@KimmoK

Quote:

And it seems the near future chips are the same, even though they are catching up. But currently, going for ARM would not benefit any desktop use.

With the exception of the top end G5s the current top end cell phone chips are faster than any Amiga available!

The A57s will be faster, but they won't just be going into phones.

ARMs have always been designed for low power use so that limits the performance possible. With the move into servers that all changes because the power available is much higher. This lets you add things like big wide memory busses and things like PCIe interfaces.

Have a look at the new AMD dev board. It's like no ARM board I've ever seen before.

_________________
Whyzzat?

@KimmoK

Quote:

KimmoK wrote: @cdimauro >>So far ARM has been totally inferior for high end embedded use. >OK, so in your opinion this will NEVER change, right? No. I said so far. And it seems the near future chips are the same, even though they are catching up.

See what minator said. Also, take a look at Apple's A7.

Anyway, I stated a different thing at the beginning, and you replied with something else. Take a look at the old messages.
Quote:

But currently, going for ARM would not benefit any desktop use.

Neither PowerPC, since we have only embedded / networking chips. And POWER chips, of course, but only for big servers.
Quote:

>>(IIRC, fastest ARM chips have been 1/6 of the speed of the fastest PPC or intel chips and fastest ARM integrated GPUs have been similarly slow when compared to RadeonHD series) >You're comparing apples with oranges. Why don't you compare similar products? I compare what can be compared.

If you want to know how good is a micro-architecture, you have to take a look at single core / thread, or something similar.

Multicore / thread is also important, but you cannot always squeeze the most of the multiple processing units.

For Amiga o.s./like, SC/T performance is the most important thing.
Quote:

For desktop, we do not want cell phone chip.

But embedded stuff for routers is OK, right?
Quote:

For high end, only Cavium/MIPS and Power Architecture are in the same ballpark with high end i7.

I agree.
Quote:

>For example the dual core Cortex-A15 with a QorIQ (which represents the future of Freescale's PowerPCs)? You'll find some surprises... Even the coming ARM offerings seem to be behind. (for example, look the freescale ARM+PPC roadmap. their PPC core powered offerings will be superior to ARM, in the near future at least. And for example AMD's coming ARM offerings does not yet seem to beat PPCs from freescale.)

What do you mean with "offerings"? SC/T performance? MC/T performance? Features? All of them?

For the AMD stuff, what's your PowerPC chip of comparison?
Quote:

>>"Freescale has disclosed that it'll invest in R&D only for ARMs. That's the crude reality..." >We already discussed some time ago, but I can report again the source: http://www.statesman.com/news/business/freescale-makes-changes-for-stronger-push-into-arm/nTkjs/ >"...We still do have products based on other architectures, but virtually 100 percent of our research and development is directed toward 32-bit ARM-based (chips)." Ok, that's old info from some interview, meant for PR. Official 2013 material says otherwise. (Power Architecture will be developed in/for the high end and customer can select Power/ARm in the mid/low end)

Actually Power is available only for high-end and embedded / router segment. Neither are good for desktop usage.
Quote:

(and 32bit ARM is not high end, etc etc...)

As I said, it's a question of implementation. Take a look at how fast is 32-bit code on Apple's A7.
Quote:

>>To my understanding past ARM chips have been pretty incompatible with each other. >>Meaning that users need to recompile their apps for every ARM chip, or to use JAVA. >The situation is very similar the PowerPCs one. No it's not.

See below my answer.
Quote:

> The only certain thing is that PowerPCs isn't the way to go.... x64 is better for desktop.

I agree.
Quote:

(not 100% sure if it's best for custom desktop)

Can you explain, please?
Quote:

For our niche, PPC gives some advantages.

What? You're talking about desktop performance, but PowerPC are only for high-end servers and embedded / networking (routers, as I said).
Quote:

>You can have extensions to the ISA and/or missing components, as happens to PowerPC too. No. You can run the same binary from PPC601 with FPU etc, on almost every PPC chip. It's not like that on ARM.

You can run the same ARMv2 binary (if the o.s. allows it) on almost every ARM chip.

For PPC601, since some PowerPC chips have NOT compatible FPUs, it's not possible.
Quote:

>>how long will we see fully binary compatible x64 chips?) >What do you mean? The trend seems to be non-binary compatible in the mainstream. (IMHO)

This trend is only for PowerPCs and ARMs. x86 never suffered it, albeit with x64 the ISA was cleaned-up a bit.
Quote:

So, at some point x64 makers will start t6o drop the HW legacy to compete with ARM etc.

