Overview Features Coding ApolloOS Performance Forum Downloads Products Order Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.
Please visit our Apollo-Discord Server for support.



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Documentation about the Vampire hardware

PowerPC In FPGA - How Much Sense Does It Make?page  1 2 3 

Wawa T

Posts 695
16 Dec 2019 20:18


Marian Nowicki wrote:

  PowerPC is something that used in Amiga for more than twenty years.
 

  by very minor group of people, to which i accidentally belong. i assume you have a lot experience with it yourself and know the whole lot of software ppc platform offers for amiga users?
 
 

  People want to use something that they use when they are young, that's ok.
 

 
  do you have any figures at hand? because every now and then individuals come along and claim what other people want. the problem is this that they make it up simply to back up their own demands. they dont even realize that they do.
 
 

  Even if today it is useless, so what?
 

  you are asking people who work mostly voluntaryli and have a lot of work on their hands to do something you personally want, even if it doesnt make sense, simply because you want it? you think its a decent attitude? do it yourself if you desire so.
 
 

  What is important is it was in their Amiga twenty years ago.
 

 
  the presence of ppc in amiga context is marginal, it likely doesnt even reach 1%. what you want is an amigaone along with xmos hardware, because it almost as relevant. simply get it and stop these rants.
 


Marian Nowicki
(Needs Verification)
Posts 22/ 1
17 Dec 2019 05:05


Motorola was not good enough to compete ALONE with Intel so Motorola have to cooperate with IBM and cancel 68k and 88k.
It was history and first PowerPC Mac was available in 1994 even before Comodore banckrupt.
This anti PowerPC crap should stop and in Amiga community we should cooperate.



Kyle Blake
(Needs Verification)
Posts 108/ 1
17 Dec 2019 05:35


If you want people to cooperate the first step is not to denounce justified views as "anti powerpc crap".

Think about it sensibly. People have said why it's not a good idea, and you react by saying you don't care, that you want it just because.

Nobody is here to please your personal whims. If you want it, learn FPGA work and PPC instruction set and make it.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
17 Dec 2019 08:04


Marian Nowicki wrote:

It was history ...
This anti PowerPC crap ...

You might not be aware but several of the APOLLO Core developers are IBM engineers which did develop / work on many generation of PowerPC Cores.
You can say that know the history well as we did experienced and lived it, we also are experts of PowerPC ASM and know by heart the PowerPC architecture.

First of all MOTOROLA was not competing with INTEL.
MOTOROLA and INTEL were not in the same market segments.

If you look back in computer history then there are clear trends.
At first the CPU design all favored a "CISC" design.
CISC main attributes are:
- Strong powerful instructions
- Very good code readability
- good match to how programmers think
- good code density
== Good for Programmers

- complex to pipeline
- complex to super scalar
- difficult to get high clock
- huge test effort
- long development cycle
== Bad for companies building chips

As you will recall Chip process technology did make fast progress at this time. Every other year the transistor size was shrunk smaller and smaller. This was great as smaller transistors means higher clock. But it also meant that new chip needed to be developed for the new process technology at an accelerating speed.

The Chip companies easily saw that while CISC is optimal for the programmer - it was complex to develop and this made it very hard for the companies to keep pace with the race of new transistor size technologies.

Chip companies did find a solution they designed RISC
RISC was a trade. The RISC main attributes are:

- not so strong instructions
- not good code readability
- not good match to how programmers think
- bad code density
= hard and not optimal for Programmers

- easier to pipeline
- easy to super scalar
- easier to get high clock
- medium test effort
- shorter development cycle
== Good for companies building chips

To make this crystal clear.
RISC was not done for programmer and users.
RISC was done to make live easy for the companies producing the chips.




Kyle Blake
(Needs Verification)
Posts 108/ 1
17 Dec 2019 08:37


I think it's probably good to clarify what market segments Motorola and Intel were in.
   
    Motorola had basically two markets, the first is high end powerful systems, Workstations. This is where the original 68000 lived in very early 80s.
   
    Then as 68000 becomes obsolete from workstations, it gets a new life in uses where being the fastest CPU is not important, only being cheap and reasonably useful. So it shows up in microcomputers, then in calculators and other small embedded things. Workstation role moves to 68020,040 etc but doesn't last because of SPARC etc taking over. So all that's left for 68K is embedded.
   
