Click Here
home features news forums classifieds faqs links search
6071 members 
Amiga Q&A /  Free for All /  Emulation /  Gaming / (Latest Posts)
Login

Nickname

Password

Lost Password?

Don't have an account yet?
Register now!

Support Amigaworld.net
Your support is needed and is appreciated as Amigaworld.net is primarily dependent upon the support of its users.
Donate

Menu
Main sections
» Home
» Features
» News
» Forums
» Classifieds
» Links
» Downloads
Extras
» OS4 Zone
» IRC Network
» AmigaWorld Radio
» Newsfeed
» Top Members
» Amiga Dealers
Information
» About Us
» FAQs
» Advertise
» Polls
» Terms of Service
» Search

IRC Channel
Server: irc.amigaworld.net
Ports: 1024,5555, 6665-6669
SSL port: 6697
Channel: #Amigaworld
Channel Policy and Guidelines

Who's Online
0 crawler(s) on-line.
 79 guest(s) on-line.
 2 member(s) on-line.


 Hypex,  matthey

You are an anonymous user.
Register Now!
 Hypex:  29 secs ago
 matthey:  1 min ago
 Hammer:  16 mins ago
 amigakit:  17 mins ago
 Rob:  19 mins ago
 pixie:  21 mins ago
 eliyahu:  35 mins ago
 Marcian:  50 mins ago
 keonhacaikto23456:  1 hr 9 mins ago
 BigD:  1 hr 37 mins ago

/  Forum Index
   /  Amiga OS4 Hardware
      /  32-bit PPC on FPGA
Register To Post

Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 Next Page )
PosterThread
michalsc 
Re: 32-bit PPC on FPGA
Posted on 18-Feb-2024 20:08:28
#341 ]
AROS Core Developer
Joined: 14-Jun-2005
Posts: 377
From: Germany

@matthey

Quote:
Does ARM64/BE (AArch64/BE) still work on newer ARM cores where ARM32 has been dropped? Have you tested Emu68 on one of these newer cores?


Of course I was not able to test Emu68 on everything. From what I have tested it works on all raspberry Pi's models supporting aarch64 as well as on RockChip RK3399. At some point in the future I will look at other cpus too.

 Status: Offline
Profile     Report this post  
agami 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 0:10:47
#342 ]
Super Member
Joined: 30-Jun-2008
Posts: 1691
From: Melbourne, Australia

@Hans

Organising a multi-national union is no more challenging than organising a local one, and there's no reason why volunteers couldn't join. Unions aren't just about money.

As a person who's written their fair share of NDAs, and reviewed and signed more than I'd care to remember, it is not uncommon to see a clause among the confidentiality clauses which precludes the parties from disclosing information, to include the NDA itself.
While I too think it's a little excessive, my impression of Ben is that he is partial to legal excessiveness.

From a legal recourse point of view, NDAs generally aren't worth the paper they're written on. Mostly it places the disclosee on the outside, and all that that entails. Meaning that in the case of Hyperion, it cuts a developer off from any resources shared by Hyperion, including access to sources. Whatever that's worth these days.

If I'm picking up what you're laying down, what you're saying is that there is no useful information guarded by the Hyperion NDA, and the community at large would not benefit from it being made available to the public.


_________________
All the way, with 68k

 Status: Offline
Profile     Report this post  
Hammer 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 2:17:31
#343 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5452
From: Australia

@Gunnar

Quote:

Gunnar wrote:
@Kronos

Quote:
A proper ASIC on a semi current process (so you have a chance to beat existing PPC)


With 90nm process from year 2003, you can reach up to 4 GHz clock.
So even a 20 year old process is perfectly good to make a very competitive ASIC


Intel Prescott (90 nm) is 31 stages pipelined microarchitecture and it's purpose-designed for high clock speed marketing.

4Ghz CPU will require front-end improvements to close the gap between 4 GHz CPU and slow system memory.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A ECS, KS 3.2, PiStorm/RPi 3A+/Emu68)

 Status: Offline
Profile     Report this post  
Hammer 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 2:46:10
#344 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5452
From: Australia

@ppcamiga1

Quote:

ppcamiga1 wrote:
@Karlos

ppc was chosen as new amiga cpu in 1995

PowerPC was selected by Escom-era AMIGA Technologies GmbH's Petro Tyschtschenko in cooperation with Phase 5, a subsidiary company of AS&S (Advanced Systems & Software).