The x86 history speaks for itself: it was always retrocompatible. I don't think that it changes, especially now that we have BILLIONS of transistors. Cutting stuff is unlikely to happen.
Quote:

.net makes that possible.

Yes.
Quote:

But for native SW fans (like Amigans) the trend might bring problems.

Absolutely. It's totally different from the usual way.
Quote:

anyway... UPDATE: -I'm more familiar with PPC and due to PPCs desktop legacy, it seems to have better binary compatibility etc. (when compared to ARM offerings) but the truth might be anyway different.

They both suffer of some form of binary incompatibility. In this thread you find something about PowerPC. If you think that ARM is worse, well, you can show me why.
Quote:

-only IMO: there is no forcing need to change CPU ISA for AOS. We have plenty of PPC technology to explore and there are plans for new cores as well.

Routers?
Quote:

-It is possible to do affordable HW around PPC.

Expensive and not for desktop.
Quote:

-PPC has enough CPU muscle for most of the needs, new cores are "only" nice to have

With single-threaded o.s., your multiple cores are sitting down. And SC/T performance is not that good.
Quote:

-Amigalike systems need to start using multicore etc. so the biggest challenge is in SW, let's not make it even harder with a ISA change at this point

When? You don't know. And software needs to be changed. Multicore systems are an expensive and useless option now.
Quote:

-if AOS is planned to start using off-the-self motherboards, then x64 is the way

I don't think it'll happen. I think the AmigaOS4 will die with PowerPCs.
Quote:

-if we are going to have custom Amiga boards, then PPC is ok (or best) for now

I don't agree. See above.
Quote:

But that's just my view. (debating does not bring further new to the topic IMO)

Relax: it's a forum. It's made for exchanging facts, ideas, opinions.

Last edited by cdimauro on 06-Feb-2014 at 07:28 AM.
Last edited by cdimauro on 06-Feb-2014 at 07:21 AM.
Last edited by cdimauro on 06-Feb-2014 at 06:28 AM.

@itix

Quote:

itix wrote: @cdimauro You van have virtual memory without introducing private address space. Both OS4 and MorphOS implement virtual memory but they don't provide private address space to ABox environment.

Yes, you're right.

Unfortunately this way applications consume too much space from the shared virtual address space. But right now it cannot be avoided.

@minator

The qualifier "any Amiga available" is completely new to this discussion. We were discussing what chips to use in the future. The top end PPC SoCs coming out of Freescale RIGHT now are faster than any current cell phone chips. More to the point, they are also faster than any PLANNED cell phone chips. With the exception of video offloading which we SERIOUSLY need to figure out how to do on AmigaOSes of all kind.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

ARM really doesn't matter since amiga uses PPC.

_________________

Quote:

Rob wrote: @gregthecanuck Quote: I suspect the reason behind no on-board audio for the X1000 successors may be due to future plans to leverage the HDMI audio on Radeon graphic cards. I doubt that is the case. HDMI wouldn't be useful to those who don't have a monitor that supports it and would in most cases require a seperate break out box for those that prefer to hear there audio through a seperate set of speakers. Far easier to just plug in a sound card.

Hi Rob/all -

Any modern receiver/amplifier or sound bar has HDMI inputs. If you want to take advantage of 5.1 audio or even 2.0 you need post-amplification. Sound cards on their own won't drive much. That's why they have digital outputs.

Why not use the HDMI support built into the Radeon and bypass the need for a sound card completely? Pick up a nice sound bar and you're good to go.

I would gladly support a bounty for HDMI audio support.

@gregthecanuck

"I would gladly support a bounty for HDMI audio support. "

+1

(sometimes I have had dreams about triple head games console for cost and space efficient multiplayer games ... HDMI to helmet would be optimal...)

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@gregthecanuck

I guess Hans doesn't want to work on it as the only info on how to enable it is in GPLed code. But if he can do everything he has docs for, then someone else could do the reverse engineering of what is needed from GPL, and document this in a way he can use without looking at the original. I think.

At this point we're often stuck in a "noone will do it" situation rather than a "noone will pay for it" situation

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

I'm just hashing out the reason I think HDMI audio support is "a good thing". The practicality of coding and funding such an effort is something outside the scope of this thread. If someone wants to spin up a new thread be my guest.