    Meanwhile Intel has got very lucky and become the industry standard for "normal" computers. Not small cheap home computers, and not specialist workstations, but the wide majority of computers the PC. Intel's role is to power the fastest possible "standard" computer in the $2000 range, and to sell the half-broken reject chips for the $500 range.
   
    So the time that x86 and 68K competed directly is very brief. 68K fell from Workstation heaven and only was in the Mac and Amiga because they were the half way point on it's fall down to it's final resting place, in a games console or 90s mobile phone.

Even then, it's only competition in the sense of two computers at a similar price. Nobody who needed DOS was going to buy a 68K computer.
   
    It's just not comparable. Intel was thinking "Compaq wants a chip that will make next year's computer 30% faster for the same price". Motorola was thinking "keeping up is too hard, just make what we already have cheaper".
   
    In that way PowerPC switch makes perfect sense, they had IBM to lean on. When PPC failed in the market they did the exact same thing again, just take existing chip and make it cheaper to put in modems or whatever.
   
  Intel is in the fast CPU business, Motorola is in the easy money business.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
17 Dec 2019 08:47


Kyle Blake wrote:

Intel is in the fast CPU business, Motorola is in the easy money business.

 
Thank you Kyle, you explained the history very good.
 
 
Regarding CISC/RISC/EPIC design trends.
 
The original goal for all companies was to make fast and great chips.
 
What companies did what we developers called this "change the subject"
The original goal to make a fast and good to code CPU was very hard to reach - so the engineer team tried to cheat "by changing the goal".
 
I always compared it to being in school and being ask about the book we had to read over the summer holidays.
If you could not answer the original question, you could try to cheat and speak about the new dog your parents bought this summer.
 
This is basically what RISC did.
The problem to SuperScalar CISC was a tough nut to crack.
And if you not could solve it - a possible cheat was to give up on CISC - and to something simpler.
 
"Teacher, no I did not finish my math homework, but look I painted you a nice picture instead" ....
 
The same repeated with EPIC, again a problem which was tough to be solved in engineering  - was simply not done at all.
Who can name the technical challenge which EPIC simply not does?
 
 
 
 


A1200 Coder

Posts 74
17 Dec 2019 09:20


All these years the hardware chip manufacturers have cheated us. 68k CPU is the best CPU design ever made, and we are not giving it up for a crappy PowerPC design. I can't understand why anyone would want a PowerPC CPU these days. AmigaOS 4 and PowerPC Amigas may be faster than the current 68k Amigas, but they are obsolete as well, and not compatible at all with 68k classic Amigas. Not with CPU and not with custom Amiga chipset. For a fraction of the price, you can as well get yourself a fully Amiga incompatible x64/ARM and run Linux+UAE on it.

When hardware is simple and elegant and we don't have several architectures to support the software side gets also much easier to manage, and this is very good, since there are very few Amiga developers around.

A new Amiga design has to satisfy three key points:

1) A 68k compatible CPU
2) OCS/ECS/AGA compatible chipset (no foreign GPU add-ons can be accepted)
3) AmigaOS 3 compatible OS



Kyle Blake
(Needs Verification)
Posts 108/ 1
17 Dec 2019 10:19


PowerPC fans are a kind of zealot - they have fought for it so hard they forgot what they were trying to achieve in the first place.

We were supposed to get our "PowerAmiga" in 1996, with Phase 5 bringing along the old machines. But it was delayed, and cancelled, and uncancelled, and delayed again...

Until it's the middle of the 2000s and it's called "AmigaOne" and it's a really buggy motherboard but OS4 will leave beta soon and finally it'll have been worth it, just another month or two!

And then Apple goes intel and PPC is totally dead buried and irrelevant on the desktop.

I'm reminded of VMS. Where did VMS go after VAX ended? Itanium!


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
17 Dec 2019 10:23


A1200 coder wrote:

All these years the hardware chip manufacturers have cheated us. 68k CPU is the best CPU design ever made.

 
Cheating is too strong.
The hardware manufactures gave options.
 
Chip manufacturers continued all the time to produce CISC chip for customers that were demanding them.
 
