@WolfToTheMoon

I was referring to this:
http://amigaworld.net/modules/news/article.php?storyid=6649
(c60 dualcore 1Ghz in desktop build)

@franksexy

Those MIPS system are not desktops caliber (and not produced in volume like PPC computers in 2004-2005).
(Unless I'm mistaken they are 800-900Mhz lilliput systems that make Amigas look pretty awesome powerhouses. And neither the server board is a desktop (also it uses puny 1Ghz chip).)

Last edited by KimmoK on 25-Jan-2013 at 08:41 PM.
Last edited by KimmoK on 25-Jan-2013 at 06:57 AM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@matthey
Quote:

What is the name and website for information on your project?

http://www.tinaproject.it

Last edited by wawa on 25-Jan-2013 at 06:49 AM.

@simplex

I meant to come across as agreeing with you, actually

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

Quote:

wawa wrote: @cdimauro thanks for explanations. ill observe the project even though i still wonder how its going to be financed. i mean the employees want to be paid even if they are working on a chiefs pahntasy. but lets assume it has been thought of.

It's not my business.
Quote:

Quote: I hadn't an MMU on my Amigas (2000 and 1200), so I don't remember details on how and what kind of usage make of it tools such as Enforcer. what concerns mmu it is actually very useful with muforce imho. it has had some application in higher end amigas even what concerns system functionality afair, but im not sure what it was. if there is actually a way to implement this in an easier way, im all for it.

We'll see, but it's not a priority. I have other ideas to enable some useful debugging functionality that doesn't require an MMU (or a "full MMU"). Right now, it's premature to talk about it.
Quote:

now: Quote: But if you are using an x86, it has enough power to emulate a complete Amiga machine, not only a CPU, at very good speed and beyond original machines. first of all even on x86 there is considerable amount of power lost on chipset emulation that can be spared if you have it available in hardware as on genuine amigas.

So you're interested on having a more powerful CPU that can supplant the one used on an original Amiga, leaving the latter to just act as a "chipset". Right?
Quote:

of course, the amiga chipset philosophy could be explored and induced into its possible alternative patch of development as it was attempted and talked about with natami. however what would come out of it? 24 bit hdres planar screens? strange obscure pixel formats? an alternative attempt at 3d acceleration not compatible with anything else developed while last 15 years? lets face it, at the end of the day every aspect of the implementation available by means of contemporary gfx and sound card wold beat it by performance, software portability and any other factor i can think of.

Sure, and that's why I think that we can provide just some expected features.

Chunky pixel modes (8, 16, 32 for sure; 24, 4, and 2 can be an option, to be evaluated later) doesn't require so much work, because we are essentially "skipping one step" (internally the display logic "chains together" the bitplanes to compute the effective index to access the CLUT). The Blitter will be updated accordingly.
And some audio enhancements to bring 16 bits and 48+Khz. 8 channels will be evaluated later.

They are "bare minimum" enhancements for a new "Amiga-like" machine in my opinion, that can everyone expects, and that can be quickly implemented as a Picasso96 driver and AHI driver.

But if the project becomes a success and people demands for better things, more important enhancements will come.
Quote:

therefore i think the amiga chipset can very well be left as is for legacy reasons. it isnt even necessary to reproduce it, as genuine hardware it is still available in quantities that likely may cover the demand of the scene. it is unlikely to produce any custom hardware in comparable numbers.

But Alberto needs to develop it, for his company. So we are just picking an opportunity to build a new Amiga-like hardware. Alberto will sell the know-how (and may be motherboards too) to his customers this way, while amigans can have a new toy to play with (that's an hobby, in the end).

Think about ACube: they build a PowerPC motherboard with an FPGA because they sell it to their customers (industries), but OS 4 users can also buy it.

It's the same here with TiNA: why don't get the chance?
Quote:

therefore i think it is most sane to take your amiga and use what remains usable of it, which is kickstart, keyboard, power supply, legacy interfaces, legacy chipset, floppy drive, and the genuine (labelled) housing (!) and extend it with whatever tech can be introduced to it. be it an x86 board with sata, pci and usb, if it behaves like an ultra fast 68k accelerator, then what makes it inferior to, say, a cyberstorm. from your point of view, one could argue an accelerator extended with pci board, gfx and sound card doesnt actually need amiga underneath, yet such extensions have been succesfully built, people bought them, are proud of their frankenstein configs and the demand is still high. thats why i think an x86/arm accelerator board might not only be sane but perhaps even be a commercial success, feasible with least design effort, in shortest time and with least investments.