Phase 5 was active in PowerPC upgrades for the Apple Mac market. https://everymac.com/upgrade_cards/phase5/

Amiga Technologies showed the 68EC030 @ 33 Mhz to 40 Mhz Amiga Walker prototype in late 1995 and early 1996 time period. This is like flogging a 386DX-40 class computer in the Pentium era.

Ex-Commodore Germany's personnel was the wrong Commodore group to continue the Amiga. Amiga's development involves custom ASIC/FPGA level R&D, not just PC clone box shippers like Escom. Escom is only interested in the Commodore PC clone brand.

Commodore Semiconductor Group was purchased by management and reformed as GMT Microelectronics. However, in 2001 the EPA shut the plant down. GMT ceased operations and was liquidated.

Last edited by Hammer on 19-Feb-2024 at 03:02 AM.
Last edited by Hammer on 19-Feb-2024 at 02:50 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A ECS, KS 3.2, PiStorm/RPi 3A+/Emu68)

 Status: Offline
Profile     Report this post  
Hammer 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 3:25:50
#345 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5452
From: Australia

@matthey

Quote:

matthey wrote:

The emu68 and PiStorm projects are fine to use ARM32/BE but a proper port of the AmigaOS to ARM should use ARM64/LE with reduced 68k and PPC AmigaOS compatibility (sandbox for compatibility). ARM32/BE is being removed in newer ARM cores and LE is needed for many GPUs on ARM devices as Hans has stated before. Do I think most AmigaOS users would move to ARM64/LE hardware? No. I think it would divide AmigaOS users between 68k, PPC and ARM flavors of AmigaOS with some choosing to convert to MorphOS or AROS x86-64 instead.

AArch64 lost the Aarch32 SETEND instruction for switching endianness from user mode.

This problem is not an issue for hypervisor or supervisor-level software.

Vulkan API requires LE.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A ECS, KS 3.2, PiStorm/RPi 3A+/Emu68)

 Status: Offline
Profile     Report this post  
Gunnar 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 6:50:47
#346 ]
Cult Member
Joined: 25-Sep-2022
Posts: 511
From: Unknown

@Hammer

Quote:
4Ghz CPU will require front-end improvements to close the gap between 4 GHz CPU and slow system memory.


Hammer as always you gave us the impression, that you not understand what you say.

 Status: Offline
Profile     Report this post  
Karlos 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 11:11:31
#347 ]
Elite Member
Joined: 24-Aug-2003
Posts: 4430
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@NutsAboutAmiga

Quote:

NutsAboutAmiga wrote:
@Karlos

64bit addressing is not necessary, AmigaOS4.1 use MMU for memory objects. In the upper memory.

I know you have been out of OS4.1 development team for while so you don’t know what’s going on.


This is a workaround. Its certainly better than nothing but it's not a substitute for flat 64-bit addressing in the same way 16 bit pointers with segments were no substitute for flat 32-bit addressing either.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
matthey 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 18:33:11
#348 ]
Elite Member
Joined: 14-Mar-2007
Posts: 2101
From: Kansas

Kronos Quote:

Judging from the last AmigaMetting in NeuMĂĽnster. all MorphOS users either had an PowerBook or G5 (sometimes both) and most of team Red was running X1000/5000 with a 1 or 2 midrange system (Peg2, A1-XE, SAM460).

I'd say most people with lower class HW have either given up or already upgraded to one of the above.

That fake 200MHz PPC would only be a "sidegrade" for people using Phase5 HW or emulation, so it's quite clear that it would fail hard if it was just that.


There are only low thousands of active MorphOS and AmigaOS 4 hardware users and the hardware is outperformed by a $35 USD RPi 4 with a more powerful $60 USD RPi 5 recently introduced. The only advantage I see of PPC Amiga NG hardware is the possibility of a better GPU but PPC CPU performance levels can be found in $3 USD SoCs. There are of course RPi competitors with better integrated GPUs but GPUs usually are underwhelming for low cost SoCs even though significantly better integrated GPUs are possible.

