Poster | Thread |
BigD
| |
Future 68k development and TMSC Fabs Posted on 12-Jun-2021 0:03:05
| | [ #1 ] |
|
|
|
Elite Member |
Joined: 11-Aug-2005 Posts: 7307
From: UK | | |
|
| I've just started researching TSMC the Taiwan chip fabricator that pretty much dominates the manufacture of Advanced Logic Chips (ALCs) today!
Quote:
Intel continued to hold its position as the world’s largest semiconductor company until 2017 when Samsung overtook it by revenues, and TSMC caught up on manufacturing process technology. |
Source:https://en.wikipedia.org/wiki/TSMC
Is there any way to harness this new powerhouse in global chip technology? We are all seemingly hoping for a port of AmigaOS to x86-64 and/or ARM but maybe we can now design a new 68k chip and get TSMC to mass produce it? Could the Apollo Team do this if they wanted? What is possible with 68k? I know PPC design is open source now but what is the situation with 68k?
What do people think? The falling apart of Moore's Law and lessening of Intel dominance has opened up an opportunity for 68k I think.Last edited by BigD on 12-Jun-2021 at 12:04 AM.
_________________ "Art challenges technology. Technology inspires the art." John Lasseter, Co-Founder of Pixar Animation Studios |
|
Status: Offline |
|
|
simplex
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 0:29:13
| | [ #2 ] |
|
|
|
Cult Member |
Joined: 5-Oct-2003 Posts: 896
From: Hattiesburg, MS | | |
|
| @BigD
(A) I really appreciate being able to choose more than one option.
(B) I think there's a market because the technology industry has diversified. In particular, CPUs for embedded computers are often now more powerful than anything the Amiga ever had, e.g., a Raspberry Pi 4 with PiOS makes for a fairly amazing computing experience, considering its cost anyway.
(C) Motorola and Freescale were naturally more interested in the higher-margin embedded moneymaking market, and that was a problem for the development of 68k and even PPC on the desktop.
(D) All that aside, I don't think the Amiga will make a comeback, and definitely not on a non-existent but merely hypothesized 68k CPU. _________________ I've decided to follow an awful lot of people I respect and leave AmigaWorld. If for some reason you want to talk to me, it shouldn't take much effort to find me. |
|
Status: Offline |
|
|
AmigaMac
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 1:25:36
| | [ #3 ] |
|
|
|
Super Member |
Joined: 26-Oct-2002 Posts: 1094
From: 3rd Rock from the Sun! | | |
|
| @BigD
Why can’t 68k make a comeback?! _________________
|
|
Status: Offline |
|
|
matthey
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 2:01:44
| | [ #4 ] |
|
|
|
Super Member |
Joined: 14-Mar-2007 Posts: 1968
From: Kansas | | |
|
| BigD Quote:
I've just started researching TSMC the Taiwan chip fabricator that pretty much dominates the manufacture of Advanced Logic Chips (ALCs) today!
Is there any way to harness this new powerhouse in global chip technology?
|
There are many fabless chip development businesses today. It's never been easier or cheaper, at least before COVID-19 disruptions.
BigD Quote:
We are all seemingly hoping for a port of AmigaOS to x86-64 and/or ARM but maybe we can now design a new 68k chip and get TSMC to mass produce it?
|
There may be other fab options to consider than TSMC. A 68k Amiga SoC would not be using a 5nm process at the start. It may be better to look at an older process which offers a cheaper price/transistor. There are older fabs which will need to shut down or retool if they don't find enough business that offer very good value.
BigD Quote:
Could the Apollo Team do this if they wanted?
|
Gunnar doesn't want to and he speaks for the team. The N68k and SAGA were not originally written by him though. Others could likely negotiate to obtain HDL sources to have something to start with. I would want someone very experienced in processor design to lead development anyway.
BigD Quote:
What is possible with 68k?
|
The 68060 and the Apollo core are the most advanced 68k designs. Both designs have strong single thread performance for the frequency compared to their competition which is a good trait but they may not clock up as much as some simpler designs. A quick and cheap ASIC may not gain much performance but just taking advantage of half way modern processes for the 68060 or improved routing for the Apollo Core should result in a several times performance boost. Both would probably benefit from an on chip L2 cache at that point and gain further performance from that. We don't know the limits of a good modern 68k design but it should be close to that of x86. The 68060 was a better processor than the Pentium both of which were the first of the modern superscalar CISC processors with RISC cores and deep enough pipelines to be clocked up.
