A project I'd like to see from Jens...

[biggun]

Hi Matt,

our new Cyclone 5 accelerator cards are done already.

Cheers
Gunnar

[biggun]

for what system? a 600?

For A500/ A1000/ A2000 / CDTV.

Card specs are:

* very fast 68K CPU
* 128 MB DDR3 Fast-memory
* SD-card usable/bootable as IDE-device
* Network interface
* RTG Graphics Card (chunky/Hicolor/truecolor) with HDMI out

The physial card design is done.
Testing / Driver development needs to be done.

[biggun]

so is it designed or is the prototype already fabricated?

The boards are in production/shipment.
We expect to receive them this week.

[biggun]

Why can't you cannibalize old processors for new devices .

We use a new processor for several reasons:

a) The new processor is a faster than existing 68K processors.
b) The FPGA solution allows a lower price than a 68060 CPU solution would have
c) The new processor in the FPGA needs less power.
    Therefore the card can run in a normal A500.

[biggun]

Ideally this could be an upgrade board for the Replay instead of the '060 board it might one day have. FPGA chipset and modern board + FPGA CPU card.
Andy

Yes - in theory.


On the other hand, all you need is make a standalone system out of the CPU card
is to add "Joyport", audio jack,  and a keyboard connector.
This is no rocket science.
This solution would be very cost effective and for the saved money you can get yourself a big ass Flatscreen. :-D

[biggun]

Hi Olaf,

Lets be totally open and very honest about the project.
Who we are, where we are and where we are heading.

Today there is the Vampire600.
The Vampire600 has 64 MB fastmem and an FPGA which you can load a new fast 68K CPU into it.


The next card, the new Cyclone5 card is in production right now.
The Cyclone 5 card offers
* Faster FPGA  = about twice as fast CPU compared to the Vampire600
* 128 MB Ram
* Video out HDMI
* SDcard / IDE
* Network

As soon as the Cyclone 5 card is fully tested and we see that all works without yellow wire
and as soon as the sales/distribution is set up - it will be sold and
we will offer the CPU upgrade file for the Vampire600 owners to speed up their CPU as free download.

Then Vampire600 owners will then get a CPU comparable to 68060 in speed for free.

If the A500 card runs fine we will make card for A1200,A4000 all other models etc.
We as team have recently bought several A1200 and have already stocked up mostly every made AMIGA here for testing the cards: E.g  A500,A1000,A2000,A3000,A4000,A1200,CD32

Lets be honest:
The APOLLO CPU is an envolvement of the 68050 CPU done for the Natami project.
The RTG Chunky/truecolor display is basically a steal from the SAGA/Super-AGA chunky stuff we prepared for the NATAMI.


You don't need to be an Einstein to see that you can also take the CPU card out of your A500, and if you add a Joyport to it - and you basically have a new NATAMI - for little money.

To be clear: I'm not promising anything here.
This is not a market announcement.

The post was started as question whether we need help to design a new CPU card - and the answer to it is "No" we have designed it already and the above is our current status....

[biggun]

of course thomas had a different goal, namely to build a complete system, but i was speaking of the "softcore" part of team.

also thats where the fun is. testing, submitting reports, reading what commits have been submitted day by day.

Actually this is exactly how the Softcore is tested/developed.

[biggun]

crowth?

What is the most nice one to program

The most nice to program is 68k - by far.

Clock by clock a good 68K can also stand its ground and even beat the other performance wise (clock by clock).

[biggun]

Will that stand even to other one-chip embedded architectures

A good 68K can outperform them, hands down - on a clock per clock comparison.


But what do you really want?

If you have an A500 or A600 or A1200 then you ideally look
* for a good CPU card that does not run to hot in your Amiga case. Am I right?
* and you want a card which works well with your Amiga power supply,
  so that you do not need to use a 100 Watt PC power supply for it.


This means for many chips your AMIGA does not have good enough cooling and not good enough power supply.

An FPGA based 68K would be happy with this cooling requirements.

And performance wise it can beat PowerUP PPC cards
and does beat hands down stuff like 240 Mhz Coldfire or low clocked ARMs.

So with the low end chips a good 68K has no problem to compete.

Do you want to run 68K code?
Then this means you need to run 68K in emulation on your ARM.
In this case a good 68K can easily compete and beat the absolute high end ARM chips.

[biggun]

I was quite saddened when the Natami project became stalled, or stopped (I don't know if it is cancelled, or just on hold).

