Ease off bashing AmigaONE X1000?

[OlafS3]

it isnt necessary nor possible to rewrite whole aros in asm to gain speed. aros needs to be platform independent and portable and therefore source needs to be kept in a high language for the most part. but critical parts are in assembly already, or may be yet replaced by such where it is necessary.

also there are initiatives to rewrite and optimize essential libraries in asm, consult eab for that.

however toni wilen confirms that this work on details is mostly of minimal influence for the whole especially in comparison to effort. i can confirm it from own experience when testing and helping alain thellier to improve storm mesa (68k). i dont remember which compiler exactly he has used, i guess it was gcc3.4.0 (which is particularly bad, current 4.6.x versions, used for aros, are better again). alain let out the genuine asm inlines, because they were not compatible with gcc, but it had no negative influence on overall speed.

I think he not meant to rewrite whole Aros in Assembler (only critical parts)

[vox]

thanks for nice words (even for the "messy" part :-) ). Your part testing AROS 68k on real hardware is very important too. Thanx for that.

the lack of interest is because most 68k fans do not realize how important AROS 68k will be in future. It is a standardized, extensible and opensource 68k platform. It will run not only on classic hardware and emulation (including AROS distributions) but also on new FPGA based hardware. So you can write a 68k software and it runs on all supported platforms. You miss functionality, it can be added. Most 68k fans only see it is it 100% compatible, answer No. Is it running as fast as AmigaOS, answer No. Then they say no interest. That is shortsighted from my point of view but it is how many people think. That will not change before software will exist for AROS 68k that is not available for any other 68k platform.

I do understand it is AROS for all as it can be run via UAE. Could you do some version that can be fast and hosted inside MorphOS/AmigaOS 4 68k emulation? I am not keen of running emulated AROS inside QEMU on Linux on X1000 ... So I see point here if possible, too ... AROS for MorphOS and OS4 users.

Sadly, Natami is unlikelly to hapen, but its good to see FPGA Arcade and others for Classic lovers. If there was no AROS 68k, there would be no point in further 68k development.

And hope we are done with X1000 bashing :-)

[save2600]

Huh? Where's the A2000 accelerator? And I mean accelerator, not some underpowered completely useless 020 or 030? My last achievement was a used Blizzard 2060 which cost me 2/3 of a new Sam440ep. So much about "affordness".

Look around. Amibay, eBay, etc. More 2060's, 1260's, and PPC cards for the A3000/A4000 the last couple of years being sold than ever it seems. 2060/1260 still in the $350-$500 range. But yes, the PPC boards are going for crazy silly prices.

And hell, the "measly" 020/030 will probably forever be in more peoples hands than X1000 + Sam systems combined. Don't understand what's "completely useless" about either chip??

[OlafS3]

I do understand it is AROS for all as it can be run via UAE. Could you do some version that can be fast and hosted inside MorphOS/AmigaOS 4 68k emulation? I am not keen of running emulated AROS inside QEMU on Linux on X1000 ... So I see point here if possible, too ... AROS for MorphOS and OS4 users.

Sadly, Natami is unlikelly to hapen, but its good to see FPGA Arcade and others for Classic lovers. If there was no AROS 68k, there would be no point in further 68k development.

And hope we are done with X1000 bashing :-)

X1000 is there and all was bashed what could be bashed :-)

You can only run AROS 68k in UAE. QEMU is very slow from my experience. I think someone run AROS in QEMU on a X1000 (I think I saw a screenshot). But when I tested AROS X86 running with QEMU it was too slow. I installed it in VMWare. "Hosted" is also very fast.

[wawrzon]

Could you do some version that can be fast and hosted inside MorphOS/AmigaOS 4 68k emulation?

i fear its not up to olaf to make aros68k any faster. however the quest is ongiong. how fast is uae on ppc systems available for you?

[vox]

X1000 is there and all was bashed what could be bashed :-)

You can only run AROS 68k in UAE. QEMU is very slow from my experience. I think someone run AROS in QEMU on a X1000 (I think I saw a screenshot). But when I tested AROS X86 running with QEMU it was too slow. I installed it in VMWare. "Hosted" is also very fast.

Yes, and since AOS4 (and MorphOS) have 68k CPU JIT, is it possible to make some faster box then having to use UAE (or QUEMU/UAE on Linux) to run it? I know you don`t maintain the PPC release - but it does it run hosted on PPC Linux too? (natively only on SAMs?)

[Acill]

I still stand firm on my feelings that this is a ridiculously priced machine for what you get out of it. You cant compare how MOS runs on the G5 to it at all, sure it only uses one core, but it does that on the G4 too and not because they cant get it working with both, but because they chose to maintain compatibility.

As for the 1GB of RAM, thats still a hell of a lot better over what your getting from the X1000 hardware being used on AOS. All the hardware that came with a G5 is used and working on MOS, if it shipped with the models that are supported, the hardware has a driver. This was all done with a bounty for less than a single X1000 machines cost.

