ARM or x86 with FPGA emulator

[TeamBlackFox]

On the OP, this sounds like a dumb idea. As said, you're basically doing what Nintendo did with the SA-1 in the SNES game Super Mario RPG: Enhancing one part while leaving the rest untouched.

I really wish Commodore would have diversified the Amiga market. If they made a low cost server with a 68k and OCS combo chip around the time of 1992 then they'd have gotten a toe hold in the low cost server market, especially for people new to server administration. It may have not been super secure, but since Amiga in the 1990s was a poor man's SGI more or less, take a page from their book.

I would like to see the guys at A-EON do a small desktop version of the SGI Onyx2 cube design, and do something like this:

PowerPC or SPARC or MIPS quad core
DDR3 RAM up to 32GB
NUMA bus system

I'm a believer in keeping it simple and staying away from needlessly complex architectures. SPARC, for instance, is RISC, Big Endian and has open designs, plus Fujitsu manufactures tons of them for the server market. SPARC compares favorably to x86_64 if you compare similar die sizes and CPU classes. They're also a parallel orientated design which is where technology is going to move. Moore's law has been hit for clock speeds and the marketing behind that is slowly fading, but we can thank Intel and NetBurst for that, bunch of morons. So parallel, moderately clocked designs will become the norm. Even x86 is now becoming more RISC-like internally, converting the more orthogonal instructions to simpler ones with microcode.

68k was good for the time, where orthogonal instructions were useful, but today all this FPGA design effort is largely fruitless when the fact is the hobby of Amiga computers is going to die out. I like having Nia around, but usefulness is taking priority these days. Orthogonal instructions are largely useless waste of die space when MIPS for instance is simpler by a magnitude and is able to work with the fact that memory access is very fast now and we don't need to have very lengthy instructions fed to the CPU when it is faster now to do short concise instructions. Even Itanium and EDGE emphasize this with instruction parallelism, which while applying CISC concepts to RISC actually compliments the architecture to the point that one engineer I know described it as Super-RISC.

[TeamBlackFox]

Actually Fiz I do know what I'm talking about.

I think AmigaOS 4's best chance of survival is not emulation or on x86, but on an open design based on the NUMA architecture ( Basically it is a decentralised DMA system which is widely copied, Intel uses QPI which takes a lot of the same ideas ) and I fully support breaking both API and ABI as well as architectural compatibility for a sustainable architecture. Both MIPS and SPARC have open designs, which means A-EON could improve them without a license.

The primary advantages of AmigaOS are that it is lightweight without being impractical, and that has nothing to do with its backwards compatibility. Instead, make the OS fully 64-bit and use FS-UAE as an integrated sandbox for old software, add a ports system like in BSD as well as a frontend to it, which will make software dependency and distribution easy, NUMA architecture would increase bus performance and make a CPU that isnt as fast be less of a hindrance. CPU frequency is pretty much BS when a 1GHZ R16000A MIPS CPU wipes the floor with a 3GHz Pentium 4, all higher frequency does is make more heat and power consumption.

The only reason I don't support ARM is that up until recently most designs have been 32-bit, which in today's world of cheap memory doesn't make sense. In addition most ARM CPUs are for mobile and embedded devices. I have a Nexus 7 tablet, I love it, but it isn't any better than my 600MHz Octane2 at computational tasks, however in memory access and graphics, there is no contest, but that is to be expected.

[TeamBlackFox]

Are you aware of "OPEN POWER"?

Yes, but as far as I know there are differences between the parent POWER architecture and the PowerPC architecture, no? In addition, I'd think the cost to produce a MIPS or SPARC processor would be less.

If someone can build a NUMA/DDR3/PowerPC board and AmigaOS 4 is fully 64-bit on it, I'd be willing to pick it up at a price point of $1200 or so. I'd pay more if they donated boxes to the BSD projects though.

Currently the X1000 I would not pay over $600 for considering its specifications. Not to insult anyone but I can get a used HP C9000 PA-RISC for $300 that is as fast as a G5 but has SCSI on it ( I prefer SCSI for build quality )

[TeamBlackFox]

No offense taken! I realise my opinion may not be the norm at all.

PowerPC is not very common in the consumer market, same goes for SPARC, MIPS et cetera. However in the server market SPARC and POWER, the parent architecture of PowerPC are both well known. MIPS is currently embedded but has server grade and consumer grade designs available such as R10000 and derivatives.