Once the project is complete, and may be some expansion cards for existing Amigas can be build. Using an x86 or ARM for such task will require a lot of work too, and I don't think that we can reach the same performance of a new 68K core, especially for ARMs, which are less powerful.

Also, you must take into account the power consumption: an expansion card can't draw a lot of it, so you must resort to low-power x86 or ARM CPUs to save it, providing lower performance. Intel's Haswell core can change the picture, but it's a complex and expensive CPU, and developing a new board can be very difficult too.

You want MMU support too, but you know that emulating it greatly slows down performance, because there's no x86 or ARM (virtualization) technology that can be used to accomplish this task, since 68K MMUs were very different from the ones implemented by x86s or ARMs.

Finally, the next time that Intel's changes the socket you must begin another project. FPGAs give much more stability and convenience from this PoV.

That's just my opinion, but we are open to any idea. And... I'm not the head of the project as I said.

@cdimauro

How do you calculate the costs for the FPGAs? Is it not becoming too expensive? I know (from Natami) that FPGA is the most expensive component on such a board and you want to use even a couple of them.

Quote:

simplex wrote: @cdimauro Quote: Sorry guy, but I don't see any desktop or server PowerPC which is produced or planned by Freescale. Really? Navigate to www.freescale.com. Hover mouse over "Products". Drop-down menu has several choices. One is Microcontrollers, another is Processors (e.g., PowerQUICC, Power Architecture). Freescale's website, at least, acts as if these are different.

Haven't you took a look at it?

For PowerQUICC: "PowerQUICC Communications Processors

Through a wide-range of low-power and high-performance solutions, the PowerQUICC families built on Power Architecture® technology support the entire spectrum of embedded networking equipment applications, from the core to the edge to residential access."

A router!

And for Power Architecture: "Host and Integrated Host Processors (8xxx, 7xxx, 7xx, 6xx)


General-purpose processors based on the e600 core, built on Power Architecture technology, offer performance from 500 MHz to 1.8 GHz, including a dual-core option. Ideal for high-performance processing in networking, aerospace and defense, storage and pervasive computing applications, such as home media, printers, computer clusters, blade servers, thin clients and gaming systems."

An high-performance single core built for... embedded market: that's what's the e600 core, as you should know.

It's based on the G4, OK, but it WAS a powerful processor, since the development is stopped. The last core revision is 2008: FIVE years ago. Do you have any evidence of future core enhancements by Freescale? Take a look at KimmoK's link: do you see any future improvement for this core? I don't.

Freescale has also 6800 based SoCs, and our beloved 68000 ones too, in his catalog, as you can see, but they are NOT developed anymore...

Sorry, but that's the cruel reality that we must accept.
Quote:

The guy quoted in your article (Lees) is the director of microcontroller technology: Quote: Geoff Lees, Freescale's new general manager for microcontroller products... His statement can be read to say that R&D in microcontrollers will be nearly 100% ARM, not R&D in all products.

We can say that at least microcontrollers is the segment where we know for sure that Freescale is doing R&D.

There's nothing similar about its "processor" segment.
Quote:

Hence my question. I don't have a dog in this fight; I just want to know if anyone who actually knows something about Freescale can clarify. Cherry-picking ambiguous press interviews doesn't clarify.

Right, but I think that the PowerPC situation is quite evident, as I have already explained in my article.

Quote:

matthey wrote: @cdimauro Hi Caesar, I am Matt Hey from the Natami Forum. Remember discussing 68k enhancements where we both argued for the block memory copy/clearing instructions and EA update addressing modes?

Hi Matt, I remember you. Nice to meet you again.
Quote:

We certainly had similar ideas about what the 68k needs. I created some documentation for a new open 68k ISA standard. It's all preliminary and needs some work but it has some good ideas. Since you seem to be helping with this Italian fpga Amiga project, maybe we could work together to help refine it and for possible use in your project. It can be found here: http://www.amigacoding.de/index.php?topic=273.0 http://www.natami.net/knowledge.php?b=3¬e=47318

Very good! I took a look at this, but right now I've not so much time. I'll read everything this weekend.
Quote:

I also have an older version of the N050 fpga CPU if you want to have a look. Jens is ok with it and will be releasing the VHDL source anyway when he gets it cleaned up. It's pipelined and looks well designed and coded to me but it's incomplete.

It'll be great! It can become the CPU core for TiNA project. I hope that Jens will release it sooner.
Quote:

The TG68k is no doubt more mature but simpler.

And limited too, as I read. I wasn't designed for high performance.
Quote:

What fpga CPU were you looking at using?

