One unified OS for the future?

[kamelito]

IMHO after the X1000 they should have made the X500 for 1/4th of the price.
It should have been at least 2Ghz dual core. I'm not seeing myself putting 3000Euro on any computer even the latest IMAC of MacPro even less on an Amiga but I totally understand those who do.
As for an unified OS forget about it. The nearest AmigaNG experience I'd like to try is WinUAE/FSUAE with AmigaOS 4.1 final.

Kamelito

[takemehomegrandma]

A-EON should partner with TYAN to adapt some of their CPU designs to a newer, less expensive AmigaNG system,since TYAN already manufactures POWER hardware.

OMG!!! :lol:

Did you even think that through before hitting the "submit reply" button?

[TeamBlackFox]

OMG!!! :lol:

Did you even think that through before hitting the "submit reply" button?

*Ahem* I don't appreciate being insulted and addressed in a patronising manner. I did logically think this through: For roughly the same price as a brand new X1000, I could buy a POWER8 server ( The PA6T is based on POWER5, for reference, and that is 5 revisions behind the POWER8, counting the + revisions ) I am seriously getting something worth every penny I pay, compared to the X1000, which is using 10-year old tech and passing it off as a new system.

With A-Eon's investment of more than 1 million dollars to Vari-Sys to  design and build PPC motherboards for running AmigaOS4.x (and possibly  some embedded markets), I don't see them changing directions any time  soon, or switching to a different company to do design work, or  manufacture boards for future A-Eon computers.

All very good points, but if Varisys doesn't offer hardware that is cost-effective to build, and performs poorly, what's the point? Besides you and a few others, I don't know anyone who even owns an X1000, let alone laid eyes on it. I'm sure they were the best option when A-EON was founded by far, but you have to jump ship if the current one is leaking. I can't imagine Trevor and A-EON have made much in terms of profits, and while I know none of those involved in OS4 do it for the money, I can't have expected them to survive long running a deficit, especially with the banks so reluctant to lend money.

I yesterday bought a notebook with 2 core 2.7 GHz, 4 GB RAM, 500 GB  Harddisk, GPU, WIFI and so on  incl. Win 8 for 238 EUR (used). There is  nothing that can compete right now with such offerings.

Sorry to burst your delusion, but there are: http://elinux.org/Jetson_TK1

I have experience with that board above, and it is a wonderful board that I have been given the pleasure of using, for free, for work purposes. It is quite powerful, not to mention inexpensive, and performs probably twice as good as your laptop's integrated graphics pipe, and depending on the CPU generation, comparable or better than the CPU.

Not exactly the cheapest ARM board out there, but definitely a winner in terms of cost/performance. I also bought a BeagleBone Black for $55+$11 shipping, and that is no slouch either, I have it running Gentoo until my friend, the developer of Void Linux and a former NetBSD developer, ports Void to it. It is very fast, about 4-6 times the speed of a Raspberry Pi, which is about the speed of a 300MHz Pentium II. Do the math, and you can tell it blows the transistors off of a Pentium III running at 1GHz, which I have running PFSense here, especially in power consumption.

I'd love to benchmark the TK1 against an old Mac Mini G4, I suspect that since it has twice the RAM, many times the bus speed, a faster CPU and GPU, it will beat it at nearly every record. When Project Denver, the ARM64 Tegra comes out, it will do better than almost every Intel Atom and likely a G5 CPU in terms of performance.

x86_64 is so messed up, that it has literally 63 different instructions that can perform the jump operation. An audit against the instructions used by the Linux kernel to the instructions present in a modern x86 CPU, in hardware or microcode, came to be roughly 1 instruction used by Linux for every 10 the CPU offered. And it has to maintain that massive instruction set, either in hardware or microcode, in addition to anything else AMD or Intel wants to cram into the chip. The patent-protections on these chips prevents third-party compilers, such as GCC and Clang, from fully utilising the instruction sets, forcing you to either use a proprietary compiler and the OS it will run on to build your OS and programs, or else contend with only being able to utilise a fraction of a chip's instructions.

The problem, I will remind you, with using the x86 architecture as a base, are the following:

The hardware is so diverse that with the tiny size of the community, only a small subset of that hardware will ever be supported. The BSD community, which is at least 4 times as big as the Amiga community, has trouble with this on almost every system built today. Your average GNU/Linux distro that doesn't use blob drivers not merged into the Linux kernel only supports about half the computers out there to satisfactory level, adding blobs from Nvidia or AMD raises it to about 70-80%. The GNU/Linux-libre distros? Less than 10%.