My understanding is that AMD has released public documentation - see Phoronix article here: AMD Publishes HDA Audio GPU Documentation

Likely this means someone has to develop a "RadeonAudio.resource" and then a matching AHI driver to go with it? No small feat.


Sticking to the PPC roadmap theme, did anyone notice that DDR4 support is in the new T series?

Quote:

Yssing wrote: ARM really doesn't matter since amiga uses PPC.

PPC really doesn't matter since amiga uses 68K.

@cdimauro

Quote:

PPC really doesn't matter since amiga uses 68K.

The NG Amigas obviously use PPC chips as you well know

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

I don't know of any "NG" Amiga. The Amiga brand was used by some Commodore machines, and unfortunately by some recent Commodore USA ones.

I know that there are some PC - x86 + PowerPC machines which are able to run AmigaOS 4 and/or MorphOS and/or AROS, but they are a completely different thing.

@cdimauro

AmigaOne = the NG Amiga for all intensive purposes. If anyone killed the Amiga it was Commodore and we don't owe any brand loyalty to their vision of the machine we love. Hyperion's AmigaOS continues the development of Workbench 1-3 and the AmigaOne machines are the culmination of die-hard Amigans putting their money where their mouth is. Petunia runs classic 68k software and RunUAE runs the classic games. This is as close to a back compatible next generation Amiga system as you could hope for. Commodore, should they have survived, would have moved to the HP-PA cpu and killed the Amiga as we knew it anyway

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

@olegil

Quote:

The qualifier "any Amiga available" is completely new to this discussion. We were discussing what chips to use in the future.

I was answering KimmoK who said cell phone whips were not suitable for desktops. This implies all the Amiga NG machines are not suitable desktops - I don't think many people around here will agree with that.

Quote:

The top end PPC SoCs coming out of Freescale RIGHT now are faster than any current cell phone chips.

That may well be the case but, as I pointed out ARM server chips don't have the same limitations as phone chips so will be rather more powerful - and there's going to be lots of them.

Quote:

More to the point, they are also faster than any PLANNED cell phone chips.

Only for heavy multi core stuff, I wouldn't bet on it for single threaded stuff.

However, this is all a bit academic given the next generation is already planned and announced.

Quote:

With the exception of video offloading which we SERIOUSLY need to figure out how to do on AmigaOSes of all kind.

Which ironically is what the Amiga was originally all about.

Of course you might not like to know that one of these little phone chips with a bit of help from the little phone chip GPU can do HD H.265 *without* using the video decoder.

_________________
Whyzzat?

Quote:

BigD wrote: @cdimauro AmigaOne = the NG Amiga for all intensive purposes.

That's YOUR definition. For a more precise answer, take a look at the PowerPC emulator thread.
Quote:

If anyone killed the Amiga it was Commodore and we don't owe any brand loyalty to their vision of the machine we love.

So you're admitting that you have no right to something else as "Amiga", right?
Quote:

Hyperion's AmigaOS continues the development of Workbench 1-3 and the AmigaOne machines are the culmination of die-hard Amigans putting their money where their mouth is.

They are putting money in a something which is NOT an Amiga, both legally (no right for the brand) and technically (no ad hoc o.s. running on a custom-made hardware).

Also, take a look at the list of machines which can run AmigaOS4: http://en.wikipedia.org/wiki/AmigaOS4#Compatible_hardware

Only SOME machines are called "AmigaOne". And that's the point. That's because AmigaOS4 is only the o.s. which survived to the Amiga MACHINES.
Quote:

Petunia runs classic 68k software and RunUAE runs the classic games. This is as close to a back compatible next generation Amiga system as you could hope for.

Really? I was supposed that the Amiga machines were A BIT different from a classic PC. Just a bit, eh!
Quote:

Commodore, should they have survived, would have moved to the HP-PA cpu and killed the Amiga as we knew it anyway

HP-PA = custom hardware with custom o.s. running on top of it. So, again, something which is A BIT different compared to what you're using now with your AmigaOS4...

@cdimauro

I've become interested of what you would like Amiga NG to be...

Any HW with the name "Amiga"?
A HW that outperforms all others?
Something else?

for me
1) Amiga can be any HW running Amigalike OS (=should "feel" like Amiga when used)
2) it's better if it cand do what others do + more
3) it's better if it takes better use of HW available than others

The name "Amiga" is problematic because it collect flies called "lawyers". Lawyers are far worse than M$ Gates & Ballmer or Commodore Irwin Gould + Medhi Ali.