The reasons I'm opposed to moving to x86 are simple: It is a pedantic, relic architecture which uses dirty hacks and workarounds to overcome the design limits of the original 8086, which by all accounts is a terrible processor. x86 is very orthogonal compared to all but a few mainframe architectures in common use today, but beyond that I've ran servers for 3-4 years and in that time my x86 boxes have died catastrophically, one caught fire and destroyed two XServe in the same cage. The build quality is just atrocious and delicate. I run computers hard and for a long time at high load ( I have computers I loan out for remote access and compiling ) and I've had no RISC system fail because of that. Therefore x86 ends up costing a consumer a lot more in time and money as one has to continuously replace broken hardware. Not the case for my RISC boxes.

I'm not opposed to a high performance, 64-bit ARM design at all, but those do not seem to exist. 32-bit is an engineering dead end, 4GB RAM just doesn't cut it. Need I remind anyone about the Y2038 bug that will render most old UNIX and similar systems inoperable? 64-bit is the future, even at the cost of compatibility.

Current offerings from ACube and AEON are overpriced and underpowered, so why not move to an open architecture and mature the operating system at the same time. MIPS is being widely produced in China and Japan which means low production costs, SPARC has open designs as currently stated as well. The PA6T chip is pretty much a dead end, newer POWER ISA chips are expensive as hell so I don't see a future here.

[TeamBlackFox]

Therein lies our differences in development style:

I mostly use Vi and Vim as a text editor, and use simple makefiles ( I want to find a simpler build system though ) with compilers like MIPSPro, PCC and Clang. I don't dare touch Windows development because I just can't take it. In addition IDEs are useless to me because they breed laziness. I also don't use C++ as I think it doesn't keep code concise and optimised.

I don't like the fact that even Intel or Microsoft compilers produce crap code, and again I like compilers to be smart and optimise code as much as possible. The day Amiga goes mainstream as it stands is the day it goes the way of BeOS. Haiku I've lost hope for as well because binary compatibility with the outdated ABI of BeOS has slowed them down.  You're free to take that as bull, but think about it: All successful in terms of numbers OSes are supported and dominated by a company who restricts the software's openness. Apple and Microsoft already do this, and Canonical and RedHat have already dominated GNU/Linux and used the GPL to lock software to GNU/Linux in an attempt to strangle the BSDs, which are supported but not dominated by software companies. And honestly RedHat keeps adding useless software to GNU/Linux to Windowsify it and make it "easier". They can't admit that BSD has the leg up on stability and so they're strangling it and the rest of the open market by adding dependency for PulseAudio, systemd, policy it etc.

The only way to break the cycle is to keep your projects in the niche where you aren't risking money in a gamble to beat the big dogs. It amazes me how arrogant most people on here are in this effect.

[TeamBlackFox]

I'm using the same method and tools other people are using so there is no point other than my method is console based and doesn't have a GUI. I'm not using an ancient compiler: the oldest compiler I've used is from 2006 - and it works fine with the C 99 standard. An IDE is just a graphical front end with proprietary or compiler specific libraries which hinder portability... Tell me again how that makes a programme any better?

Processor architecture does make a difference because RISC specific compilers like MIPSPro are designed to directly compile highly optimised code for use on their architecture. There is a reason -O3 is recommended in the documentation of Nekoware, an open source suite of software for SGI MIPS systems running IRIX.

Kremlar,

Its been a while so I don't remember the brand, but it was a dual Xeon server refurbished from the manufacturer. It failed three days after me and one of my friends who shares the cost of server cage space verified it for its 90 day evaluation. Everything was working fine on it. We had scarcely had it two years when the fire occurred. The fact is not that it was x86 that caused the fire, it is the fact we paid for a server grade hardware solution and got cheap rubbish instead. Nobody here I think has run their servers as hard as we did, and we do it because the servers are marketed to take this. Now we run increasingly more MIPS hardware converted for server use. No matter that some of it is 20 years old, our Sun and SGI racks have not had a single failure of a server in the same time that various servers running x86 and x64 hardware have given up the ghost. Luckily most just stop working, but there is a reason our hardware is now segregated by type: makes it easier to contain costly losses. The fact is I can get a few racks of SGI Origins for dirt cheap and when craylinked properly I get all the performance I need out of them and most of the time if there is an issue, it is an easy fix like a fan replacement or a cable reseat.