The Amiga community won't be able to utilise blobs, because Nvidia, AMD, Realtek, Intel etc. won't write them, unless you're referring to the Linux-hosted version of AROS, so you can count on their support being less than 10%, even in a unified state.

Furthermore, most members here don't even use AmigaOS or an OS that in any way resembles Amiga OS, the majority uses Windows, with lesser minorities using OS X and GNU/Linux, and then you have the esoteric ones like me who use BSD, IRIX and Solaris, operating systems most people aren't even aware exist. So what's the point? Most are on here to discuss WinUAE or maintaining old hardware, so one could argue that the apathy and lack of participation among userbases would effectively stifle any effort to modernise the OS or the hardware underlying it, let alone UNIFY it. You're kidding yourself if the switch to x86 will do anything other than piss in the well of the users who sunk money into NG hardware, frustrate new users with lack of hardware support, and fragment the community even more, and perhaps even cause MorphOS or OS4 to be forked again, making yet more players into a market the size of a quark.

[guest11527]

x86_64 is so messed up, that it has literally 63 different instructions that can perform the jump operation. An audit against the instructions used by the Linux kernel to the instructions present in a modern x86 CPU, in hardware or microcode, came to be roughly 1 instruction used by Linux for every 10 the CPU offered. And it has to maintain that massive instruction set, either in hardware or microcode, in addition to anything else AMD or Intel wants to cram into the chip.  

Why do you actually care so much about that? It's utterly irrelevant when it comes to performance or programming. You have the compiler to keep care about the code generation, problem solved. It might have been an issue 1987 when people still had to program partially in assembly, for some parts at least, but that's almost 20 years ago. We don't work this way anymore. Actually a lot of code is in high-level languages these days, with many levels of software and translation between the actual CPU, and the only thing that really counts is compatibility to legacy code and performance.  

The patent-protections on these chips prevents third-party compilers, such as GCC and Clang, from fully utilising the instruction sets, forcing you to either use a proprietary compiler and the OS it will run on to build your OS and programs, or else contend with only being able to utilise a fraction of a chip's instructions.

Pardon me? Since when does a patent disallow a compiler to emit a specific instruction? Please backup this claim - just the opposite, I'm pretty sure the intel folks do the best to provide customers with full support for their hardware. At least, this was my experience - the intel open source folks were always very helpful and open when I approached them.    

The hardware is so diverse that with the tiny size of the community, only a small subset of that hardware will ever be supported. The BSD community, which is at least 4 times as big as the Amiga community, has trouble with this on almost every system built today. Your average GNU/Linux distro that doesn't use blob drivers not merged into the Linux kernel only supports about half the computers out there to satisfactory level, adding blobs from Nvidia or AMD raises it to about 70-80%. The GNU/Linux-libre distros? Less than 10%.

Probably you then shouldn't even attempt that, and depend on hardware by vendors that release specs. See above, I had good experience with intel these days. Sure, you don't create a gaming machine with intel chipset graphics, but it's way ahead compared to what Amiga could do ages ago, and what "hand-rolled" hardware can provide.  

You're kidding yourself if the switch to x86 will do anything other than piss in the well of the users who sunk money into NG hardware, frustrate new users with lack of hardware support, and fragment the community even more, and perhaps even cause MorphOS or OS4 to be forked again, making yet more players into a market the size of a quark.

That's an argument that always comes up when you want to throw good money after bad money. Exotic hardware was a bad investment in first place. At some point, you'll have say that it isn't worth spending anymore and stop the nonsense. The error doesn't get better by making additional investments - the error gets worse. If you do, you're only trying to lie to yourself by finding justifications for having made a mistake in first place, and - if you ask me - all the Amiga NG stuff on proprietary hardware *was* a big mistake.   Two options exist: Either make this a complete software thing and depend on off-the-shelf hardware, or make it a complete retro thing and try to build on the 68K/custom chip hardware from old times.

[Blinx123]

@TeamBlackFox
Always interesting to read the posts of someone like minded.

Personally, I'd love to see Amiga going down the ARM road. It just makes sense, if you think about it.
ARM is inexpensive, quite powerful and can easily be adapted to something like an A1200 case.