EP4CE30F23C8N for all three chips. But I think that EP4CE30F23C6N can be affordable too: it's 500Mhz (instead of 400Mhz), and costs only 2 more euro.
Quote:

Is anyone working on the CPU currently?

The Minimig one (which is TG68 based, if I remember correctly), since we have found nothing else. A forum user has tried to contact Minimig v2 / FPGArcade people, but hasn't received any answer, unfortunately.
Quote:

Quote: cdimauro wrote to wawa: That's all. I wanted to make some clarifications because I read too many things about it. I don't want to talk again here, since we have an open forum where anyone can subscribe and make questions and/or contribute (as we hope). What is the name and website for information on your project?

Someone has already gave it: www.tinaproject.it

You're welcome. I'm one of the admins, and I've created both Italian and English forums. The project was started by Italians, but we are very happy that other people embrace it.

@cdimauro

Quote:

So you're interested on having a more powerful CPU that can supplant the one used on an original Amiga, leaving the latter to just act as a "chipset". Right?

exactly. think of it like the reverse of the amiga on a pci card idea, which is more sensible imho, after all you want an amiga on your desk, right? also think of it like a sort of amithlon, only equipped with the legacy hardware for full backwards compatibility. isnt amithlon supposed the best what may have happened to amiga? well, what im talking about is even better!

Quote:

and that can be quickly implemented as a Picasso96 driver and AHI driver.

ahi will be simple i think. chunky rtg a little more a challenge. i talked to thomas richter who implemented natami saga modes and he advised to completely resign on amiga os "monitors" approach. afair he said legacy ocs-aga software side philosophy is almost impossible to extend sensibly for chunky gfx. even p96/cgx carrying extended planar legacy had impact on performance due to it. also as both systems are closed source you will likely need to base your solution on what winuae provides, which of course is a general problem affecting all other projects too, aros68k for that matter.

Quote:

But Alberto needs to develop it, for his company. So we are just picking an opportunity to build a new Amiga-like hardware. Alberto will sell the know-how (and may be motherboards too) to his customers this way, while amigans can have a new toy to play with (that's an hobby, in the end). Think about ACube: they build a PowerPC motherboard with an FPGA because they sell it to their customers (industries), but OS 4 users can also buy it.

that really intrigues me. i understand there might be a market for ppc industry boards, but 68k compatible softcore? okay thats still imaginable, but what about genuine amiga chipset? who needs it besides nerds?

Quote:

Using an x86 or ARM for such task will require a lot of work too, and I don't think that we can reach the same performance of a new 68K core, especially for ARMs, which are less powerful.

look at amithlon on contemporary hardware. lets imagine the expansion will be an all in one module, i think its safe to say it will be able to emulate 68k numerous times faster than an consumer level fpga. id love to be proven wrong.

Quote:

You want MMU support too, but you know that emulating it greatly slows down performance, because there's no x86 or ARM (virtualization) technology that can be used to accomplish this task, since 68K MMUs were very different from the ones implemented by x86s or ARMs.

toni wilen is just about to finish the complete mmu support for winuae (030/040/060):
http://eab.abime.net/showthread.php?t=46934&highlight=winuae+mmu
they are testing it with linux distros. lets wait till it really works, then we will see how much performance it costs. if you ask me id make mmu a non-default option in early startup. you could turn it on for debugging, regular user might sacrifice it for speed.

Quote:

Finally, the next time that Intel's changes the socket you must begin another project. FPGAs give much more stability and convenience from this PoV.

you need to develop a board for the fpga along with support chipset and solder it. fpga pinout may change as well while the development. on the other hand the solution im in favour of only demands a standard interface like pci>amiga, and then you can connect just any pci equipped module running 68k emu to your amiga. x86 side code doesnt even need to be necessarily updated, as the standards are backward compatible. with the fpga you might need to adopt your implementation when switching verndors or chip generations.

Quote:

OlafS25 wrote: @cdimauro How do you calculate the costs for the FPGAs?

Here are:

EP4CE30F23C8N 30K LE, 328 I/O BGA, 400Mhz, €48,16
EP4CE30F23C6N 30K LE, 328 I/O BGA, 500Mhz, €50,89

That's for a customer that wants to pick just one.

There are discounts for quantity, so the price will go down, especially it you take into account that using the same FPGA doesn't require to buy big batches. For example, with 100 of them we can build 33 motherboards, which is an affordable investment.
Quote:

Is it not becoming too expensive?

It depends strictly on how much do you want to gain from it.

However the target has raised from an initial 200€ prevision, to about 300€ for the last configuration.