BigD Quote:
I know PPC design is open source now but what is the situation with 68k?
|
The 68k trademarks are expired. Any new 68k design should be fine to develop.
BigD Quote:
What do people think? The falling apart of Moore's Law and lessening of Intel dominance has opened up an opportunity for 68k I think. |
The 68k would compete more against ARM than Intel. ARM moved up with AArch64 to compete better in performance with Intel forgetting about the good code density of Thumb2 which allowed them to gain the embedded market. There may be room for the 68k to sneak in and compete against Thumb2 again. The embedded market is certainly big enough for more players and there are surprisingly few startups trying to compete in this space considering how little of capital is required and how long good chip designs can last now with Moore's Law ending.
Last edited by matthey on 12-Jun-2021 at 02:07 AM.
|
|
Status: Offline |
|
|
matthey
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 2:29:21
| | [ #5 ] |
|
|
|
Super Member |
Joined: 14-Mar-2007 Posts: 1968
From: Kansas | | |
|
| simplex Quote:
(C) Motorola and Freescale were naturally more interested in the higher-margin embedded moneymaking market, and that was a problem for the development of 68k and even PPC on the desktop.
|
Actually, the embedded market has lower margins than the desktop market which has lower margins than the server market. Embedded market margins are lower but high volumes can make up for it. It's kind of like with oil companies which have low margins but high volumes. Some people complain about oil company profits while other companies have several times the profit margins and are robbing them blind.
simplex Quote:
(D) All that aside, I don't think the Amiga will make a comeback, and definitely not on a non-existent but merely hypothesized 68k CPU. |
There are 68k cores and Amiga compatible custom chips which exist! They are developed in FPGA in a similar way to how ASICs are designed. They have been enhanced in some cases further than what CBM was able to develop and use. There are some very talented HDL FPGA programmers already in the Amiga community. The technology is real and here whether the Amiga owners keep ignoring it or embrace it.
Last edited by matthey on 12-Jun-2021 at 03:03 AM. Last edited by matthey on 12-Jun-2021 at 02:31 AM.
|
|
Status: Offline |
|
|
simplex
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 5:57:30
| | [ #6 ] |
|
|
|
Cult Member |
Joined: 5-Oct-2003 Posts: 896
From: Hattiesburg, MS | | |
|
| @matthey
Quote:
Actually, the embedded market has lower margins than the desktop market which has lower margins than the server market. Embedded market margins are lower but high volumes can make up for it. |
Sorry, I abused terminology. I basically meant that Motorola / Freescale was making more profit from embedded.
Quote:
There are 68k cores and Amiga compatible custom chips which exist! |
Are they the sort of thing that someone who isn't simply into retro computing would want? What turned me off to Vampire was not just their pointing out (entire reasonably) that running Linux & therefore Firefox on the Apollo Core was unreasonable, but their apparent lack of any interest in developing something to that effect. I don't mean to misspeak for them; maybe I just misunderstood._________________ I've decided to follow an awful lot of people I respect and leave AmigaWorld. If for some reason you want to talk to me, it shouldn't take much effort to find me. |
|
Status: Offline |
|
|
billt
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 7:26:45
| | [ #7 ] |
|
|
|
Elite Member |
Joined: 24-Oct-2003 Posts: 3205
From: Maryland, USA | | |
|
| @BigD
TSMC is one of many great fabs. How much money you got??
What about this alternative... Maybe not as advanced, but consider value to price ratio and the available market... See what you can do with tg68k, ao68000 or Suska CPU cores... What other open options are there?
https://efabless.com/open_shuttle_program/1
Last edited by billt on 12-Jun-2021 at 07:31 AM.
_________________ All glory to the Hypnotoad! |
|
Status: Offline |
|
|
Rose
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 8:12:36
| | [ #8 ] |
|
|
|
Cult Member |
Joined: 5-Nov-2009 Posts: 982
From: Unknown | | |
|
| @BigD
Quote:
What do people think? The falling apart of Moore's Law and lessening of Intel dominance has opened up an opportunity for 68k I think. |
I think that like usually, you don't have a clue. |
|
Status: Offline |
|
|
BigD
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 8:52:57
| | [ #9 ] |
|
|
|
Elite Member |
Joined: 11-Aug-2005 Posts: 7307
From: UK | | |
|
| @Rose
Quote:
Rose wrote: @BigD
Quote:
What do people think? The falling apart of Moore's Law and lessening of Intel dominance has opened up an opportunity for 68k I think. |
I think that like usually, you don't have a clue. |
Why can’t the FPGA designs we have be tweaked and fabbed on a ASIC chip? Why can’t this be scaled up and eventually reach 5nm size and beyond? Why not try?Last edited by BigD on 12-Jun-2021 at 08:54 AM. Last edited by BigD on 12-Jun-2021 at 08:53 AM.