Kronos Quote:

As for "Gunnar" being "MEGA_RJ_MICAL" I see 3 options:

1. the similarities are only in your mind
2. MEGA has far more technical expertise as he had shown previously
3. MEGA was Gunnar's troll account from day 1


MEGA is sadly an intelligent person with mental issues. Much of the information Gunnar has posted here is copied from the AC forum and correct. There are lies, some copied from the AC forum, and major tech blunders made as I have pointed out in my previous post. Gunnar did not answer my questions because he can't and attacked instead as I predicted (I won't respond to his trolling unless he can answer my questions). The real Gunnar is a narcissist but isn't so irritable that he attacks everyone endlessly. Don't be fooled.

Kronos Quote:

A proper ASIC on a semi current process (so you have a chance to beat existing PPC) isn't even remotely viable unless you find a huge untapped market outside Amiga.

Everyone (except IBM with POWER) has dropped PPC in favor of ARM suggests that such a market does not exist.

Same goes for 68k only with a 10 year headstart.


No. THEA500 Mini is only scratching the surface of the retro 68k Amiga "gaming" market yet sold hundreds of thousands of units. Better 68k hardware could target 68k retro (gaming and productivity), hobby, embedded and educational markets. Amiga, Atari ST, 68k Mac, x68000, Sega Genesis/Mega Drive, NeoGeo and 68k Linux/BSD markets could be reached. Cost could be reduced and performance/cost vastly improved from THEA500 Mini.

Either 68k or PPC ASICs could be produced. Good deals could likely be found to obtain professionally developed and fairly modern PPC cores. Upgrading an OoO PPC core to compete with a $35 USD RPi 4 would be expensive and a $60 USD RPi 5 would be very expensive. PPC has inferior code density to AArch64 resulting in a larger footprint that increases the cost of hardware compared to ARM hardware. The 68k has better code density than AArch64 resulting in a smaller footprint, especially with the minimalist AmigaOS, and giving a competitive advantage. Retro gaming and hardware is hot right now and many low cost retro consoles have sold well. Most have used emulation reducing up front costs but an ASIC lowers production costs, can provide an improved experience and can use more modern software. Most consoles are oppressive closed systems that limit what can be done despite being high performance computers. The RPi Foundation may eventually take this market but they (ARM/Acorn/RISC OS) have minimal retro gaming appeal while the 68k Amiga alone has thousands of games. Pessimistic closed minded Amiga users proclaim there was no 68k Amiga market and some still stick there head in a hole ignoring the popularity of THEA500 Mini, FPGA hardware, WinUAE, PiStorm, etc.

michalsc Quote:

Of course I was not able to test Emu68 on everything. From what I have tested it works on all raspberry Pi's models supporting aarch64 as well as on RockChip RK3399. At some point in the future I will look at other cpus too.


Maybe Emu68 will be ok by using AArch64 and taking over the hardware. I would still recommend AArch64 LE for a native port of AmigaOS 4. Most integrated GPUs for ARM SoCs only support LE and a BE CPU with LE only GPU is asking for trouble.

The RPi 4 used only a 3 year old core design while the average for RPi 3-5 uses an average of a 5 year old core design. Estimates based on this data could predict a RPi using a 64 bit only core as early as 2025 and on average by 2027. The older RPi versions will likely remain available for years after new hardware is introduced which is adequate for Emu68 and the PiStorm but an AmigaOS 4 port could take more than a year and not being able to use the latest RPi hardware would be bad.

Big OoO ARM cores
2012 Cortex-A57
2016 Cortex-A72 (2019 RPi 4)
2016 Cortex-A73
2017 Cortex-A75
2018 Cortex-A76 (2023 RPi 5)
2019 Cortex-A77
2020 Cortex-A78
2021 Cortex-A710
2022 Cortex-A715 (64 bit only)

Little in-order ARM cores
2012 Cortex-A53 (2018 RPi 3)
2017 Cortex-A55
2021 Cortex-A510 (originally 64 bit only but refresh had optional 32 bit support)
2023 Cortex-A520 (64 bit only)

 Status: Online!
Profile     Report this post  
Kronos 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 19:03:58
#349 ]
Elite Member
Joined: 8-Mar-2003
Posts: 2584
From: Unknown

@matthey

Wether some random ARM is faster then the average NG PPC is irrelevant to the question wether a fake 200MHz PPC would be a down, up or sidegrade for the average NG PPC user.