I still have a Beaglebone Black lying around that I spent 50 euros on but haven't used since.
If I have time to spare over the holidays, I might give it a go, install a patched Linux (or perhaps something BSDish) on it and try and run E-UAE through the framebuffer.

[guest11527]

Personally, I'd love to see Amiga going down the ARM road. It just makes sense, if you think about it.

ARM is a nice processor family, and  not outdated as PPC or 68K, but that doesn't change the problem. As soon as you're depending on non-standard hardware, prices will be high and products will be non-competative because you're operating in a very small market. Unless you see AmigaOs working on a smart phone or a tablet, but it's not exactly designed for that, neither the power management (which power management?) nor the user interface. So yes, it's a nice solution from an engineering perspective, but it won't fly either.

[Blinx123]

ARM is a nice processor family, and  not outdated as PPC or 68K, but that doesn't change the problem. As soon as you're depending on non-standard hardware, prices will be high and products will be non-competative because you're operating in a very small market. Unless you see AmigaOs working on a smart phone or a tablet, but it's not exactly designed for that, neither the power management (which power management?) nor the user interface. So yes, it's a nice solution from an engineering perspective, but it won't fly either.

As far as I know, companies like TI actually build hardware to spec (i. e. they do contract work).
Even if they didn't, nothing wrong with choosing an existing board and providing a custom firmware.

The Amiga family ought to concentrate on software more than hardware. It's not in the interest of such a small community to concentrate on expensive, custom hardware.

[Kronos]

As soon as you're depending on non-standard hardware, prices will be high and products will be non-competative because you're operating in a very small market.

There is a rather long list of small&cheaps ARM based boards readily available, stating from the rather weak RPi up to some quadcore versions, so HW surely ain't a problem when going ARM.

[TeamBlackFox]

Exactly! When you look at it from my perspective, there is no reason to *not* seriously consider ARM. I've worked for MS, Dell, and a few other companies, and all the higher ups say that there all pushing their chips in the ARM bucket due to Windows' decreasing user base giving a very good opportunity to release more high end ARM CPUs. Qualcomm, Nvidia, etc are all working furiously on getting more powerful boards and CPUs ready.

[guest11527]

Exactly! When you look at it from my perspective, there is no reason to *not* seriously consider ARM.

The question is just that: What exactly do you plan to do with the machine, and which problem is it supposed to solve? What can ARM do what PPC cannot? In a sense, you are advocating to repeat the same error as before, namely trying to establish a hardware platform. This type of approach would have worked in 1987, and it did work for some time for CBM, but it is now 2014 and CBM went downhill for a reason. You sell solutions nowadays, a complete product families or full infrastructures - see the all the App-shops in the M$, Apple and smartphone market. Thus, you are advocating to  replace one niche market by another niche market - for which reason? Why should ARM be any better than PPC?

Like I already said, either Amiga is going through a transformation offering solutions for real world problems working on real-world hardware people can buy cheaply - or already have at home - or they should address the retro market and then need to consider compatibility as ultimate goal (unlike what they do today). The strategy of selling custom hardware is no longer working today because you can no longer create a competative product - existing off-the-shelf hardware is miles ahead from what a small vendor can possibly do. That was different when the Amiga was created, and it sold quite well because it did offer a very competative hardware indeed, and sold software by that.

[Blinx123]

What can ARM do what PPC cannot?

Not costing you an arm and a leg, for one.
On the other hand, it's also evolving far quicker than any other hardware platform.

If Amiga is to survive, what the community needs is inexpensive, extensible hardware with a low footprint. ARM can do that better than any other platform (including x86).

Inexpensive hardware is what made Amiga great in the first place, btw. Compared to similar offerings from Apple, an Amiga 500 wouldn't set you back all that much.

[guest11527]

Inexpensive hardware is what made Amiga great in the first place, btw. Compared to similar offerings from Apple, an Amiga 500 wouldn't set you back all that much.

The problem is that you are still applying a business logic from 20 years ago. Back then, you could create a great hardware and sell that to freaks, no matter whether it comes with software or not. It's enough if it comes with Basic.

But look, this doesn't work anymore. Any ARM hardware you would get would be behind the x86 "off the shelve" cheap boards you can get. Hardware isn't relevant anymore, you can get excellent hardware for little costs almost everywhere. Whatever you pick, it wouldn't stand out, and you cannot sell by the hardware.