But you have plenty of power under you hands, as I've explained before, and a lot of room for enhancements.
Quote:

I know (from Natami) that FPGA is the most expensive component on such a board and you want to use even a couple of them.

We want to use three. :)

Yes, I know that Natami wasn't cheap. I don't know why, since Alberto has stated that 300€ for the last configuration is the target price.

@cdimauro

300€? That sounds good. Natami was much much higher so I am a little surprised. For 300€ there might be a market (if it offers more than f.e. FPGA Arcade).

Quote:

wawa wrote: @cdimauro Quote: So you're interested on having a more powerful CPU that can supplant the one used on an original Amiga, leaving the latter to just act as a "chipset". Right? exactly. think of it like the reverse of the amiga on a pci card idea, which is more sensible imho, after all you want an amiga on your desk, right? also think of it like a sort of amithlon, only equipped with the legacy hardware for full backwards compatibility. isnt amithlon supposed the best what may have happened to amiga? well, what im talking about is even better!

I never used Amithlon, so I don't know, but I understand the concept.
Quote:

Quote: and that can be quickly implemented as a Picasso96 driver and AHI driver. ahi will be simple i think. chunky rtg a little more a challenge. i talked to thomas richter who implemented natami saga modes and he advised to completely resign on amiga os "monitors" approach. afair he said legacy ocs-aga software side philosophy is almost impossible to extend sensibly for chunky gfx. even p96/cgx carrying extended planar legacy had impact on performance due to it. also as both systems are closed source you will likely need to base your solution on what winuae provides, which of course is a general problem affecting all other projects too, aros68k for that matter.

Yes, that's the way to go. AROS is the o.s. to look at for TiNA.

However, I've just some ideas on how to implement chunky modes, that need to be verified. Thanks to FGPAs, implementation can change a lot before going in production.
Quote:

Quote: But Alberto needs to develop it, for his company. So we are just picking an opportunity to build a new Amiga-like hardware. Alberto will sell the know-how (and may be motherboards too) to his customers this way, while amigans can have a new toy to play with (that's an hobby, in the end). Think about ACube: they build a PowerPC motherboard with an FPGA because they sell it to their customers (industries), but OS 4 users can also buy it. that really intrigues me. i understand there might be a market for ppc industry boards, but 68k compatible softcore? okay thats still imaginable, but what about genuine amiga chipset? who needs it besides nerds?

I don't know how Alberto will use it, but I think that a lightweight multitasking o.s. that can also leave easy direct hardware access can have some value. NASA used it for the same thing, if I remember correctly.
Quote:

Quote: Using an x86 or ARM for such task will require a lot of work too, and I don't think that we can reach the same performance of a new 68K core, especially for ARMs, which are less powerful. look at amithlon on contemporary hardware. lets imagine the expansion will be an all in one module, i think its safe to say it will be able to emulate 68k numerous times faster than an consumer level fpga. id love to be proven wrong.

I never had the chance to use Amithlon, so I don't know, but in my opinion a modern 68K core can give a lot of power even compared to top x86 CPUs. Running 68K code, of course. We'll see it.
Quote:

Quote: You want MMU support too, but you know that emulating it greatly slows down performance, because there's no x86 or ARM (virtualization) technology that can be used to accomplish this task, since 68K MMUs were very different from the ones implemented by x86s or ARMs. toni wilen is just about to finish the complete mmu support for winuae (030/040/060): http://eab.abime.net/showthread.php?t=46934&highlight=winuae+mmu they are testing it with linux distros. lets wait till it really works, then we will see how much performance it costs.

It's interesting because we can have a reference for MMU implementation. I read something, and it seams that the JIT is disabled, so performance will be very slow right now.
Quote:

if you ask me id make mmu a non-default option in early startup. you could turn it on for debugging, regular user might sacrifice it for speed.

May be we can have a consumer and a developer firmware.

I already exposed my idea to the Natami team some years ago: we can have different firmwares for different computing sceneries.
Quote:

Quote: Finally, the next time that Intel's changes the socket you must begin another project. FPGAs give much more stability and convenience from this PoV. you need to develop a board for the fpga along with support chipset and solder it. fpga pinout may change as well while the development.

Yes, it can happen, but I think it's a normal thing for such projects.
Quote:

on the other hand the solution im in favour of only demands a standard interface like pci>amiga, and then you can connect just any pci equipped module running 68k emu to your amiga. x86 side code doesnt even need to be necessarily updated, as the standards are backward compatible.

The problem is that it's what Alberto doesn't want / need for his company.

I don't think that we can take into consideration this solution.
Quote:

with the fpga you might need to adopt your implementation when switching verndors or chip generations.