IBM for example produces since ages 2 chip ranges:
They are called "Z" and "P"
 
IBM own high end CISC CPU design is called Z.
IBM's real high end systems are CISC "Z" systems.

P which is IBM's name for PowerPC is their RISC "middle-range" offer.
 
 
INTEL continued to produce CISC chips all the time.
INTEL also tried to produce RISC.
They did this to save cost.
INTEL Risc chip was named "960".
The 960 chip was used as controller on many SCSI cards.
INTEL also tried to make even make simpler chips than RISC.
This range was called EPIC and the product name was ITANIUM.
 
In hindsight INTEL gave up on RISC and EPIC and focuses fully on their superior CISC design.
 


Vojin Vidanovic
(Needs Verification)
Posts 1916/ 1
17 Dec 2019 18:35


Thanks. While we loved 68k was beating XT all the way, at 286 was competitive to 020 and about 030 we were last time catching. 040 and 060 were too little, too late when compared to 486dx2,5x86 and Pentium and cyrix/amd competition.

I always ment RISC is a reduction, and now I have knowledge to back it up :-)


Kenji Irie

Posts 86
17 Dec 2019 19:24


A1200 coder wrote:

  A new Amiga design has to satisfy three key points:
 
  1) A 68k compatible CPU
  2) OCS/ECS/AGA compatible chipset (no foreign GPU add-ons can be accepted)
  3) AmigaOS 3 compatible OS
 

Interesting discussion from all. From my perspective,I view the Amiga PPC accelerator movement as a starting point from transitioning the Amiga to a new processor. Obviously this ultimately failed as it just got started, and now I see the PPC as a sub-branch of the Amiga history that is now laid to rest.

Now, I completely agree with the above quite from "A1200 coder" (have you got a name? :)  ).  However, I believe the new Amiga must go farther than this. It needs these, but should also have the new functionality of an updated GFX chipset, faster processor, modern tech interfaces etc, along with the software/firmware to get the system running out of the box. This is everything the V4SA does. This is why I find the V4SA so exciting. IMO, it's the exact evolution from the A1200, and is the only real way I can see that the Amiga can actually be invigorated for the future.

On a slight tangent: I believe this is the true niche of the Amiga in the modern world: a boutique piece of 100% compatible hardware (or near to it), that is pure fun to develop and use. My dreams come true if the V4SA becomes a stable platform that brings demo coders and developers back to develop new Amiga software, enabling the subsequent engineering of the next gen V5SA in 2 years time... of course, this all links back to the V4SA being commercially successful, which comes down to the V4SA focusing on the core abilities above everything else so that the majority of ex-Amigans come back to the platform.




Wawa T

Posts 695
18 Dec 2019 11:48


Kyle Blake wrote:

  If you want people to cooperate the first step is not to denounce justified views as "anti powerpc crap".*
 

 
  calling something crap, is the strongest argument this poster seems usually be able to present, so he does it quite frequently. also ranting the forums on something exclusively he considers right, backing it up with claims that everybody wants that. he is ppc fan. he has been regularly talking down m68k, genuine amiga as such, in comparison to amigaones, and also vampire, which he claims to own.  but right now paradoxically he is trying to convince everybody of his idee fixe, being mui wrapper on top of a mainstream system as a future of amiga. so far for sanity, consistency of views and technical expertise.


Gunnar von Boehn
(Apollo Team Member)
Posts 6207
18 Dec 2019 12:05


wawa t wrote:

  but right now paradoxically he is trying to convince everybody of his idee fixe, being mui wrapper on top of a mainstream system as a future of amiga. so far for sanity, consistency of views and technical expertise.

I not recall to have seen this MUI proposal of him.
I recall some discussion about how Doom works.


Wawa T

Posts 695
18 Dec 2019 12:07


Kyle Blake wrote:

If you want people to cooperate the first step is not to denounce justified views as "anti powerpc crap".*

calling something crap, is the strongest argument this poster seems usually be able to present, so he does it quite frequently. also ranting the forums on something exclusively he considers right, backing it up with claims that everybody wants that. he is ppc fan. he has been regularly talking down m68k, genuine amiga as such, in compson to amigaones, and also vampire, which he claims to own.  but right now paradoxically he is trying to convince everybody of his idee fixe, being mui wrapper on top of a mainstream system as a future of amiga. so far for sanity, consistency of views and technical expertise.