"Gunnar"'s post here are 100% in line with what I would expect from the real one.
So he is the real thing or he is even better then the real thing


As for "THEA500 Mini" selling 6 digits, I got my doubts, but again 100% irrelevant as the discussion was about an hypothetical 1GHz++ 68k or PPC ASIC which would make no sense for that product since you would still need an emulation layer.
A basic 68000 ASIC on the other hand wouldn't be able to deliver the extra features.
In the end there is no reason to believe why either would be cheaper than the off the shelf ARM, hence no reason to go that way.

Also don't forget that rPi used off the shelf SoCs for the 1-4 generations and only after selling 8(9?) digits they saw a point in going custom but still using existing cores,
I don't see the demand to redo the same as a rPIi does on a different architecture that would need serious investment just to make it viable.
RISC_V does have such a chance, but is "free" and already exists as viable cores.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

 Status: Offline
Profile     Report this post  
matthey 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 20:29:01
#350 ]
Elite Member
Joined: 14-Mar-2007
Posts: 2101
From: Kansas

Kronos Quote:

Wether some random ARM is faster then the average NG PPC is irrelevant to the question wether a fake 200MHz PPC would be a down, up or sidegrade for the average NG PPC user.


Right. There is no question that NG PPC is a downgrade from a $35 USD RPi 4.

Kronos Quote:

"Gunnar"'s post here are 100% in line with what I would expect from the real one.
So he is the real thing or he is even better then the real thing


I know Gunnar well enough from my Apollo Core team experience. I don't think the Gunnar here is the real Gunnar unless he has declined mentally.

Kronos Quote:

As for "THEA500 Mini" selling 6 digits, I got my doubts, but again 100% irrelevant as the discussion was about an hypothetical 1GHz++ 68k or PPC ASIC which would make no sense for that product since you would still need an emulation layer.
A basic 68000 ASIC on the other hand wouldn't be able to deliver the extra features.
In the end there is no reason to believe why either would be cheaper than the off the shelf ARM, hence no reason to go that way.


Adding a cycle exact 68000 core and/or using a 68000 FPGA core would be a good idea as high 68000 compatibility is desirable. A 68000 ASIC core could be used for I/O processing too. A 68000 core is relatively small even for a FPGA. There is an open source cycle exact 68000 core available although it may need more testing and verification.

Kronos Quote:

Also don't forget that rPi used off the shelf SoCs for the 1-4 generations and only after selling 8(9?) digits they saw a point in going custom but still using existing cores,
I don't see the demand to redo the same as a rPIi does on a different architecture that would need serious investment just to make it viable.
RISC_V does have such a chance, but is "free" and already exists as viable cores.


The RPi Foundation was initially surprised by the demand for the RPi. The first SoC only had a single scalar CPU core but pushing the low cost envelope was a game changer. The RPi Foundation had the advantage of commodity SoCs but there is no competitive ARM production advantage other than economies of scale. Investment is required for an ASIC and there is a gamble whether there is enough demand for mass production but the necessary investment is within reach of small businesses, provided they are not just IP squatters.

 Status: Online!
Profile     Report this post  
OneTimer1 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 20:37:13
#351 ]
Super Member
Joined: 3-Aug-2015
Posts: 1000
From: Unknown

@Gunnar

Quote:


With 90nm process from year 2003, you can reach up to 4 GHz clock.


Well if you have a 1GHz 68k core with ~512MB of cache, you might attract 68k fans in the Amiga, Atari, Macintosh and SinclairQL retro community, but the rest of their hardware doesn't need AGA.

And maybe you could find industrial 68k users, who wont recompile their code ...

But it would remain retro even with a 4000 Mips, the old software remains old and can't compete with modern systems ...

----

AmigaNG wasn't only about faster CPUs, it meant HiRes GFX, 16-Bit audio, 3D APIs and acceleration, Ethernet support, Web Browsers, Printers, better Filesystems, UTF8 ...
Just think about all the programs you had on a standard AOS3.9 distribution and see them on a level where they could nearly compete with a standard Windows or Linux installation.

Last edited by OneTimer1 on 19-Feb-2024 at 08:45 PM.

 Status: Offline
Profile     Report this post  
Kronos 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 21:00:19
#352 ]
Elite Member
Joined: 8-Mar-2003
Posts: 2584
From: Unknown

@matthey

Look at THEA500Mini or rPi are priced, do some reasonable guesswork and realize that the ARM SoC can't be more than 5-10$.