Vendors like Altera are top companies that gives us relative solid guaranties even if we think to move to cheaper of bigger FPGAs. I already took at its catalog, so I'm quite confident that there'll be no surprises.

Quote:

OlafS25 wrote: @cdimauro 300€? That sounds good. Natami was much much higher so I am a little surprised. For 300€ there might be a market

Yes, I think so. It's not cheap for an Amiga hardware (PCs are unbeatable), but affordable for an hobby.
Quote:

(if it offers more than f.e. FPGA Arcade).

TiNA will be an open project, so other platforms can be implemented as well. There's plenty of resources to make it suitable for complex systems.

@cdimauro

Quote:

It's interesting because we can have a reference for MMU implementation. I read something, and it seams that the JIT is disabled, so performance will be very slow right now.

if you read exactly, toni says its going to stay like this. jit and mmu are mutually exclusive, so okay, some sort of fpga implementation may be a better choice if achievable. yet i heard its very complicated, gunnar of natami team didnt even want to talk of it.

Quote:

The problem is that it's what Alberto doesn't want / need for his company. I don't think that we can take into consideration this solution.

for gods sake, im not trying to talk you into it. im only portraying an interesting alternative. actually i prefer you follow your roadmap as is, the decisions start to sound reasonable. would love it become reality.

Last edited by wawa on 25-Jan-2013 at 10:45 AM.

@cdimauro

Quote:

The Minimig one (which is TG68 based, if I remember correctly), since we have found nothing else. A forum user has tried to contact Minimig v2 / FPGArcade people, but hasn't received any answer, unfortunately.

the cooperation here might be most desired. fpga arcade seems to be mature baseboard solution, allowing for extensions. lets not underestimate the whole finetune components and interface work. the problem is it gets constantly delayed. one reason is the sourcing components and fabrication problems as it seems, another perhaps that mikej tries to hit too many targets at once. even though claims the fabrication problems been solved at last im not sure if i trust it.

the company like this of your friend may help with with fabrication while benefit of building upon available baseboard and chipset+cpu core. just an appropriate daughterboard with huge fpga would cost much less time and ressources, and may be most suficcient to develop 68k core individually.

one would have to work out an agreement though. this is what i mean by modularizing projects. im aware though even if reasonable from the pov of amiga community it might be another alternative that doesnt fit into the your friends concept.

@cdimauro

In my opinion, in all your articles about the modern Amiga you are never unbiased thus not reliable at all!
You are very negative and always come to the same expected conclusions!

You are boring (seccante) while I would prefer something more constructive instead.

Regards

_________________
SAM440EP-FLEX @ 733 Mhz, AmigaOS 4.1 Update 1

@cdimauro

Quote:

Yes, I know. But if you talk about licenses, even a lighter with a sticker applied can be legitimately called "Amiga".

Of course.

Quote:

Right, but the Amiga platform itself died.

Too much talk about "dead" platform. I think it is rather un-dead.

Quote:

but not about a complete machine, which is what Amiga was in the past.

You favourite computer - X1000 - is complete machine.

@pavlor
Quote:

Too much talk about "dead" platform. I think it is rather un-dead.

Quote:

You favourite computer - X1000 - is complete machine.

you mean actually un-complete? guess why..

@wawa

Quote:

you mean actually un-complete? guess why..

Complete in sense specialy designed, manufactured and bundled with OS by default, exactly like Amiga models from Commodore era.

If hardware development could be very intersting, from the other side they haven't no software to run.
Oh yes, TiNA is retro compatible with AmigaOS 3 or AROS 68k, but this synthetic cpu could be not enough powerfull to run 3D complex engine like some 3D Games, 3D demos or 3D rendering.
And maybe not enough power to surf in internet with OWB and Netsurf the require much power on 68k side.

I think this project hasn't future because AROS users have their x86 machines, Classic users have their classic machines and moders users have their emulated machines.

The Team has moneys and times to spend for their dreams, so let them dream

_________________

@cdimauro
Quote:

EP4CE30F23C6N 30K LE, 328 I/O BGA, 500Mhz, €50,89

from alterashop you have to buy 60 minimun pieces at 59,87 USD for a tolal or 3,592 USD..

http://www.buyaltera.com/scripts/partsearch.dll?Detail&name=EP4CE30F23C6N-ND

here you can see that minimun order is 60
http://www.buyaltera.com/scripts/partsearch.dll

Last edited by Seiya on 25-Jan-2013 at 08:27 PM.
Last edited by Seiya on 25-Jan-2013 at 08:23 PM.
Last edited by Seiya on 25-Jan-2013 at 08:22 PM.

_________________