Marian Nowicki
(Needs Verification)
Posts 22/ 1
19 Dec 2019 05:57


One can read a lot of conspiracy theories here.
Why?
Why You people just not accept reality?
Motorola was not good enough to compete with Intel, so they have to get money from IBM, so they have to switch to IBM technology.
It is sad, but it is history. It was done more than 25 years ago.
Why this anti powerpc crap? After so many years?
It is time to stop this.




Nixus Minimax

Posts 416
19 Dec 2019 06:05


Marian Nowicki wrote:
 
  Why this anti powerpc crap? After so many years?

I believe that most of the "anti-PPC" Amigans of today did welcome the switch to PPC some time in the mid-90s even if they didn't follow the trend personally with the absence of "official" PPC-Amigas. The question thus should be "why still this pro-powerpc crap?" In hindsight the PPC part of Amiga history has proven to be an irrelevant dead end.



Gunnar von Boehn
(Apollo Team Member)
Posts 6207
19 Dec 2019 08:52


Marian Nowicki wrote:

Why this anti powerpc crap? After so many years?

 
Several of us have worked for IBM PowerPC development and know everything about PowerPC family.
 
PowerPC is by itself an interesting RISC architecture.
 
If you compare different architectures there are clear design goals
  visible.
 
ALPHA most important goal was high clock rate.
The ALPHA chips surpassed all others in clock rate.

For this goal the designers dropped even a lot useful instructions.
This made coding the ALPHA in ASM a nightmare.
For many operations there was not an ASM instruction in the ALPHA CPU.
For example: Division was missing and the developer needed to write  subroutines for this on ALPHA - while other CPUs could do this with a single instruction.
So coding ALPHA in ASM really was a pain.
 

MIPS most important design goal was simplicity of the decoder and the pipeline to make the chip as cheap as possible.
To reach this goal, MIPS did not implement any dependency checking inside the CPU.
This means the programmer needed to take care of dependency own its own. And if he did do this or overlooked a clockcycle then the MIPS CPU would not care and just calculate wrongly without giving a warning or error.
 
Lets make an oversimplified example to illustrate this:
Lets for discussion purpose say the ADD D0,D1 needs 2 cycle.
You can on 68K write:
  ADD D0,D2 
  ADD D1,D2 
  D2 now contains the sum of D2+D0+D1,  (time taken 4 cycle)
 
On MIPS this would not work as the CPU would not wait the 2 cycle by itself, as it would not track this.
To get the correct result the programmer would need to count the cycles and need to insert NOP instruction by himself to make it calculate correctly
  ADD D0,D2 
  NOP
  ADD D1,D2 
  D2 now contains the sum of D2+D0+D1,
 
The example is of course rubbish - but explains the concept
 
 
 
So ALPHA was a true nightmare to program, but could reach extrem clockrates. And MIPS was a rubbish CPU which was a nightmare to program but was dirt cheap.
 
 
IBM goal for POWER (which was originally called RS6000) were different.
IBM wanted a RISC CPU but with complete instructions set, having  MUL, DIV and other useful instructions. IBM also wanted the CPU to track all dependencies like a real "CISC" chip does.
As tracking in the CPU is really important to prevent program errors.

So POWERPC is from a architectural design much better to program than ALPHA and MIPS. And also was never as cheap as MIPS and never reached the insane clockrates of ALPHA.
The name POWER comes from this idea - IBM tried to keep CISC features in their RISC design to keep the CPU programmable and keep it strong.
So for a RISC - the PowerPC is indeed  very good.
There are far more ugly and weaker RISC designs than PowerPC.
 
If you of course compare the programmability of a true CISC like 68k with PowerPC than PPC is not as nice to program and the code is much harder to read and the risk to make errors is also much higher.

The AMIGA is a combined of 3 things:
A) Clever and elegant chipset
B) A CPU which is very good to code and very human readable
C) Slim and Fast OS - which is easy to use and understand

The combination of these features allowed coders on the AMIGA to write demos/games in ASM and to do very impressive stuff.

