Welcome, Guest. Please login or register.

Author Topic: NatAmi 68070 design draft  (Read 25088 times)

Description:

0 Members and 2 Guests are viewing this topic.

Offline AJCopland

Re: NatAmi 68070 design draft
« on: June 24, 2008, 08:25:35 PM »
I like their enthusiasm but I much prefer the simplicity of using an existing and known working cpu like the 68060.

Still this isn't a thread about what I like :-D

So, and I've been reading Gunnars post on the Natami forum, they've decided that in the future they'd like to make a 68070. It'd be better off being called a 68080 since it sounds more like they want to implement a 68060 and then take it further but that another aside.

Perhaps a good approach to this would be to take the tg68k and try adding features such as pipelining to it's design. it's integrated with an existing platform and behaviour can be compared against an unmodifed version to illuminate compatibility issues.

Ok so it'd still be a 16bit 68k core but it'd give the designer some experience at doing the work.

Andy
Be Positive towards the Amiga community!
 

Offline AJCopland

Re: NatAmi 68070 design draft
« Reply #1 on: June 24, 2008, 08:30:58 PM »
Power6 at TheRegister - IBM Power6 at 5GHz

... now if only they'd actually ship 'em :-D
Be Positive towards the Amiga community!
 

Offline AJCopland

Re: NatAmi 68070 design draft
« Reply #2 on: June 24, 2008, 09:02:43 PM »
Well because the two aren't equivalent.

#1 is taking a 68k CISC instruction which can (but doesn't have to) involve several steps to accomplish and breaks it down into those discrete steps. Those are then fed through a RISC core which has been specifically designed to handle the aforementioned discretised 68k instructions.

PROs:
Its how modern x86 cores work.
You can apply lots of fancy things to that simple RISC68k core at it's heart.
To the outside world it looks _exactly_ like a really flipping quick 68k cpu complete with addressing, instruction, bus design. The whole lot, it could be a drop in replacement for a real 68k chip
CONs: hard to do, CISC->RISC decoder takes up transistors and space, decoder extends pipeline causing branch prediction penalties to increase etc etc etc

#2 is completely different. An x86 is just a completely alien device, it has a different endianess, it's buses are different so the way it communicates is different. I wouldn't even know where to begin. Not only that but the instructions that it actually processes are different so it can't even take the 68k instruction and do anything with them. x86 emulators work because they pretend to be the entire machine including the ram where the executable resides.
Attempting to attach an x86 cpu onto a 68k bus is like trying to get a tortoise to go faster by strapping an angry wolverine to it :lol:
Be Positive towards the Amiga community!
 

Offline AJCopland

Re: NatAmi 68070 design draft
« Reply #3 on: June 24, 2008, 09:45:06 PM »
Quote

alexh wrote:
Quote

koaftder wrote:
it's going to be really cool, and it's going to exceed the 060.

In numerical numbering at least ;-)

And in quantity, after all you could have maybe more than 10 of these in existence :-D

Ah, I wish they'd just investigate the ColdFire further over this option. Or at least start from another base and try extending it rather than make sweeping statements about 200Mhz 070's being easily achievable etc.

An 060 and a real, working, purchasable(!) Natami that's all that anyone wants.

Well somehow persuading Freescale to make some more 060s so the price comes down a bit would be handy too but lets not go nuts now ;)
Be Positive towards the Amiga community!
 

Offline AJCopland

Re: NatAmi 68070 design draft
« Reply #4 on: June 25, 2008, 04:28:44 PM »
@wolfchild && LoadWB

Don't get me wrong I'm glad that they want to do it, I'm just worried that it becomes a matter of overreach. ASIC, or even a structured-ASIC, looks like it could be out of their reach which also precludes ColdFire even if Gunnar is right that it can be compatible enough to 68k code.

I want to see the Natami do well and initially that seems to mean the Natami60. The N68070 is a cool idea but is also an enormous undertaking.

How much do FPGAs with a couple of million gate equivalent cost? About $522 according to DigiKey (122-1350-ND). Which bearing in mind that the 68060 used 2.4million gates is a reasonable assumption to have about target size which doesn't compare all that well to the remaining 35 68060s also at DigiKey (its easy to get these prices from there which is why i'm using it).

All of this is why I suggested they not speculate but actually try to implement some of the things that they want in an existing verifiable core like the tg68k such as Instruction Pipelining and Superscalar Execution that the 68060 had.

Just my 2 pence.

Andy
Be Positive towards the Amiga community!
 

Offline AJCopland

Re: NatAmi 68070 design draft
« Reply #5 on: June 25, 2008, 04:31:33 PM »
Quote

jlariv8957 wrote:
Hi,

I just want to know what is the 68070/68080. I always tought the 68070 was a microcontroller chip used on CDi what did not share anything with the 68k familly.

Thanks


The N68070 in this case is a design for a chip that the Natami team members would like to implement after they've done the Natami060. See Natami 68070.

Andy
Be Positive towards the Amiga community!
 

Offline AJCopland

Re: NatAmi 68070 design draft
« Reply #6 on: June 26, 2008, 05:56:10 PM »
All fair points but I don't want to discourage the Natami guys or believe that they should be discouraged.

Look at the argument for the natami as the argument for the DS over the GBA, or the GBA over the GB. Many years between them, lack of straight compatibility (not entirely true of GBA->DS but for the sake of argument...). They're simply upgraded with newer designs but no less '80s/early '90s in their implementations.