Thats what you would need to compete against. Both products are 100% CPU agnostic on the SW side, so sure you could enter that market with an 68k, PPC, MIPS or x86 SoC.
But that also means that those have 0 advantage from not being ARM so would either compete on price or performance.
Competing on performance means a good multicore design running on a current node, not viable when you are the new kid,
Competing on price would require large economics of scale hitting before you have even come close to hitting those numbers.

A cycle correct 68000 would only fix a small amount of issues in THEA500Mini while adding complexity (since it won't be able to run the extra and you'd also need a HW/FPGA chipset). A non fix for a non issue.

There isn't really much demand for 68k outside of retro computing, which always also need additional HW to be emulated, something that can be done with 1 chip ARM board, a bit of SW and x variations of plastic cases for each 68000 (and non 68000) system you want to mimic.

FreeScale could have done an rPi style HW back with ColdFire, they could have done it with some of the later PPC (one might say bPlan's EFIKA did aim in that direction) but they failed to realize that market.
The last meaningful updates to 68k or PPC core were decades ago and you can't expect to leapfrog all that by putting old or untested hobbiest core on a smaller node.
And you surely can't expect the markets that buying ARM in masses to switch just because you have something that MIGHT be almost as good for almost as cheap.


Ergo, Amiga fantasy land ist still going strong 30 years later.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

 Status: Offline
Profile     Report this post  
Kronos 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 21:03:10
#353 ]
Elite Member
Joined: 8-Mar-2003
Posts: 2584
From: Unknown

@OneTimer1

Quote:

OneTimer1 wrote:


And maybe you could find industrial 68k users, who wont recompile their code ...


Those guys had over 30 years for switching and the option to go ColdFire.

So that market might be a few 100% cycle correct pin compatible 68030/40 chips to fix broken HW, but surely not to build any new products.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

 Status: Offline
Profile     Report this post  
Karlos 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 22:00:37
#354 ]
Elite Member
Joined: 24-Aug-2003
Posts: 4430
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

There is zero market for a high end ASIC 68K solution outside of the retro enthusiast community that are both allergic to emulation in any form and want the hypothetical machine of their dreams from 30 years ago. They are quite literally dying for it.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
agami 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 22:32:35
#355 ]
Super Member
Joined: 30-Jun-2008
Posts: 1691
From: Melbourne, Australia

@Karlos

Quote:
Karlos wrote:
There is zero market for a high end ASIC 68K solution outside of the retro enthusiast community that are both allergic to emulation in any form and want the hypothetical machine of their dreams from 30 years ago. They are quite literally dying for it.

Not that long ago there was zero market for black pearls, until one day there wasn't.

_________________
All the way, with 68k

 Status: Offline
Profile     Report this post  
kolla 
Re: 32-bit PPC on FPGA
Posted on 19-Feb-2024 23:01:01
#356 ]
Elite Member
Joined: 21-Aug-2003
Posts: 2975
From: Trondheim, Norway

@Karlos

What’s funny is how active m68k support in Linux (and NetBSD) is, as well as in gcc and llvm tool chains, all things considered. If a fast “proper” 68k board showed up tomorrow, it would have instant support from those camps.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

 Status: Offline
Profile     Report this post  
Karlos 
Re: 32-bit PPC on FPGA
Posted on 20-Feb-2024 1:01:27
#357 ]
Elite Member
Joined: 24-Aug-2003
Posts: 4430
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

I didn't say there was no market, I just don't think it's going to extend far beyond retro/alt computing fans. And sure, personally speaking, yes, I'd love one.

However, who else has a need for a single core 32-bit scalar-only CPU? I'm sure there might be some embedded cases but they are probably already comfortable with ARM or even RISC-V.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
Hammer 
Re: 32-bit PPC on FPGA
Posted on 20-Feb-2024 1:40:36
#358 ]
Elite Member
Joined: 9-Mar-2003
Posts: 5452
From: Australia

@Gunnar

Quote:

Gunnar wrote:
@Hammer

Quote:
4Ghz CPU will require front-end improvements to close the gap between 4 GHz CPU and slow system memory.


Hammer as always you gave us the impression, that you not understand what you say.

For example, Intel Prescott Model 524 (3.06 Ghz) has a 1 MB L2 cache while Model 670 (3.8 Ghz) has a 2 MB L2 cache.

Your AC68080 doesn't have a 1 MB L2 or 2 MB L2 cache to close the gap between a 4 GHz CPU and slower system memory.

