Excitement about NatAmi

[Gulliver]

I am just wondering if this fpgaarcade/Minimig AGA device is going to drive the Natami to its knees and leave it in agony?
I am just seen it from a pure real, practical & economical point of view:

Fpgaarcade has already got production units
Cost is 200 euro (Later batches are announced to be at least 50 euro cheaper due to NRE)
It is open source
Does not require a real 68k in its design

On the other hand the Natami:
Is closed source
There are only few prototypes that dont fully work
Cost is near 700-1000 US$
Hasnt got implemented the so called new features yet

Please dont take this as bashing, it is just what I actually see.

[jj]

A does a 100Mhz 060 real = 100Mhz PPC (Assuming they mean G3 here)
 
And will a 100mhz 060 be able to decode divx/xvid/dvd ?

[Karlos]

A does a 100Mhz 060 real = 100Mhz PPC (Assuming they mean G3 here)

Not really, though it's hard to give an exact comparison given the different architecture of the two parts. A 68060 running typical 68K object code will get through around 1.3 instructions per cycle. A PPC750 (G3) will get through around 2.3 for typical PPC object code. The PPC on average has a slightly lower code density, but on the other hand, has some instructions that can do more work than you can do in a single instruction on 68K (floating point multiply-add would be an example that might see use in codec code).

Code properly tuned for each can achieve significantly better in those circumstances where you can leverage parallel execution across multiple functional units, properly fold out branches and make the best use of pipelining.

And will a 100mhz 060 be able to decode divx/xvid/dvd ?

Sure, just not necessarily in real-time for any watchable resolution.

[vidarh]

I am just wondering if this fpgaarcade/Minimig AGA device is going to drive the Natami to its knees and leave it in agony?

I think there's space for both. The FPGA Arcade Replay board is not meant for maximum performance - it looks like it'll do quite well, not least thanks to Yaqube optimizing the TG68 core (adding cache etc), but we're talking somewhere between  68030 and 68040 speeds here, while the Natami team appears to at least aim for something far faster. We'll see once they actually deliver something, though.

Does not require a real 68k in its design

Neither does the Natami. It has a CPU slot, but one of their goals is to use their own 68050 soft-core on the FPGA.

On the other hand the Natami:
Is closed source

I agree I like the open approach of the Replay better, but I doubt this is going to be an issue for most users.

Cost is near 700-1000 US$

What's that based on?

Hasnt got implemented the so called new features yet

My understanding is that SuperAGA is there, and that the 68050 is pretty much there, though they keep doing refinements. They've also got a lot of things such as a far faster memory interface etc.

We'll see once they actually demonstrates something, though.

[asymetrix]

@thread

What is this cycle exact rubbish I keep reading. There IS NO SUCH THING.

All Commodore Amigas A1000, A1200, A600, A4000 etc are never cycle exact to each other - they technically cannot be because they all run at different speeds !

The are compatible but never CYCLE EXACT - ask any engineer.

[Louis Dias]

@thread

What is this cycle exact rubbish I keep reading. There IS NO SUCH THING.

All Commodore Amigas A1000, A1200, A600, A4000 etc are never cycle exact to each other - they technically cannot be because they all run at different speeds !

The are compatible but never CYCLE EXACT - ask any engineer.

Funny, Gunnar said the said thing.  Imagine that...

@the_leander
Can you prove to me the sun will rise tomorrow?
Mathematically, if all goes to plan it will.  However, it could go super nova over night...
Seriously, stop being ignorant.  MikeJ's board is already blitting something like 40x faster than a real Amiga because of the memory bus.  Natami's has already been demostrated to be faster on the LX board.  But oh yeah, you can't buy the LX board so it doesn't exist and is vapor.

[bloodline]

@thread

What is this cycle exact rubbish I keep reading. There IS NO SUCH THING.

All Commodore Amigas A1000, A1200, A600, A4000 etc are never cycle exact to each other - they technically cannot be because they all run at different speeds !

The are compatible but never CYCLE EXACT - ask any engineer.

You're weird...

Buy a hobby microcontroller kit, have a play... Lean how digital electronics works

[Cosmos Amiga]

>The PPC on average has a slightly lower code density

It's a joke from you ??

Watch this : http://library.morphzone.org/An_Introduction_to_MorphOS_PPC_Assembly

[the_leander]

@the_leander
Can you prove to me the sun will rise tomorrow?
Mathematically, if all goes to plan it will.  However, it could go super nova over night...