Now the Natami with a 68060 or even ColdFire+SuperAGA ASIC (yes yes yes I know they'll never do that one) is not going to be a handheld suitable one. However the design is simply damned interesting and even with a 060 would be faster than the PSone. If they did go for a ColdFire based option then it'd be up within the speed range of a PSP/PS2.

So mostly it's interesting in it's own right, perhaps only as an extended hobbyist project but it's still fascinating and judging by the interest in it there are people who want it.

Andy
Be Positive towards the Amiga community!
 

Offline AJCopland

Re: NatAmi 68070 design draft
« Reply #7 on: June 26, 2008, 08:27:13 PM »
Quote

bloodline wrote:
No. I quite agree here. NATAMI is a brilliant hobby research project for the people involved and would be happy to read about it and see it in action... it's the motivations/long term goals that is suspect!


Yeah the long term dreaming is what I'm reading a lot of but honestly I don't think it's from the developers too much. Occassionally Gunnar goes off on one but if he wasn't having those flights of fancy in public then he'd be having them in private :-D how can you resist dreaming about "what-could-be" when you're helping to make something.

Quote

Come come now!!! You and I both know the GB is a totally bad example! That is a product for which software development never stopped and had a massive installed user base which needed to maintain backward compatibility with old software... The market for which the GB served was never overtaken by other systems, the technology for mobile devices has only in the past few years really started to catch up with desktop...

The Amiga, perhaps 10 years ago, would have had some software that other systems couldn't really match... but now, the very concept of Amiga software for anything other than a small group of hobbyists is laughable.

The GB is not a good argument for "SuperAGA".

Ah ha not so fast, GB->GBA is a marvellous argument (or we wouldn't be having on :lol:) because I don't care about the actual success of the platform only the hardware differences :-P
You see we're discussing really two overlapping things, the market for something with the hardware and design of something.

If we compare the GB to the A500, the GBA to the A1200 then think of the Natami as the DS-that-never-was. That's all that some of us fanboys want really to see a sort of what-could-have-been, a continuiation of the series...

..maybe that's a better analogy actually, think of it like a TV series that gets cut halfway through, its still got a fanbase so after a few years another production company picks up the right and they make the rest of them. Now they can either keep the original staff that everyones familiar with (Natami-MiniMig-CloneA) or they can hire new actors to play the same roles :-o

Quote

Errr.. well, ok... but the PSone is, what, 14 years old... And the NATAMI would be expensive, limited and buggy compared to a cheap off the shelf GFX chip...


Yeah actually I'd like it if they DID put in an extra off-the-shelf GFX chip on there too. My point about the PSone was that its a reasonable benchmark to surpass. Slow 33Mhz CPU, weak 3D chip (by todays standards) but a potent machine when used right. Even an A1200 with a 3D chip would have given it a PSone a run for it's money at the time if commodore had ANY idea that 3D was coming.

See below...

Quote

Quote

If they did go for a ColdFire based option then it'd be up within the speed range of a PSP/PS2.

I'm not really sure how you are making this comparison... I simply don't know where the information for such a conclusion could have come from?

I code for the PSP and PS2 so know them fairly well. The PS2 would be harder to surpass but the PSP would be relatively trivial for a CF based machine at 300Mhz to 400Mhz.

Thing is even a 68060 60Mhz based machine with good memory access (something that particularly hobbles the PSP) would get close to PSP standards if they'd give it a basic 3D GFX chip, leave the 2d stuff to the AGA/SuperAGA side. They'd only have to give it something with multi-texturing support to improve on the performance of the PSP rendering.

Hmm gone off on a tangent, damn, was beta for our game today so just got home :-D

All of the above is trying to say that you don't have to do much to get a hobby platform upto older-gen's console power and that's only been the Natami's stated goal, to give the Amiga (AGA) chipset a bit of a boost up whilst maintaining as much compatibility as possible.

Quote

Which is what the NATAMI is interesting for! I don't understand, why people want to big it up into something it isn't... when it already has a decent reason.

Agreed, what they're saying though for the NatAmi with 68060 isn't that it's going to be the greatest thing ever.

Hell even getting the Natami upto even the level of the PSone would be an upgrade, after all 14 year old technology is better than 16 year old (A1200==1992!) technology :-D

Andy

PS: sorry I'm knackered so the above is all confused, I'm gonna go watch anime and get some sleep!
Be Positive towards the Amiga community!
 

Offline AJCopland

Re: NatAmi 68070 design draft
« Reply #8 on: June 26, 2008, 10:02:45 PM »
@HenryCase
Quote

Cue Garth:

 :lol:  :lol:  :lol:  :lol:  :lol:
awesome :-D
Be Positive towards the Amiga community!
 

Offline AJCopland

Re: NatAmi 68070 design draft
« Reply #9 on: June 26, 2008, 10:19:33 PM »
Quote

HenryCase wrote:
Quote
AJCopland wrote:
Yeah actually I'd like it if they DID put in an extra off-the-shelf GFX chip on there too.


FYI, Gunnar has added a 3D core to the SuperAGA chipset.

I wonder what it's like, I found this thread but it just made me want to headbutt the desk in frustration at these people until my brain leaks out of my eyesockets nuuuuuurgh. :crazy:

If I could actually stick my oar in I'd say that one of the most annoying features of working on the PSP, not PS2, was that you can only do single pass texturing. I know it sounds strange but even just doing two textures in a single pass would have almost doubled our framerate for a lot of titles. Without it we had to resort to multipass (redrawing the scene with different textures and blend modes) which of course means drawing every triangle twice for a single frame.

Ah well anyway we'll see.
Be Positive towards the Amiga community!