_________________ "Art challenges technology. Technology inspires the art." John Lasseter, Co-Founder of Pixar Animation Studios |
|
Status: Offline |
|
|
Rose
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 14:15:54
| | [ #10 ] |
|
|
|
Cult Member |
Joined: 5-Nov-2009 Posts: 982
From: Unknown | | |
|
| @BigD
For at least last 10 years it has been explained at least twice a year why it's not feasible.
For once, how about YOU explaining why it would be a great idea and economically viable.
|
|
Status: Offline |
|
|
matthey
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 16:07:19
| | [ #11 ] |
|
|
|
Super Member |
Joined: 14-Mar-2007 Posts: 1968
From: Kansas | | |
|
| simplex Quote:
Sorry, I abused terminology. I basically meant that Motorola / Freescale was making more profit from embedded.
|
Yes, the embedded market has been the consistent source of profits for Motorola/Freescale/NXP. Even though their embedded market share has shrunk since the 68k days, the embedded market has grown enough to allow them to have a healthy position even with ARM dominating the market. In fact, they are content to pay license fees to ARM for ISA and core development which once allowed them to dominate the market.
simplex Quote:
Are they the sort of thing that someone who isn't simply into retro computing would want? What turned me off to Vampire was not just their pointing out (entire reasonably) that running Linux & therefore Firefox on the Apollo Core was unreasonable, but their apparent lack of any interest in developing something to that effect. I don't mean to misspeak for them; maybe I just misunderstood. |
The classic Linux kernel requires a MMU with virtual addressing which the Apollo core does not have. Embedded Linux can use a modified kernel adding functionality like the Amiga scatter loader for a flat memory model like the AmigaOS uses as it avoids address translation overhead and jitter. AmigaOS 4 has moved toward the slower Linux style of virtual addressing requiring an MMU, adding the Linux SLAB page based memory allocator and supporting virtual memory paging but the AmigaOS does too much memory sharing to benefit from program isolation where each program has its own virtual memory pages so errant pointers won't affect other programs.
|
|
Status: Offline |
|
|
matthey
| |
Re: Future 68k development and TMSC Fabs Posted on 12-Jun-2021 16:22:45
| | [ #12 ] |
|
|
|
Super Member |
Joined: 14-Mar-2007 Posts: 1968
From: Kansas | | |
|
| billt Quote:
What about this alternative... Maybe not as advanced, but consider value to price ratio and the available market... See what you can do with tg68k, ao68000 or Suska CPU cores... What other open options are there?
|
The authors of the FPGA Arcade CPU core (not tg68k) and N68k core (Apollo Core before it was superscalar) have shown interest in making their cores open source. Both of these authors are professional HDL programmers and have experience in CPU core programming. The N68k VHDL sources are very neat and well documented.
|
|
Status: Offline |
|
|
fishy_fis
| |
Re: Future 68k development and TMSC Fabs Posted on 13-Jun-2021 13:02:02
| | [ #13 ] |
|
|
|
Elite Member |
Joined: 29-Mar-2004 Posts: 2156
From: Australia | | |
|
| Lol.
TMSC new? Haha. They've been *the* fab for many years. No-one can compete currently. What research have you been doing? Doesn't take much to know this, that they've been around for many, many years, or to discover theyre well and truly tapped out capacity for a long time and into the foreseeable future.
|
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: Future 68k development and TMSC Fabs Posted on 13-Jun-2021 13:30:35
| | [ #14 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12795
From: Norway | | |
|
| @matthey
Quote:
but the AmigaOS does too much memory sharing to benefit from program isolation where each program |
the amount of available graphic/texture memory has increased hugely, by adapting virtual memory. remember the memory only needs to be available while changing something.
The expectation between new and old programs does not need to be the same as well.
We are not running this on a 7mz computer, so we can afford a better memory system, however as program you be considering pre-allocation, using the memory on stack, avoid variable augments functions as they map to FunctionTagList(a,&myTags) functions that does the same, so your kind looping threw the arguments twice, when using variable arguments functions, FuncionTags(a,…, TAG_END).