ASUS P5GDC-Pro LGA 774 (Pentium 4) supports 64-bit DDR2-675 memory modules which is a 5.4 GB/s memory bandwidth in single channel (64-bit) mode or 10.8 GB/s in dual channel (128-bit) mode.

Your Vampire 4+ SA has 609 MB/s memory bandwidth (BusTest ReadL) via 32-bit DDR3 (up to 1GB/s).

With a 90 nm process node,
IBM's PowerPC 970FX reached 2.7 GHz (April 2005) and 970MP reached 2.5 Ghz.

AMD's K8 Athlon 64 FX-57 (E4 stepping) reached 2.8 Ghz, and K8 Athlon 64 X2 6400+ (F3 stepping, Black Edition) reached 3.2 Ghz.

Intel Prescott (3.8 Ghz, 90 nm process node) is designed for very high clock speeds.
Pentium IV ALUs are double-dumped i.e. a single-cycle integer ALU fabricated in 90nm dual-Vt CMOS technology operating at 4 GHz in the 64bit mode with a 32bit mode latency of 7 GHz.


Last edited by Hammer on 20-Feb-2024 at 02:35 AM.
Last edited by Hammer on 20-Feb-2024 at 02:04 AM.
Last edited by Hammer on 20-Feb-2024 at 01:43 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A ECS, KS 3.2, PiStorm/RPi 3A+/Emu68)

 Status: Offline
Profile     Report this post  
matthey 
Re: 32-bit PPC on FPGA
Posted on 20-Feb-2024 1:52:27
#359 ]
Elite Member
Joined: 14-Mar-2007
Posts: 2101
From: Kansas

Kronos Quote:

Look at THEA500Mini or rPi are priced, do some reasonable guesswork and realize that the ARM SoC can't be more than 5-10$.


Sure. RPi Pico RP2040 SoC is less than $1 USD. Original RPi and RPi Zero SoC maybe $2 for the SoC. The RPi 3 SoC maybe $3, RPi 4 SoC maybe $5 and RPi 5 SoC maybe $10. A 68060+AA+ SoC would use fewer transistors than the $1 RP2040 SoC and is more interesting as it could support HDMI unlike the RP2040 which has no video output. Enhancements are in order, especially increased CPU caches, extra CPU cores (68060x2, 68000?), more modern I/O and a 3D GPU would increase this and likely be worthwhile but may push it into the $3-$7 range.

Kronos Quote:

Thats what you would need to compete against. Both products are 100% CPU agnostic on the SW side, so sure you could enter that market with an 68k, PPC, MIPS or x86 SoC.
But that also means that those have 0 advantage from not being ARM so would either compete on price or performance.


Avoiding ARM royalties is a competitive advantage for chip production. The royalties may not sound like much but they are significant in a competitive market.

Kronos Quote:

Competing on performance means a good multicore design running on a current node, not viable when you are the new kid,
Competing on price would require large economics of scale hitting before you have even come close to hitting those numbers.


There is competition for smaller in-order cores as well as large OoO cores. These are different markets as the OoO cores are many times larger than in-order cores and more expensive. I believe an enhanced in-order 68060 core could compete in the higher performance in-order market. The SiFive U74 core design is like the 68060 core design but the 68060 can execute more powerful instructions than RISC-V. A 3 DMIPS/MHz in-order core that is smaller and lower power than low end OoO RISC cores is interesting and should be cheaper to develop than OoO cores.

Kronos Quote:

A cycle correct 68000 would only fix a small amount of issues in THEA500Mini while adding complexity (since it won't be able to run the extra and you'd also need a HW/FPGA chipset). A non fix for a non issue.


I guess you haven't read reviews of emulation and FPGA hardware. A FPGA for chipsets is cheap as even a $10 FPGA is large enough for most 16 bit retro chipsets and good enough for 8 bit CPUs and chipsets. There are other options that may make this more economical like eFPGA blocks or using flash memory integrated in a FPGA for other purposes to minimize the chip count.

Kronos Quote:

There isn't really much demand for 68k outside of retro computing, which always also need additional HW to be emulated, something that can be done with 1 chip ARM board, a bit of SW and x variations of plastic cases for each 68000 (and non 68000) system you want to mimic.


A native CPU results in smaller footprint hardware that is cheaper to produce and could be just as versatile as an ARM SoC. Other developers may be interested in native 68k capabilities where no developers are interested in emulated 68k capabilities.