Thanks, this is nice to hear.
Let me answer it this way.

Our plan is to bring out the card for every AMIGA model.
We also plan to then create a stand alone version of the card.

As you might know the people doing this CPU card previous worked in the NATAMI team...
Does this answer your question?

[biggun]

Everyone can take what he/she wants.

If you want an 68K fine.
If you want a PC, fine too.



My view on ARm is this:

I own both AMIGA and Archimedes since their beginning.

When A500 came out the 68K run at 7 MHz and most instructions needed 4 - 20 cycles.
The average instruction was around 8 cycles maybe.

When the Archimedes came out it ran at 8 MHz and each instruction needed 1 cycle.

Yes the 68K supports more complex instructions.
For some 68K instructions the ARM needs 1 instruction.
For many it needs 2, or 3.
Sometimes the ARM needs even many more like 4, 5 or 6

But even when the 68K in the A500 could do the work of 5 ARM instruction with a single 68K instruction - this single 68K instruction took like 20 cycles on the 68_000

This means the ARM was always faster.
If you compared the 8MHz ARM with the 7MHz 68000 the ARM was always much faster.
And at this time it good some myth of performance.

Now today the situation is very different.
Today a good 68K CPU can execute even complex 68K instruction in a single cycle.
For example:
ADDI.L #1234567, $1234(A7)
A good 68k can execute this instruction in 1 single cycle today.

A good ARM needs about 6 instruction to do the same work.

As you see today - clock by clock - the 68K is much stronger.

[biggun]

Hi Dave.

thanks for the nice words. its appreciated.

The Amiga commity seems a lot smaller to me.
Its hard to estimate how many people are still around.
Would be good to know how big the market really is today.
Producing 100 or 2000 boards would make a difference in the price.

[biggun]

OK lets be very honest, ok?

We do this thing not to make a lot of money. ;-D
We do this because _WE_ want it ourselves.

We work on our CPU Core for about 5 years now.
Frankly, I regard the Motorola 68060 Core as very good core,
but I strongly believe that our Core is better.
And our Core is faster, its even faster than all overclocked 68060.

We do this project because we want to use our Core together with in a nice OS.
I personally regards Amiga OS as much nicer and friendlier than Linux.

I think that our CPU Core has enough power to do quite a lot of interesting stuff.
I believe its strong enough to play games like Wing Commander 4, or Northland or Diablo 2.
And for me also playing 640x480 AVIs is an important goal.

I speak here a lot about "me".
Because I want you to understand what my and my friends motivation is.
We want to surf internet a little bit, to be able to download some ADF or buy something on EBAY,  
And we want a low cost, silent harddrive,
also I want a nice screen resolution when working with my computer.
720p minimum screensize with 16bit colors is what we want.
In this resolution you can play games in nice quality, and have a nice looking Workbench ...
And we want to attach a big LCD TV's without loosing quality over scart.

These are the features that we personally want from our "toy" computers.
And this is why our card is like it it.

There are certainly many wishes other people might have.
But if we start adding the stuff that others like - then this will be a project which will never end.
Frankly, we did put all the stuff and the card which WE need and WE want.
And we added nothing else because we want to start using our cards next month - and not next year.

This is the honest story behind this.
If you like our card then you are welcome to join our boat and sail along. :-D

Lets also be honest what we want to do in the future.
This card is our toy system to continue to develop the CPU core.
We believe that we have by far not reached all that is technically possible.
And we also have our own "work in progress" version of a next Gen AMIGA chipset.
I call it Super-AGA, and the card will also also us to tinker around with it.
Please get me right: This does not mean that we will ship the card with this.
I just tell you what we want ourselves to use the card for next year...

[biggun]

I have a design question.
If I understand correctly, the new card will go into the cpu socket.
Is there any impact to not use the CPU slot (when it's avaible)?

The card goes into socket in A500/A1000
This means the card is internal for these AMIGAs.
While a card using the A500 slot would be external

For the A1200 the card will be adapted to use the CPU expansion connector.
This means the card is internal here.

For the A4000/A3000 the card would use the CPU slot.
This means the card is internal here.

Does this answer your question?

[biggun]

What about the Amiga 2000?

This CPU upgrade card fits several AMIGA models:
A500
A1000
A2000
CDTV

This image is from the first batch of boards.
For the production run we will evaluate a breakout options to allow moving the HDTV and NETWORK connectors to a plate e.g for the backside of the A2000.