(I've been diehard fan of AOS implementation. In late nineties I saw that it's the OS that is important on my Amiga4000, not the HW. Nowdays ... I even prefer Amigans to be more important than the original AOS SW code living further. But we must preserve Amiga ideology and to me that is "enable everything, in smart, userfriendly and simple way" or something...)

and for available HW:
-there are powerfull enough chips available for CPU and GPU (no matter what ISA one uses, except 68k)
-only "below EUR100" (desktop like capability) systems are impossible to have with some ISAs (ARM rules on sub EUR100 HW ... with almost desktop like capability)

and from history:
-Commodore chose HP-PA-RISC for "Amiga" 5000 CPU and for 3D
-Commodore "Amiga" became "PowerPC" when phase 5 released PPC card for it
-Escom/AmigaTechnologies "Amiga" chose PowerPC as the NG "Amiga" CPU
-Gateway chose MMC as the NG "Amiga" CPU

To me the logic to glue oneself to retroHW is silly. In modern world using off-the self parts is the way.

Last edited by KimmoK on 07-Feb-2014 at 07:18 AM.

_________________
- KimmoK
// For freedom, for honor, for AMIGA
//
// Thing that I should find more time for: CC64 - 64bit Community Computer?

@KimmoK

It would be nice if we could have ONE technical PPC roadmap discussion without falling into the old pattern of discussing what an Amiga is.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@minator

Quote:

Quote: The qualifier "any Amiga available" is completely new to this discussion. We were discussing what chips to use in the future. I was answering KimmoK who said cell phone whips were not suitable for desktops. This implies all the Amiga NG machines are not suitable desktops - I don't think many people around here will agree with that.

You were muddling the discussion of PPC roadmap. Cell phone chips are unsuitable for desktops not because they lack performance, but because they lack the connectivity you expect on a desktop. Like SATA, PCIe, Ethernet.

Quote:

Quote: The top end PPC SoCs coming out of Freescale RIGHT now are faster than any current cell phone chips. That may well be the case but, as I pointed out ARM server chips don't have the same limitations as phone chips so will be rather more powerful - and there's going to be lots of them.

I haven't seen any documents hinting that these ARM chips will be powerful, I've seen a few hinting that they will have connectivity, though. Making them suitable for desktops, but still not powerful.
Quote:

Quote: More to the point, they are also faster than any PLANNED cell phone chips. Only for heavy multi core stuff, I wouldn't bet on it for single threaded stuff.

Sort of true. Due to stupid freescale using the e5500 without Altivec on the T102x and T104x. But it's not like ARM is taking on i5 and i7 in single threaded performance here, we're still around 2GHz. A15 and A57 are the first ARM cores to beat a same-era PPC on DMIPS/MHz, so they will be faster than e5500 on theoretical performance. However, if there's one thing the P4 showed us, it's that a longer pipeline isn't a good thing for practical performance. e5500 and e6500 has a two-wide 4-issue, 7-stage pipeline, while A57 has a 3-wide, 8-issue, 15+-stage (what the + means I'm not entirely certain) pipeline. How this compares in practical life we have yet to see. e6500 brings back good old Altivec, which to my knowledge beats NEON on efficiency. But it's a bit weird that we keep seeing DMIPS/MHz numbers but nothing on SIMD performance (both architectures). I don't think even CoreMark uses SIMD. SIMD needs more attention, methinks.

However, I only know of 2 things me and my colleagues do daily that wouldn't benefit from rather extensive parallell processing, one is my electronics design software and the other is FPGA design software. Other than that, most things are of a modular nature that can benefit to some degree from having more cores. And if the user (me me me memememmememe!!! Pick meEeee!!11onetwo ) also happens to be doing more than one thing at a time, like browsing the net while something compiles, then more cores than makes sense for each task actually makes sense for the system.

Also, for some reason they support DDR4 on those, but only DDR3 on the ones with more cores. There are some stupid decisions being made there, and I'm worried that's gonna cost them customers. Hopefully the LS-P family will fix some of these shortcomings, but first up is (as has been mentioned before) LS-A, which takes some time to get out due to the peripherals getting an upgrade. Which I'm all for, but there's a limit to any mans patience.
Quote:

However, this is all a bit academic given the next generation is already planned and announced.

Next generation what?
Quote:

Quote: With the exception of video offloading which we SERIOUSLY need to figure out how to do on AmigaOSes of all kind. Which ironically is what the Amiga was originally all about. Of course you might not like to know that one of these little phone chips with a bit of help from the little phone chip GPU can do HD H.265 *without* using the video decoder.