Now PPC is pretty weak as replacement here - simply as coding it in ASM is difficult and a pain.
So if you look at the AMIGA scene which were full of ASM coders -
that loved inventing new routines and writing cool demos in ASM.
Now taking their great to use CPU away,
and giving them something which is no fun to code in ASM - did not match the AMIGA spirit. And therefore really close to nothing was developed for POWERPC on AMIGA. If you simply count the demos, games  etc which the AMIGA scene did wrote then 99% was developed for 68K as the 68K did match the spirit of the community.




Michael Borrmann

Posts 140
19 Dec 2019 09:21


I don't get why you'd want a PPC CPU in the Vampire, when we can finally have 68k back at a reasonable speed in our Amigas by FPGA tech...
 
  There are faster alternatives for PPC if you want to use MorphOS or AmigaOS4.
 
  And as a fan of the 68k design, and as someone who had to code on a RISC design before, I am really happy about this development.


Vojin Vidanovic
(Needs Verification)
Posts 1916/ 1
19 Dec 2019 11:24


Nowicki wrote:
 
        Why this anti powerpc crap? After so many years
     

     
      I love PPC Macs and G4/G5 designs.
     
      However, on Amiga side PPC is a crap in implementation (driver support, OS advancements, compatibility etc.).
     
------------- Real world examples----------------
      DJ nick owned blizz 604+040 , expensive toy, buying 060 plus SCSI alone was better investment in real world performance with exception of few effects and PPC native wos proggies and games.
     
      I own A1-x1000. PA Semi Mytic CPU Is nice dual core SOCC, quite power efficient and made for laptops, not desktops. So in its category it performs well, and Linux is well usable if it did not drop support (which was never made official).

Its CFE and AOS4 and Linux support failing. No MOS for x1000 planned, x5000 territory, even its doable.
     
      I have prepaid AOS 4.2 only to be forced to change gfx card , buy os4.1 Fe and enhancer and in Real world test os4.1.8 plus enhancer 1.3 looks fine but Is incompatible to wos and m68k way less then mos and even to older os4 sw.
     
      I wish vamp existed and I have Never gone A-eon route.
      -----------------------
      So no, PPC is fine.
     
      Amiga ppc Is crap in realization, so I look forward to Aros x64, Mos x64 and 080 as m68k 64 bit.
     
      My pa Semi Is bi endian and 64 bit with altivec , 1gb lan and dual core, but os4 just uses altivec poorly
   
    Pa Semi Has fast ddr2 mem controller for ppc, but 080 blows it away.
  --------------------
   
    Similar 080 design would be way better and less pricey and sw kinky then my low quantity poor Nemo board - would run more AmigaOS sw, be more Amiga etc.

  Fast 080 could emulate 604 at half clock and Make some wos, Mos or older os4 sw avail. But Ask for that 2025
     
     


Kyle Blake
(Needs Verification)
Posts 108/ 1
19 Dec 2019 19:10


Marian Nowicki wrote:

  One can read a lot of conspiracy theories here.
  Why?
  Why You people just not accept reality?

 
  Where is the conspiracy theories? We've just talked about how the further development of product lines is decided by business reasons. This isn't conspiratorial, only common sense.
 
 
Marian Nowicki wrote:
Why this anti powerpc crap? After so many years? It is time to stop this.

 
  It makes more sense to ask why all this pro-powerpc crap? after so many years?
 
  25 years ago, Amiga PowerPC sounded like it made sense. But the PowerAmigaOS took so long to arrive that the PowerPC boards were ten years obsolete and too slow for it, and the PowerPC was leaving the desktop computer market forever.
 
  PowerPC isn't relevant to modern computers anymore, and never was much use to Amiga either. So why obsess over it?
 
  Even as a Co-Processor for offloading workloads, Amiga PowerPC was not very good, because it was wired as an alternative main CPU in the hope of PowerAmigaOS. So we had context switching disasters ruining the potential performance.
 
  I had a BlizzardPPC with Bvision, overclocked and I enjoyed it because it was so unusual. But it was also clunky when PPC and 68K had to work together. When it became so much money I sold it and bought an 030 card, and in daily use I don't see much difference. PPC did nothing for 95% of software.
 
  Now we have Apollo 68080 and it's at least as fast as ppc was in use, but the speed is for everything, not just some datatypes or mpeg decode. Wipeout 2097 is better on Playstation. So what have we actually lost?

posts 43page  1 2 3