The point goes back to the original gripe I had and getting back on topic, selling a machine for a premium price that doesnt even have full support of the hardware stinks of "give me your money" and isnt in the spirit of "Amiga" at all. This is what killed the damn Amiga in the first place!

They should be building a machine that is better priced, fully supported and easily purchased by the average user to sell more copies of OS4 thus increasing the number of users and making some money at the same time.

[vox]

I still stand firm on my feelings that this is a ridiculously priced machine for what you get out of it. You cant compare how MOS runs on the G5 to it at all, sure it only uses one core, but it does that on the G4 too and not because they cant get it working with both, but because they chose to maintain compatibility.

As for the 1GB of RAM, thats still a hell of a lot better over what your getting from the X1000 hardware being used on AOS. All the hardware that came with a G5 is used and working on MOS, if it shipped with the models that are supported, the hardware has a driver. This was all done with a bounty for less than a single X1000 machines cost.

The point goes back to the original gripe I had and getting back on topic, selling a machine for a premium price that doesnt even have full support of the hardware stinks of "give me your money" and isnt in the spirit of "Amiga" at all. This is what killed the damn Amiga in the first place!.

Well Trevor could hold back and wait for OS 4.2 (as machine was announced) and we wouldn`t have all the shame and blame, but instead I do believe quite a lot was gained even by premature, expensive and blah blah system: path to develop more advanced features, experience to correct mistakes such as PA Semi, and possibility for users to have high end system.

All the hardware in X1000 is usable with limitations of 2D driver, no second core and 2GB RAM limit - which stands for all AOS 4.1.6 users.

As of G5 and MorphOS bluetooth and firewire I believe are not (yet) support.

I on my own have calculated what I get with SAM 460 and X1000 and decided to go for X1000 - just because it will offer way more once oS 4.2 is out. Not that I am happy with pricing - but I am aware that it is future of AmigaOS 4 and that I will see better days with dual core, 4GB RAM and nice RadeonHD 7000. Even there will be a better X2000 system about time of OS 4.2, I would take it now, and show appreciate for Trevors effort to provide a path for AmigaOS to grow mature - where I do count RadeonHD 3D driver and OpenOffice ports. So I count my money went to both OS 4.1, OS 4.2, and Trevors efforts.

I was nearly buying Mac G5 for about fraction of what I will give in the end for X1000 - but sadly was beaten in bidding, Even higher price I will pay then people in EU (when I count in import taxes, postage and packing and enormous interest I will in the end pay to the bank loan) but I do believe its path towards seeing AmigaOS coming to quite usable state. Meanwhile I will have to use Linux for daily work.

If there was no Trevors effort, I would never dare to do that. And I believe everyone that buys X1000 (or later on X2000) knows what he is buying it.

They should be building a machine that is better priced, fully supported  and easily purchased by the average user to sell more copies of OS4  thus increasing the number of users and making some money at the same  time.       

I agree that is "not Amiga spirit of old" (even Commodore did nice profits on high end models and they were never acessable to all)
but SAM 460 looks like new low end. If ACube brings new board, SAM 460 Lite should become new low end.
There I see it fitting with AmigaOne 500 model.

If it was such an easy task to get new cheap but good PPC board, why there is none for MorphOS (Efika being even lower
end then SAM 440)

Once MorphOS gets either ported to X1000 or x86 I have no doubts I will buy it. As some SAM 460 users will.

With support of boards, I do hope OS 4.2 will bring order and support, and that X2000 and beyond will come
fully supported breaking the ugly tradition of selling hardware before fully supported with software, as it
started with original AmigaOne coming with no AmigaOS at all.

So, my hope is that I have partialy invested in brighter future, as if anything gave a hope to OS4
users, it truly is A-EON as Genesi once did for MorphOS.

[Crumb]

@vox

MorphOS drives firewire stuff correctly with Helios on my Mac Mini, I don't see why it wouldn't work on G5s. As for Bluetooth I don't think I ever used it on Windows/OSX machines.

[vox]

@vox

MorphOS drives firewire stuff correctly with Helios on my Mac Mini, I don't see why it wouldn't work on G5s. As for Bluetooth I don't think I ever used it on Windows/OSX machines.

I stand corrected. Bluetooth is great thing when sharing stuff with phone or even making a wireless network, and one thing I praise Apple for is including most modern standards and quality input/outpout devices.

My point was not to bash MorphOS back - MOS usualy really supported its hardware, unlike AmigaOS 4 - but to show that breaking OS barriers with high end hardware is not an easy task.

Problem is I don`t see what will be current MorphOS plan (beside announced ISA change) will be: no dual core plan, no more graphics card to support as PCI-E gfx is out of reach and so on. So that is the place where A-EON/Hyperion strategy currently works better for the future, and at least gives some promising expectations to AmigaOS 4.

Now its to Hyperion and community developers to live up to it. And with things they have started to set up after court case (Blog,Support forum, OS4 website, frequent updates, 2D driver for RadeonHD) things started to look better, really. I kind of newer expected it to come this far back in the days when SAM 440 was only and most powerful hardware avail aside used Terron boards.