[TeamBlackFox]

@Linde

Maybe I'm not understanding your question, because I've been trying to answer it. Let me try again:

I prefer working with an architecture at a low level. The reason I give two %&$#?@!%&$#?@!%&$#?@!%&$#?@!s is because I actually would like to code some for Amiga down the line. Reason I don't now is simple: I am learning 68k assembler and that is one of the most tedious and pedantic languages I've learned because all the current compilers for C are too old and broken for OS 3.9 for me to use. I have been coding on a UNIX style C compiler such as Clang and that's how I intend to do this. GCC is a load of junk though so I may be forced to build my own compiler >_>. Anyways, back to my point: If the Amiga were to become x86 based I'd probably say screw it and not code for it because I have long given up optimising programs on x86 properly. Nothing about the tools I use exposes anything, rather, it is how I treat C as more flexible assembler. I like to be aware of the underlying hardware. With that said I'd probably stick to RISC boxes running BSD or some other UNIX if Amiga moved to x86.

@OlafS3

Ive got much different goals in mind than you: I never intend to make money off the code I write. I therefore could care less about what a user thinks. As a sysadmin as my day job I don't handle users or interact with them so I generally speaking don't care.

I'm just upset because I'd like to do some NG work since the toolchain and OS have a lot of expansion, but acquiring new and high performance hardware is impossible. I suggested a switch because it seems AEON and ACube are utterly incapable of building units at a reasonable cost/performance ratio. I could buy a whole rack of Onyx 350 IR4 graphics for the price of the X1000 and wipe the walls with it, especially since OS4 is a 32-bit, multiprocessor blind system.

AROS doesn't count for me since I dont own any powerful ARM or PowerPC hardware

@biggun

I REALLY don't appreciate being called stupid subversively. Currently there are plenty of more performant MIPS and SPARC chips available than most PowerPC gear. The R16000 of MIPS fame is faster despite being clocked lower, and a Sun Ultra 45 wipes the walls with a G5 and a PA6T. We're talking 10+ year old designs: It doesn't cut it! And if they adopt faster CPUs for the next NG Amigas, PowerPC ones I can expect the systems to be even more expensive.

As previously explained I could give a flying freak whether or not it would break binary compatibility. Sure other users care, but I care more about having a 64-bit OS with a CPU that can perform well today.

Don't feed anymore bull, you can look up the dhrystone measurements and other benchmarks yourself, but do keep in mind all of these systems are 64-bit and SMP capable so they're going to beat any slow uniprocessor 32-bit mode OS

[TeamBlackFox]

Well of course. My sentiments exactly. If I'm gonna buy something high performance it has better be a good RISC CPU is my though, and for anything recent that leaves out PowerPC, ARM is too underpowered, so that leaves SPARC and MIPS.

Of course if A EON were to get a POWER7 based system with NUMA and reasonably priced around $1000-1500 at most, sure I'd pick it up. But the current offerings seem to be limited by cost and CPU availability. Why not then move to MIPS which has plenty of high performance designs available? Or SPARC which has processors sold from Fujitsu. I'm sure between those a lower cost yet high performance option exists!

[TeamBlackFox]

You know what? It isn't even worth arguing over anymore. Screw it. I'm not being unreasonable and I'm certainly making a point that everyone seems to casually deny: that SPARC and MIPS are readily available, and offer more performance for less cost than our current designs. Sure it would cost money to move, but I can't imagine A EON is getting rich on the 100 or so X1000s sold. I also think I'm not alone in feeling alienated and cheated in that the performance offered is much less than I can get elsewhere.

The more I think about it though the more I see why nothing gets done here. Just too fragmented and resistant to change anything that may break compatibility.

[TeamBlackFox]

Gladly for MIPS: lemote.com/en/products

For higher performance stuff, a company like A-EON could take one of the open designs such as the OpenSPARC T1 and build boards off that for reasonable prices. It isn't as hard as you're postulating.

[TeamBlackFox]

If you're going to move anywhere you move to an architecture where there is an existing inexpensive desktop product that exists - like x86.

Yeah and that will be the day I move to another NG product that doesn't do that. You don't realize the scope of issues that is going to cause. Look at BeOS/Be Inc. if you want to know what lies in store for anyone wanting to play with the big boys.

[TeamBlackFox]

Kremlar,

Come now, the majority of cost of production isn't the creation/assembly so much as licensing. If someone produces an OpenSPARC derived chip and then orders at least 1,000 of them from China, then its gonna be hella cheaper than the current 300 or PA6T orders and then having to deal with licensing costs et cetera.

The Lemote computers are readily available from resellers, they sell them on Amazon and other online retailers. The entire architecture from CPU to firmware is open source, which is good because I've had to deal with the proprietary firmware on servers and coercing them to run BSD instead of RHElL or Windows can be a challenge. Having an open source from top to bottom architecture will help minimize bugs and reduce power consumption, improve ACPI stability etc.

Yes, I'm anti-x86. I'm against a backwards architecture that is poorly engineered, constantly fought over by AMD and Intel, hampered by incompatible addressing modes, uses hacks to extend the opcode possibilities.