Kronos Quote:

FreeScale could have done an rPi style HW back with ColdFire, they could have done it with some of the later PPC (one might say bPlan's EFIKA did aim in that direction) but they failed to realize that market.
The last meaningful updates to 68k or PPC core were decades ago and you can't expect to leapfrog all that by putting old or untested hobbiest core on a smaller node.
And you surely can't expect the markets that buying ARM in masses to switch just because you have something that MIGHT be almost as good for almost as cheap.


ColdFire missed out on the 68k market by not being compatible enough. ColdFire was well liked and successful for the embedded market though. It is easy to add improved ColdFire support to the 68k which improves code density.

Efika may have been the best chance AmigaOS 4 had for affordable hardware. Low end PPC hardware has always been disappointing though. The PPC e300 CPU is based on the 603 which also was poor. The SoC is only clocked at 400MHz and there is only 128MiB of memory which is too little for fat PPC. A 68k system with these specs would be comfortable as they are more than high end 68k Amiga hardware specs today. Then there is the problem that not much PPC software will run with this spec while there are huge libraries of 68k software that comfortably runs at a fraction of this spec.

The largest 68k market would be retro and hobbyist initially. There is likely some demand for 68k, ColdFire, SuperH and BE embedded system replacements. General purpose embedded use may be an uphill battle but the embedded market is huge and low enough priced hardware is likely to attract some attention. RPi Foundation has been popular for embedded use more because it is cheap than because it is ARM where it saw initial success despite being a new and unknown player. The 68k has a good reputation for embedded use where it was number one for 32 bit cores. If a "MC" qualified 68060 core could be licensed, it would instantly garner respect in the embedded market.

Karlos Quote:

There is zero market for a high end ASIC 68K solution outside of the retro enthusiast community that are both allergic to emulation in any form and want the hypothetical machine of their dreams from 30 years ago. They are quite literally dying for it.


I agree that a high end OoO 68k ASIC would be a mistake. I would aim much lower like 1-2GHz in-order CPU cores.

Karlos Quote:

I didn't say there was no market, I just don't think it's going to extend far beyond retro/alt computing fans. And sure, personally speaking, yes, I'd love one.

However, who else has a need for a single core 32-bit scalar-only CPU? I'm sure there might be some embedded cases but they are probably already comfortable with ARM or even RISC-V.


The original RPi and Pi Zero SoC have a 32 bit scalar CPU core and it was good enough to make the RPi Foundation what it is today. It would be more competitive to use at least a superscalar 32 bit CPU core like the 68060. I would prefer 2x in-order 68060 cores (and maybe a 68000 core) which are useful even without SMP (embedded systems often use AMP). I believe it is more likely to get SMP working in AmigaOS with custom hardware.

Last edited by matthey on 20-Feb-2024 at 08:35 PM.

 Status: Online!
Profile     Report this post  
V8 
Re: 32-bit PPC on FPGA
Posted on 20-Feb-2024 2:55:27
#360 ]
Regular Member
Joined: 30-Mar-2022
Posts: 134
From: Unknown

@Hans

Quote:
I see no need for that. For starters, an NDA doesn't stop anyone from doing their own independent work, and publicizing confidential information without permission is unethical except in very exceptional circumstances (of which there are none).


NDA's does not stop you directly but it does stop others from being able to work with you.

You understand that a secret NDA with litigation-happy Ben Hermans, no less, means that no one else will be able to work with or accept contributions from that person?

No one will accept or allow that person to contribute or be part of their project. You are at that point just begging to be sued by Hermans.


Sure, walking away IS an option but if you walk away you walk away from everything amiga.
You can realistically not join any other team or work on any other amiga flavor.


The best thing that could happen is if Hyperion declares all NDAs void.


@agami
Quote:
From a legal recourse point of view, NDAs generally aren't worth the paper they're written on.


Correct. But do you want to spend tens of thousands of dollars to prove that in a court of law?

Last edited by V8 on 20-Feb-2024 at 03:03 AM.
Last edited by V8 on 20-Feb-2024 at 02:58 AM.

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 Next Page )

[ home ][ about us ][ privacy ] [ forums ][ classifieds ] [ links ][ news archive ] [ link to us ][ user account ]
Copyright (C) 2000 - 2019 Amigaworld.net.
Amigaworld.net was originally founded by David Doyle