ARM has its justification, it's an excellent processor for embedded applications where you need maximum performance for as little power as possible. But is that a requirement? Is Amiga for "mobile" and "computing, everywhere"? Not that I remember.

Or to put this in a different way: Before you select hardware "by the coolness factor", you should probably better know what your requirements are. If "runs fast and can do a lot of things and doesn't have to cost much" is the requirement, then x86 is the answer these days. You get a lot of boom for little bucks.

[Blinx123]

@Thomas Richter

I'm not applying 20 year old business logic.
I'm well aware of the fact that software sells hardware.

However, in order for there to be great software we need to cater to developers first.
Not many software devs are able or willing to create software for hardware that costs several thousand dollars and is pretty outdated.

You'll have much more luck finding devs willing to code on a sub-100 dollar platform that is well documented and that plenty of people (probably even more than x86, those days) know how to program for.

Again. I'll probably start a little project over the holidays. If anyone wants to join, PM me.

[TeamBlackFox]

Thomas_Richter,

I don't know your background in the hardware industry, but I have several people in the companies I have worked for saying that the engineering divisions are putting their bets on ARM. There are ARM for mobile, ARM for the workstation and ARM for the servers. It isn't limited to mobile or embedded applications, and hasn't been for about 5 years. If engineering divisions of major companies like MS and Dell are saying to place your bets with ARM, then perhaps a major shift is coming in the next few years. With mobile devices being the norm, it may end up spreading back into the workstation and server markets, and the server market is already getting some promising designs. AMD is selling society compatible Opteron-A ARM kits for developers, Qualcomm and Nvidia are ramping up production with high performance ARM chips. Why not get ahead of the curve, and stay on top of the game?

x86 is very costs ineffective because it doesn't scale down well, Intel Atoms consume on average 20-30% more power while offering a diminishing return on power as clock speed scales up. I have a Nocona Xeon workstation at my side that besides being louder and running hotter than modern chips, does roughly as well due to having massive RAM ( My machine has 16GB ) and UW-SCSI, which is still very fast when using 15kRPM disks. The only applications it sucks are are those that are quad+ core aware, it has dual Xeons and for NetBSD usage does very well, but I simply don't use it because it is loud and hot when I have the option of using my TK1 based board from work, which is quiet and has better graphics performance.

[Linde]

In a sense, you are advocating to repeat the same error as before, namely trying to establish a hardware platform.

There are tons of off-the-shelf ARM-based boards available, some cheaper than dining out, so there is no need to "establish a hardware platform". I tend to agree that ARM isn't the right way, but that argument just seems to have no basis in fact.

[OlafS3]

Thomas_Richter,

I don't know your background in the hardware industry, but I have several people in the companies I have worked for saying that the engineering divisions are putting their bets on ARM. There are ARM for mobile, ARM for the workstation and ARM for the servers. It isn't limited to mobile or embedded applications, and hasn't been for about 5 years. If engineering divisions of major companies like MS and Dell are saying to place your bets with ARM, then perhaps a major shift is coming in the next few years. With mobile devices being the norm, it may end up spreading back into the workstation and server markets, and the server market is already getting some promising designs. AMD is selling society compatible Opteron-A ARM kits for developers, Qualcomm and Nvidia are ramping up production with high performance ARM chips. Why not get ahead of the curve, and stay on top of the game?

x86 is very costs ineffective because it doesn't scale down well, Intel Atoms consume on average 20-30% more power while offering a diminishing return on power as clock speed scales up. I have a Nocona Xeon workstation at my side that besides being louder and running hotter than modern chips, does roughly as well due to having massive RAM ( My machine has 16GB ) and UW-SCSI, which is still very fast when using 15kRPM disks. The only applications it sucks are are those that are quad+ core aware, it has dual Xeons and for NetBSD usage does very well, but I simply don't use it because it is loud and hot when I have the option of using my TK1 based board from work, which is quiet and has better graphics performance.

"bet on something" is for private people making bets on horse races, on business you "do not bet" but make predictions and try to leave open chances to turn direction if prediction fails. We all know of the bets that were lost, from Commodore starting up to what we are discussing now. So if changing direction it would make more sense to have something portable like Aros so you can support both X86 and ARM and are on the safe side.

And if you talk about companies, there are a lot of companies that set on the wrong horse and do not exist anymore. Not a very wise strategy.