Last edited by NutsAboutAmiga on 13-Jun-2021 at 02:23 PM. Last edited by NutsAboutAmiga on 13-Jun-2021 at 02:16 PM. Last edited by NutsAboutAmiga on 13-Jun-2021 at 01:42 PM. Last edited by NutsAboutAmiga on 13-Jun-2021 at 01:32 PM. Last edited by NutsAboutAmiga on 13-Jun-2021 at 01:31 PM.
_________________ http://lifeofliveforit.blogspot.no/ Facebook::LiveForIt Software for AmigaOS |
|
Status: Offline |
|
|
LarsB
| |
Re: Future 68k development and TMSC Fabs Posted on 13-Jun-2021 13:55:09
| | [ #15 ] |
|
|
|
Regular Member |
Joined: 29-Jul-2019 Posts: 104
From: Unknown | | |
|
| @BigD TMSC are playing in another league. We are talking about much higher volumes which are indeed needed to finance such a venture. And its also a technical problem. It was said that the Amiga design is completely asynchrone. But it isnt. You can easily find out. Just take UAE and tweak the perfrormance value to the max. You will find our that most software is crashing at some point and the whole systems gets out of synchronisation.
Last edited by LarsB on 13-Jun-2021 at 01:57 PM. Last edited by LarsB on 13-Jun-2021 at 01:56 PM.
|
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: Future 68k development and TMSC Fabs Posted on 13-Jun-2021 14:15:50
| | [ #16 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12795
From: Norway | | |
|
| @LarsB
Quote:
Just take UAE and tweak the perfrormance value to the max. You will find our that most software is crashing at some point and the whole systems gets out of synchronisation. |
Yes you have problem if a program unable to process the message in the queue in time, that it is unable process, because another task is taking up all CPU resources, if message is not reply quickly, it result in some part of system halting, as message is not necessary allocated when its created, it be pre allocated messages that are recycled on replymsg(), if there are no free messages to send, then service that sends the messages also stops, now the sender and receiver is blocked, because of a task that has taken up all the resources, the OS can recover, if the runaway program quits, but most likely it is stuck.
This extremely annoying on system that depends on a running USB sub-system.
You can however remedy it, by setting a lower task priority on the aggressive task.Last edited by NutsAboutAmiga on 13-Jun-2021 at 02:22 PM.
_________________ http://lifeofliveforit.blogspot.no/ Facebook::LiveForIt Software for AmigaOS |
|
Status: Offline |
|
|
DiscreetFX
| |
Re: Future 68k development and TMSC Fabs Posted on 13-Jun-2021 15:15:00
| | [ #17 ] |
|
|
|
Elite Member |
Joined: 12-Feb-2003 Posts: 2481
From: Chicago, IL | | |
|
| @BigD
No of the options are pro Vampire V4 so I won’t be participating in this poll. Have you ever seen or tried the V4? The fact that it’s FPGA based is an advantage not a disadvantage. _________________ Sent from my Quantum Computer. |
|
Status: Offline |
|
|
ne_one
| |
Re: Future 68k development and TMSC Fabs Posted on 13-Jun-2021 17:43:36
| | [ #18 ] |
|
|
|
Cult Member |
Joined: 13-Jun-2005 Posts: 905
From: Unknown | | |
|
| I still can't fathom why these discussions always revert back to trying to reinvent 40 year old hardware.
Unless you plan to mass produce a low cost Amiga to tap into the nostalgia market why would you even bother? And setting aside the obvious legal challenges, you could do a decent job at that today through low cost emulation.
Anyone who wants to use a decades old platform already has lots of options.
If you really want the Amiga to evolve, the OS has to be rewritten - it's a software problem.
|
|
Status: Offline |
|
|
AmigaMac
| |
Re: Future 68k development and TMSC Fabs Posted on 13-Jun-2021 17:44:42
| | [ #19 ] |
|
|
|
Super Member |
Joined: 26-Oct-2002 Posts: 1094
From: 3rd Rock from the Sun! | | |
|
| Should both the Amiga and Atari communities band together and stand behind building a future behind 68k? _________________
|
|
Status: Offline |
|
|
Rose
| |
Re: Future 68k development and TMSC Fabs Posted on 13-Jun-2021 18:07:56
| | [ #20 ] |
|
|
|
Cult Member |
Joined: 5-Nov-2009 Posts: 982
From: Unknown | | |
|
| @AmigaMac
There is one tiny problem with that. People in Atari community are sane and not looking for comeback. |
|
Status: Offline |
|
|