A CPU in this day and age doesn't need 1000+ op codes to do its job. And since ABI/API has long since broken legacy software there is little point in conserving the real and protected modes of operations.

You also completely blew off my warning about BeOS: what makes Amiga any different?

Let's see :

Both are niche OS with various open and proprietary solutions
Both suffer from infighting and fragmented user base
Both are hampered by backwards compatibility with the original iteration
Both had official solutions running on PowerPC
Both have open source solutions on x86, neither are very successful outside their niche

BeOS failed permanently when it switched from PowerPC, and dropped it for x86

Sounds an awful lot like history repeats itself.

[TeamBlackFox]

OlafS3,

Tekmote.NL sells some for €200-400, mostly due to import tariffs. If there was more demand I'm sure the cost would drop.

[TeamBlackFox]

Sure biggun, but it takes longer to have those instructions processed. The small amount of registers, combined with the amount of wasted silicon to useless backwards compatibility, and the overly complicated instruction set is just terrible. It makes a definitely orthogonal architecture like m68k seem RISC by comparison.

It takes about 40 cycles for the average x86/x64 CPU to access memory due to the size of the pipeline, a MIPS R16000 can do it in 11. In addition a benchmark compared a contemporary AMD Phenom to a R16000A, and it found they take about the same amount of time using a single thread to do a task. The AMD edges out as the number of threads increase till you hit the limit of CPUs a single computer can hold, then the R16000A takes the lead again with its ultra fast craylink clustering ability. Craylinked x86 computers are very uncommon whereas the R16000 is deployed with it almost exclusively. The reason why the R16000 can hold out so well has to do with its massive L2 cache in the 1GHz version it is 8MB, and it has a very streamlined instruction set.

Mikhail Kalashnikov, the late designer of the AK47 once said " All that is too complex is unnecessary and it is simple that is needed. " there is much to be said about this in the computing field.

[TeamBlackFox]

I've never heard of them.  There's ONE vendor on Amazon named "Revolutionary Books" with a 33% rating (3 reviews in the past 12 months) selling a Lemote netbook for $1500, and also for $2500?  This is the architecture you think they should port to??  I seriously doubt you'd ever get one if you ordered from that Amazon vendor.

BeOS failed to right it's ship even after porting to x86 because their sales were atrocious and Apple did not save them.  AmigaOS sales are atrocious now, so what's the difference?  Hyperion somehow seems to survive by selling a handful of licenses (if that) per month, which Be could not do.  

I'm not saying AmigaOS will survive long term by porting to x86, but it certainly won't survive on the path it's on now, and certainly would not survive if ported to ANOTHER obscure platform like you are suggesting.

I'm not a fan of the NG systems in general - I don't see the point, unless something unique can be brought to the table.  But, if you're going to do SOMETHING, don't spend incredible resources to just move sideways like you are suggesting.

Amiga was different because it was better.  There's nothing better about current NG systems, hardware or software - they are worse.  Porting to an obscure platform does not change that.  Porting to x86 could at least put them in the ballpark hardware wise, and resources could then be devoted to software instead.

Look what the move to x86 did for the Mac?  Saved the platform.  Of course the Amiga market doesn't have Apple-like resources, but with WinUAE a sandbox already exists to run classic apps on x86 - and that's half the battle.

AROS isn't successful because:
 - It's not "blessed" with the Amiga name
 - It tries to support generic x86 hardware and does not have "official" hardware behind it.
 - There are not enough resources devoted to it.

Just an FYI, Amiga and BeOS objectively are two approaches to the same question " How simple can an OS be and still be useful? " neither is objectively better than the other. If you yourself aren't interested in NG Amiga hardware then you're doing nothing more than trolling this topic. Don't put out ideas that will potentially ruin what's left of the community if you have no intention of taking responsibility for your ideas, which your lack of interest clearly indicates.

Furthermore, you don't understand that putting a proprietary OS that lacks most features new users are familiar with on the same ballgame as Windows, OS X and GNU/Linux will result in them being trampled.

The reason the BSD communities are still around are that they mutually assist each other, are ported to several architectures which diversify the risk, and they aren't money driven.

Going after an idea with a gamble such as the entire OS and company at stake is an unacceptable risk, even if the supposed pay off is big. You wouldn't gamble your house in Las Vegas or Cancun Mexico if you had a wife and kids living there was the no second dwelling. As such, the Amiga community is spread too thin for diversifying to work. Instead there needs to be an affordable, high performance niche piece of kit which both NG solutions can agree to support.

There is just too much risk.