
controller and old 3.3 voltage use.

The "Self-Made" 68K provides a lot of
advantages:
- Cost reduction
- Possible to add new faster and
  cheaper DDR memory.
- Higher clockrate of 133-166 MHz
  inside the FPGA.
- Memory latencies will be reduced.
- Powerful multimedia instructions
  could be used.
- The Natami-68070 could be made to
  be a lot more compatible with 68000
  games. Getting it to be more
  compatible to the 68060 is simple.

Q. Will Natami use a 64 bit 68070
processor?

64-Bit is no advantage for AmigaOS as
it wasn't designed for any more
memory. For integer the "golden" mix
of best performance 32bit is still
the best. It will make sense to add a
wider SIMD unit (64-128 Bit) A 64-bit
CPU will only bring one advantage:
more memory than 4GB. All structs and
pointers in AmigaOS are designed for
a 32bit CPU. A 64-Bit CPU brings many
disadvantages over a 32-Bit CPU. What
we should do is add a 64-Bit wide
databus as this will increase the
performance.

Q. Does the Natami processor need
cooling?

The 68060 runs perfectly fine passive
cooled. The new 68060 revision 6 that
the Dev-boards have, are produced in
a smaller structure; needing even
less power than the 68060 used in the
old Amiga cards. None of the Natami
chips require even a passive heatsink

Q. Is a Memory Management Unit
planned for your new 68K chip?

No, not for the first CHIP
generation. If you want a MMU; you
can use the 68060 CPU card, There are
many good reasons not to include a
MMU: A MMU adds latency to all memory
accesses. This extra latency will
always be there, even if the MMU is
not used. A chip design can try to
hide the latency in a longer
pipeline. Without such a latency the
chip will always be faster. The
performance loss caused by the MMU
can be quite heavy. For the silicon
that a fast MMU would need, adding
something like ALTIVEC into the CPU
would be a better choice. The 68060
developer boards are capable to boot
NetBSD or Linux if they wish to do
so. The great feature that Amiga
users are so proud of, is to have HW
acceleration for everything and by
design just cannot be used under
Linux.

Q. Don't you need to create a new
architecture for ColdFire or is it
the same because the ColdFire is
compatible to the 68060?

We did some research in this area,
and we came to the conclusion; that
it's possible to build our own 68k
CPU that we could include in our
FPGA. We found out that we could
build a new 68k CPU that is even more
compatible to older 68000 Software,
than the 68040 and 68060 CPU. This
new 68k CPU is more powerful than the
68040 and 68060. The combination of
these advantages makes the Self-Made
68k CPU (N68050/070) very attractive
for us. I believe that our design
will be ideal for the next Natami
design. We will ship the Natami60
board with a 68060 CPU. Later when
our 68k FPGA CPU is finished the
Natami60 board could install it into
their FPGA with a simple firmware
upgrade.

Q. How does Natami compare to PPC
accelerators used in 'classic'
configurations?

A 100 MHz 68060 has a similar
performance as a 100 MHz PPC. Our
goal is that the Natami will be
powerful enough for DVD/Xvid/DivX
playback.

FPGA

Q. How much can be gained by
transferring the FPGA to a custom
fixed function chip ?

Oversimplified answer:
A - Medium price FPGA  100-200 MHz
B - Very expensive FPGA  250-500 MHz
C - Custom Chip based on little
    reworked FPGA code  500 MHz
D - Custom Chip complete redesigned
    500 MHz - Several GHz

But creating custom chip (ASIC) does
cost a lot of money.
Option C costs some money but less
than a million.
Option D is too expensive for the
Amiga market.
For the best price/performance mix we
target solution (A) for now.

Q. Will the FPGA be possible to have
more 68070 or SuperAGA cores ?

AmigaOS is designed to run very good
with a 25 MHz CPU. The AmigaAGA
chipset is running at 3.5 MHz and was
16 Bit. In an effortable FPGA,
SuperAGA could be running in 100-200
MHz and in 32 Bit. SuperAGA Blitter
is thereby 100-200 times as fast as
the Amiga AGA.

The original Amiga always was a
multicore design with :
- the GFX Chip with sprite features.
- the blitter
- the copper
- the Audio DMA
- the CPU
- In addition to this Natami has 3D
  core with texture engine

AmigaOS is perfectly designed to make
use of the Amiga hardware features.
Natami with SuperAGA is the only
chipset worldwide making this again
possible.  PC GFX Card were never
used by any OS as efficiently as the
AmigaOS did with the Amiga chipset.
With SuperAGA and the N68070
softcore, the Amiga Workbench will
run on Natami many times faster than
it currently does on WinUAE using a
fast PC. A multicore CPU is not
needed.

Q. Regarding sprites - Have these
been enhanced? Size? Number? Etc...

Sprite capabilities are unchanged and
are the same as AmigaAGA. Using
sprites make still sense for a mouse
pointer. 8 Sprite channels with max
64 pixel width each and max 16
colours In today games BOBS have
replaced sprites. Bobs are images
that could be copied onto the
background image.

Bob support is as follows:
- 16 Million colours max
- Unlimited size and number
- Copy, rotate and scale
- New ColorKey bob don't need a
  Blitting-Mask

The SuperAGA blitter has the power to
blit the whole screen thousand times
per second. The SuperAGA supports
Gouraud shading, Z-Buffer,
Bilinier-Interpolation.

Q. Why is Natami not cycle exact?

Cycle exact was never needed on
Amiga. The Amiga OCS, ECA and AGA is
16bit, 3.5MHz cycle frequency.
Denise was fetch bitplane and sprite
data using 16bit.Lisa could fetch
bitplane and sprite data in a 32bit
reads.

Natami is running (worst case) 32bit
100-150MHz cycle frequency while
still maintaining software
compatibility.
There never was something like 100%
cycle exact on Amiga. Different Amiga
models used different speeds of
ROM/memory. So even an A600 was never
100% cycle exact to an A1000 or A2000

What is useful for compatibility is a
cycle exact Copper mode. And the
Natami support two Copper speed
(slow-compatibility) and
(new-fast).SuperAGA is faster than
AGA but also compatible.

Q. Does SuperAGA support 1080P
resolution?

If SuperAGA runs in the FPGA on the
Natami60 board at over 145 MHz then
this resolution is possible.
Technically 1080p is possible, but it
is not guaranteed that the Natami60
will run at this clockrate.  Graphics
cards are used in SLI mode to
increase frame rates on x86 PCs.

Q. Is it possible for 2 SuperAGA
FPGA's on the same motherboard to do
the same in an Amiga?

The Amiga architecture is build
around different DMA channels.
Everyone has its own distinct task.
The power of the Amiga originates in
the perfect matching and cooperation
of all DMA channels. A PC has only
one DMA channel in its concept. The
CPU today is a brute-force general
purpose calculation monster. It can
do everything on its own, so the GFX
memory is not really part of the
architectural system. On the Amiga it
is the central part.

You may translate your question in
Amiga terms like "is it possible to
add a second blitter DMA channel to
the system".

The original blitter is a pure 2D
unit, so the computations the blitter
can do are not that time-and so would
not be be any beneficial to add a
second blitter DMA. But technically
the answer is "yes".

           CONTINUED IN PART 2