NatAmi 68070 design draft

[alexh]

Zac67 wrote:

The Motorola 68060 CPU was designed that way. Its a CISC decoder in front of a RISC execution pipeline.

Err...

The '060 has no RISC core - this is garbage. RISC starts with Coldfire.

He's trying to say that it is a microcoded processor. That is each CISC instruction is in fact a mini program running on a much simpler faster core.

I tried to give some positive feedback, but it is just showing the gulf between his ambitions and his capabilities.

[alexh]

bloodline wrote:
He optimistically suggests 200Mhz speed... Something as complex as a 68k CPU is never going to get even 20Hmz real speed on an FPGA... I'd be surprised if that to be honest...

I think a bog standard 68000 would run on the fastest FPGA's we have today somewhere close to the 100MHz rate... but they are not cheap FPGA's and once you start adding all the complex stuff the speed will nosedive.

I struggle to get 75MHz for ARM7 TDMI designs using FPGA chips that cost $1000's

[alexh]

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

In numerical numbering at least ;-)

[alexh]

Plaz wrote:
I've made this suggestion before, but it didn't get much milage.

The coldfire appears to be useless for native Amiga. At least at its current power/config.

If it wasn't, we'd have the Elbox Dragon by now.

From what I've read, they couldn't solve the compatibility problems to anywhere within the speed of an 060 and so decided it was a non-starter.

[alexh]

bloodline wrote:

Crumb wrote:
Tobiflex m68k core is faster than a 20Mhz 68000 with a DE-1 board.

Yeah, and really how fast does it actually run?

What do you mean? I would imagine it runs at 20MHz. Although MHz is never a good comparison. You need something like the SPECint performance to get a good comparison. Several of the instructions in TG68 are many times faster than a regular MC68000 meaning a 20MHz TG68 is probably more than 3x faster than a 7MHz MC68000

[alexh]

Atheist wrote:
My computer of 2.26 GHz P IV with 1 Gig ram stalls (~3 seconds) when opening a 2 to 4 meg ASCII file!!!!!!!!!

AND when I save the smallest trivial web pages with FireFox 2, it take 5 to 10 seconds (not kidding, SECONDS) before I can use FF again.

Sounds like you have user error. Probably too much CRAP installed as background processes, probably some rubbish AV software monitoring everything you open, close, save etc.

50Mbyte ASCII file takes about 3 seconds to load here.
After saving webpages with Firefox I can use FF instantly.

[alexh]

Plaz wrote:
I imagined they were trying to do it all in one [Coldfire] CPU, and that can only handle so much.

And you want to have more than one?? Does the Coldfire have any hardware multiprocessing capability? Even if it does, it will be a nightmare to develop and write software for unless you abstract the entire system... and if you are going to do that... why use Coldfire?

[alexh]

wolfchild wrote:
Whenever Natami is mentioned, it seems inevitable that the project team members will eventually be ridiculed.

In order for them to be ridiculed, first they must write something ridiculous surely?

wolfchild wrote:
Without knowing how they are implementing the system, who are we to judge about clock speeds and such?

15 years of ASIC and FPGA design and a current job in a top chip company??

wolfchild wrote:
NO ONE should be telling them that it's not worth the trouble.

No, but at the same time someone should be telling them to be realistic. If they have their heart set on a soft 68k core then they should progress along the development. Start off simple and progress the design forward. Evolve their design. Not make any wild claims about prices, speeds, and structured ASIC's until they are well into their testing phase.

I wish their "team" all the best with their development. I hope that they can inspire and collaborate with the "do-ers" in this area, Tobias Gubener, Arnim Läuger, Gary Becker, Wolfgang Förster, MikeJ and the people over at PACEDev (and everyone else I missed) into taking part in debates.

[alexh]

tiffers wrote:
Next they discovered that they could get their SuperAGA FPGA stuff programmed into the Coldfire package, giving a low-cost ($20 / 'chip' was a price quoted) solution

*Cough* bullsh!t

How much was the design services going to cost(NRE)?
What was the minimum order quantity(MOQ)?

Either you are re-telling the story incorrectly... or someone is telling porkies.

I can imagine they discovered that you can use a synthesisable (Soft) Coldfire alongside your design in a regular Altera cyclone III FPGA. (but $20 sounds too cheap. The Coldfire alone requires at least 5000LE's)

Perhaps there is already an FPGA which has a Coldfire hard macro in the corner (there are a lot with PowerPC's in the corner)?

tiffers wrote:
How many people bought C64DTVs?

A lot less than were supposed to in the UK judging by the fact they were discounted down to just £2 in ARGOS!

[alexh]

wawrzon wrote:
is an a4k (with a cyberstorm060/ppc) cycle exact with a a500?

As cycle accurate as AGA can be to ECS/OCS. Certainly a lot better than MiniMig for example.

wawrzon wrote:
if it is probably natami aga chipset replacement is compatible too since there it was shown to public and there are snapshots of it running genuine amiga apps on the natami page.

While intended to be compatible, it will suffer from just as many problems as MiniMig (if not more) and will take a long time to debug. Mainly because of all the subtleties in the chipset that were never written down (AGA is worse because there was no HRM). But they'll get there. The FPGA development kit will allow for quick bug fixes.

[alexh]

bloodline wrote:
But with the AGA chipset timing was never really an issue

Not sure that is 100% true.

bloodline wrote:
it was the 2megs Chip and register gaps filled with new functions...

Yeah, and more importantly the new kickstart ROM and cpu caches.

bloodline wrote:
switch these off (in the early boot) and all but one of my A500 software would work...

You must not have had much software then ;-)

[alexh]

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

I wonder who this guy is. Either he's the best novice HDL programmer in the world. He used to work in a hardware company which developed 2D and 3D HDL and borrowed upon his knowledge from projects gone by.

Or it's bollox.

Have you any idea how long it takes to develop a rasteriser from scratch?? Add in filtered texture mapping and shading and Z-clipping and you are talking a good years worth of research and work.

I know, designing hardware 3D accelerators was my first job.

It's certainly not a two week task working evenings and weekends.

HenryCase wrote:
According to the Natami devs the SuperAGA development period is close to completion (I have even seen it described as complete). No 20 year wait, so no problem here.

You are far too trusting. If they say complete they mean features implemented... they dont mean features functioning correctly or accurately. If they do, I take my hat off to them.

HenryCase wrote:
let's take a look at one feature that is being considered that would help all Amiga software, which is the integration of Scale2x support into the hardware.

Doesn't look like it's being considered to me. A filter of that complexity would take up a lot of room in an FPGA and require much higher RAM bandwidth.

Plus scale2x sucks ass :-) If you were going to implement an output filter there are much better ones to choose.

[alexh]

Makes sense, there is no way he could have done it from scratch in the time. He's "borrowed" some of his old work for NatAmi. It was probably his inspiration.

Nice to know.

[alexh]

biggun wrote:
The ramp up cost for a Altera Hardcopy are 250,000 quit.

Which one?? $400,000 is the quoted NRE price for a Hardcopy IV.

biggun wrote:
A piece price for a hardcopy ASIC of < $20 is a correct statement.

What is the MOQ (minimum order quantity) at that price?

biggun wrote:
What we said is that we want to bring the chip design
to functional state that this could be done.

Looking forward to it. If you can further the development of the Amiga cores (68k & custom chips), give advice to open source projects such as MiniMig then everyone's a winner :-)