[wawrzon]

Bluetooth is great thing when sharing stuff with phone

pita. i prefer to connect my phone to the pc via usb any time..

[Hans_]

They should be building a machine that is better priced, fully supported and easily purchased by the average user to sell more copies of OS4 thus increasing the number of users and making some money at the same time.

That sounds great ... until you start trying to think of a way to make it happen and at least break even. Go on, try to come up with a realistic plan. Once you discover how expensive it is to develop and produce one of these boards, and how hard (and expensive) driver development is, I think that you'll give up. The length of time that it took to get onboard audio supported (both A1-X1000 and Sam460ex) should give you a hint as to how hard and time consuming driver developmennt is. Also, bear in mind that people won't be satisfied with Efika grade hardware even if it's affordable.

This may be slightly off-topic but, given the above, here's some of what I like about "Project Cyrus" (the future A1-X2000):
- Freescale have committed to long-term supply of the CPU, so any work done supporting it will last years
- There's an upgrade path to a high-end quad-core machine (via the P5040) without requiring reengineering of the board, or a single new driver to be written
- Choosing an already supprted sound-card instead of having an audio chip on-board lowers board costs, board design & testing costs, and eliminates the need to create yet another driver (or scrounge around for a supported chip that will remain available long-term)

Hans

[Rob]

i meant to recall that at least one of key aros devs was intending to buy into x1k, maybe it was someone else. not that i d consider it a bad thing if there was aros port for x1k;)

Yep, pretty sure phoenixkonsole has one.

[Acill]

That sounds great ... until you start trying to think of a way to make it happen and at least break even. Go on, try to come up with a realistic plan. Once you discover how expensive it is to develop and produce one of these boards, and how hard (and expensive) driver development is, I think that you'll give up. The length of time that it took to get onboard audio supported (both A1-X1000 and Sam460ex) should give you a hint as to how hard and time consuming driver developmennt is. Also, bear in mind that people won't be satisfied with Efika grade hardware even if it's affordable.

This may be slightly off-topic but, given the above, here's some of what I like about "Project Cyrus" (the future A1-X2000):
- Freescale have committed to long-term supply of the CPU, so any work done supporting it will last years
- There's an upgrade path to a high-end quad-core machine (via the P5040) without requiring reengineering of the board, or a single new driver to be written
- Choosing an already supprted sound-card instead of having an audio chip on-board lowers board costs, board design & testing costs, and eliminates the need to create yet another driver (or scrounge around for a supported chip that will remain available long-term)

Hans

Oh spare me the crap about how hard and expensive it is. I work at intel and see other engineers doing things here all the time for nearly nothing on hardware they just toss together to test ideas. As mentioned if the Raspberry-PI guys can build a machine as cheap as they do I dont want to hear how hard it is to get it done. I see it every single day I go to work.

[gertsy]

That sounds great ... until you start trying to think of a way to make it happen and at least break even. Go on, try to come up with a realistic plan. Once you discover how expensive it is to develop and produce one of these boards, and how hard (and expensive) driver development is, I think that you'll give up. The length of time that it took to get onboard audio supported (both A1-X1000 and Sam460ex) should give you a hint as to how hard and time consuming driver developmennt is. Also, bear in mind that people won't be satisfied with Efika grade hardware even if it's affordable.
Hans

+1. Trevor did something no other individual has done in a decade. Soz Acube. And through AON produced a complete product distributed through AmigaKit. Sure it's nigh on $3K, not cheap, but much of the community sees it as expensive when compared to mass produced intel MBs with a multi million market base. Providing a marketable product under a sustainable business plan allows you to invest further if successful. It also reduces risk of falling on your face with realistic sales modelling.  
How much was the original Amiga when it came out in 1985?; $1,300 with 256k and without monitor. I think the cost is comparative all things considered (including the distribution model).

[Hans_]

Oh spare me the crap about how hard and expensive it is. I work at intel and see other engineers doing things here all the time for nearly nothing on hardware they just toss together to test ideas. As mentioned if the Raspberry-PI guys can build a machine as cheap as they do I dont want to hear how hard it is to get it done. I see it every single day I go to work.

The Raspberry Pi is built around a single ARM SoC with everything that they need, including GPU. They also took full advantage of available Linux drivers, and manufacturing at higher volumes than we're talking about. If something similar existed in the PowerPC world, then it would definitely be much easier. But, it doesn't.

Anyway, why don't you prove me wrong then. Go and "toss together" a PowerPC motherboard with PCIe & PCI slots, make sure that every bit of hardware on the board has AmigaOS 4.x drivers before putting it on the market, and sell it for a reasonable price.

Good luck. You're going to need it. ;-)

Hans


P.S. The Raspberry Pi is pretty slow and very limited (e.g., slow SD-Card, no SATA, etc.). It suffers (amongst other things) from bandwidth constraints caused by the CPU and GPU sharing bandwidth. Comparing what A-Eon is doing with the Raspberry Pi is a bit of a joke.

P.P.S. Did you factor in the salaries of those engineers at Intel when you calculated that their exploits cost "nearly nothing"?