If it's getting help from a GPU then this is a trick we should also be able to use, so I'm very much interested in that, contrary to your beliefs. Also, that's a blatant misuse of the word "ironically", as it's not ironic at all. 20 years of being behind means catching up is hard, nothing unexpected there.

The big question you should ask is:
Would switching to ARM give us help with video decoding on GPU? (hint: no)
Would switching to ARM improve single threaded performance? (hint: theoretically, at least)
Would switching to ARM give support for multiprocessing? (hint: no, in my view it's a step in the wrong direction compared to going all-in with SMT and SMP on x86 or PPC)
Would switching to ARM give 64 bit memory model? (hint: no)

Last edited by olegil on 07-Feb-2014 at 08:19 AM.

_________________
This weeks pet peeve:
Using "voltage" instead of "potential", which leads to inventing new words like "amperage" instead of "current" (I, measured in A) or possible "charge" (amperehours, Ah or Coulomb, C). Sometimes I don't even know what people mean.

@olegil

Yes I agree, well I see lot people escaped to Facebook or Amigans, because they are bored whit this behavior.

@All

I really don't get way people have problem whit AmigaOS4.1, it is continuation of AmigaOS3.x, its runs on Classic hardware like A1200 and A4000, and runs system friendly classic software, and it can even run hardware banging stuff trow a hardware emulator like UAE.

If commodore where going to make Amiga5000 it probable end up whit a standard PC graphic cards, AGA killed Amiga, PC and Mac went the way of true color, and ended up a graphic card that did 640x480 at good speed whit we Amigans ended up whit AGA (320x200 NTSC / 320x250 PAL) at HAM8 or 8Bit, what is HAM8 anyway a hack, that makes the colors look blurry, maybe I'm heartless wanting to trow out AGA, who knows what might have been if Apple stayed whit MC680x0 instead of going PowerPC, but the problem whit that CPU's where getting HOT.

PC's solved it by putting a bigger fan on it, while Motorola / IBM solved it be simplifying the instructions of the CPU by removing stuff, this what RISC is, ARM and PowerPC they are RISC processors.

ARM is successful now only because it sold in higher volume, ARM was used by RiscOS, for many years they did not have a successor, because there was not marked for it, handheld devices where using xScale, ARM was not alternative when AmigaOS4.0 was being planed, and we did have history whit BlizzardPPC and CyberstormPPC, first it was Hagge&Partner that was going to make the new AmigaOS4.x, Hagge&Patnter was responsible for WarpUP, the software that enabled PowerPC programs to run on AmigaOS3.1/3.5/3.9, Hagge&Partner was responsible for making AmigaOS3.5, AmigaOS3.9, when the deal whit Amiga Inc failed, a new contract was signed by Hyperion, they got some of developers from Hagge&Partner, now working on AmigaOS4.x, and also most of the source code.

Yes we do understand that Amiga is not AmigaONE, because its different type of hardware, but AmigaOS4.1 is AmigaOS.

If it was commodore they most likely ended up whit some thing like AmigaOS4.1 and AmigaONE hardware anyway, there where no way AGA was enable to compete whit the Graphic chip companies.

Its a real shame that they did not include any PCI slots in Amiga4000 / 3000 computers back in the day, maybe more software might have worked on AmigaOS4.1, if people head invested in high end Audio sound cards, and modern PCI graphic cards for its time.

Last edited by NutsAboutAmiga on 13-Feb-2014 at 11:27 AM.
Last edited by NutsAboutAmiga on 09-Feb-2014 at 05:02 AM.
Last edited by NutsAboutAmiga on 09-Feb-2014 at 04:56 AM.
Last edited by NutsAboutAmiga on 07-Feb-2014 at 07:35 PM.
Last edited by NutsAboutAmiga on 07-Feb-2014 at 08:25 AM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@olegil

I really appreciate the indepth knowledge you share here with us. I find your postings very constructive and clarifying. I would be pleased if you kept up with it as your postings on a general note makes this portal so very interesting. To me you're the rather unbiassed professional with something to share with us. Please continue!

Quote:

It would be nice if we could have ONE technical PPC roadmap discussion without falling into the old pattern of discussing what an Amiga is.

I whole heartedly agree with this remark!

Lots of people here have an agenda of their own to derail a highly interresting thread into some sorry excuse. Stick to topic, please!!


OldFart

_________________
More then three levels of indigestion and you're scroomed!