Strawman.

Seriously, stop being ignorant.

I ask you (repeatedly) for evidence to back up your assertions and you utterly refuse to supply them, then you accuse me of ignorance? GTFO!

Until it is released, all your claims are just that - claims.

SamuraiCrow at least had the good sense to coach much of his response as something he hoped for rather than fact.

But oh yeah, you can't buy the LX board so is vapor.

Fixed.

Yes. Given the history of this community and all the hardware that never moved beyond prototyping (or even got that far). This is the only reasonable position to take.

[bloodline]

>The PPC on average has a slightly lower code density

It's a joke from you ??

Watch this : http://library.morphzone.org/An_Introduction_to_MorphOS_PPC_Assembly

And the first example shows PPC's lower code density:

[Karlos]

Unless he meant "slightly" was a joke.

In reality, most object code I've compiled for ppc and 68K shows between a 30-70% size increase in the PPC version. It varies a great deal. In some extreme cases, I've seen object code that's more than twice the size.

As for the example above, the increased complexity of the PPC example is not entirely down to the PPC architecture but also considerations of the OS itself.

[bloodline]

Unless he meant "slightly" was a joke.

In reality, most object code I've compiled for ppc and 68K shows between a 30-70% size increase in the PPC version. It varies a great deal. In some extreme cases, I've seen object code that's more than twice the size.

Very good point, he is probably unaware of our British sense of understatement

[Louis Dias]

Strawman.

Before I go on, you do realize that "blitting" is basically just moving memory, right?
http://en.wikipedia.org/wiki/Bit_blit

I ask you (repeatedly) for evidence to back up your assertions and you utterly refuse to supply them, then you accuse me of ignorance? GTFO!

DDR2 > DDR1 > SDRAM > Amiga DMA  now relate: Natami>fpgaarcade>MiniMig>A500
Until you have accepted this, there is no point in going on...
DDR2 is clocked higher and transfers twice the data of DDR1.
With data being transferred 64 bits at a time, DDR2 SDRAM gives a transfer rate of (memory clock rate) × 2 (for bus clock multiplier) × 2 (for dual rate) × 64 (number of bits transferred) / 8 (number of bits/byte). Thus with a memory clock frequency of 100 MHz, DDR2 SDRAM gives a maximum transfer rate of 3200 MB/s.

Now what is the memory transfer rate of an unexpanded Amiga 500?  Let's see if you know how to do your homework?
I'll even help you out:  The Draco could do 30MB/s which was even faster than Zorro3. 3200/30 is greater than 100.  Amiga's bus is 3.57 Mhz and only 16 bit on the early models.
However, I do realize that the sun may not rise tomorrow.

Until it is released, all your claims are just that - claims.

See above.

SamuraiCrow at least had the good sense to coach much of his response as something he hoped for rather than fact.

He's not wasting time justifying what is obvious.

Fixed.

however your thought process seems broken.

Yes. Given the history of this community and all the hardware that never moved beyond prototyping (or even got that far). This is the only reasonable position to take.

The developer has announce boards are almost ready for production.  It has already gone through 2 prototype stages...that have been well documented.  '060 cpu cards have been developed a long time ago.  You are fighting a useless fight.

It sad that someone make a statement on this forum.  Someone else is then in disbelief and demands proof despite have no basis for such a demand nor even has a right to make any demands what so ever.
Wouldn't it be simpler to come up with evidence to the contrary on your own rather than troll out demands that carry no weight other than in your own little black book?

[Piru]

The developer has announce boards are almost ready for production.

I remember that boards were initially planned to be released summer 2008. What does this "almost ready" mean this time?

[Louis Dias]

I remember that boards were initially planned to be released summer 2008. What does this "almost ready" mean this time?

Why don't you direct your question to Thomas?
Did you ask MikeJ this 1.5 years ago?

[SamuraiCrow]

First off, as a member of the Natami team it would be unprofessional of me to allow hopes to get excessively high.  Without a prototype to test on we can't do much profiling.

Secondly, we can test the VHDL code with a simulator so it looks like the N68050 is right on track to be clocked at 133 MHz and to be able to combine instruction sequences internally for one instruction cycle-per-clock execution for most instructions.

Lastly, I think that the "almost ready" figure was a reference to AGA chipset feature set and didn't include the N68050/070 CPU core.  Also, I've mentioned in a previous post that the Natami LX was a prototype and had some performance lapses in it preventing it from going to market.