@pavlor

I'm excited about this initiative.
I do not know if they did manage to make it work.
But it seems to me the best chance to see the Amiga reborn.

If this works ..... this is 100% Amiga hardware in 2014

Last edited by drstrangelove on 29-Jul-2014 at 05:32 PM.

@drstrangelove

Quote:

But it seems to me the best chance to see the Amiga reborn.

Price is also important factor (hopefuly not X1000 range ).

@pavlor

I think these guys do not think about selling it.
Hope I'm wrong.

I just saw ..... Nvidia 9800 GT costs over 1000 euros!!!!

Last edited by drstrangelove on 29-Jul-2014 at 05:56 PM.

@drstrangelove

Quote:

I just saw ..... Nvidia 9800 GT costs over 1000 euros!!!!

On my planet the Nvidia 9800 GT is a very old card. One can buy it on ebay for under $100.

_________________
All the way, with 68k

@68k

I doubt any 68k/emulated system can be the "main branch" of future Amigas:
http://via.i-networx.de/bench_en.html

Available PPC systems already are (up to) hundreds times faster than high end 68k.

Last edited by KimmoK on 30-Jul-2014 at 08:12 AM.

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

@agami

Upssss.

Well do you right.
In the list of hardware, the card is the Nvidia GTX 690

But even if it does not work .... why use it?

Not understand it

Quote:

KimmoK wrote: I doubt any 68k/emulated system can be the "main branch" of future Amigas: http://via.i-networx.de/bench_en.html

I think UAE on a high end x86_64 is already up there with PPC. The software benchmarks in the link likely both use floating point. The 68k FPU is extended precision and all others processors are probably cheating by using double precision. The software is probably compiled for 6888x and going through traps on the 68040 slowing down by half or more. The 68040 also removed the FINT/FINTRZ instruction which was a huge mistake that the 68060 corrected. The 68060 FPU is still no speed demon but I think it could be up there with low end UAE emulation and Efika emulation in a fair comparison. The small cache size hurts the 68040 and 68060 more for floating point performance more than integer performance. I get ~25fps in QuakeGL 640x480x16 on my CSMK3 68060@75MHz and the code is still way less than optimal. An fpga FPU can solve the cache problem and provide more memory bandwidth for moving fp numbers around but it has to be highly pipelined and the latencies are killer because fpgas can't shift big numbers very fast. Moving to 64 bit operations like the other processors that cheat would help the speed some. Floating point performance in an fpga is likely to be a weak point and disappointing. Integer and memory benchmaks are a different story though. Moving to an ASIC changes everything.

Quote:

Available PPC systems already are (up to) hundreds times faster than high end 68k.

In some areas no doubt. The PPC is good at floating point and Altivec performance. We don't have to compete in performance any more than AmigaOS 4.x computers have to compete in price. Making responsive, classic compatible and cheap Amigas may be enough to out sell AmigaOS 4 PPC computers ;).

@KimmoK

no they are not...

a couple of benchmarks:
http://www.aros-platform.de/html/benchmarks.html

regarding comparation PPC and FPGA that has to be seen when the accellerators are working and workbench booting. PPC will still be faster but not hundreds of times

Besides main branch does not depend on hardware but on user numbers. In my view it will be a mixture between different platforms, expecially X86/X64/ARM native and different 68k options like FPGA and emulation. PPC will because of price have problems to win new users and will play no big role anymore.

@KimmoK

You use Benchmarks from 2011?

19.1.2011

Sorry to say that, that are more than 3 years. Might be that this is not much in "amiga-terms" where people measure progress in decades but that is a long time in normal world. So hardware has become much faster since that so your conclusions is not valid anymore.

BTW I own Cinema4d. If I find the scene that was used here I will repeat the test with a more recent configuration. And even my configuration is not brandnew and I am pretty sure that with new hardware from 2014 the results will be different again. As long as AmigaOS and MorphOS only support one core they will loose more and more ground. And now I only talk about 68k (both emulation and FPGA because both are hardware that is in development).

Last edited by OlafS25 on 30-Jul-2014 at 10:21 AM.
Last edited by OlafS25 on 30-Jul-2014 at 10:11 AM.
Last edited by OlafS25 on 30-Jul-2014 at 10:10 AM.
Last edited by OlafS25 on 30-Jul-2014 at 10:10 AM.

@drstrangelove

Me neither. I'm not a fan of ATI, normally, but at least on AmigaOS RadeonHD + the driver works fine and performance is fine as well so for AmigaOS RadeonHD is the way to go IMHO.

_________________
Proud user of AmigaOS 4.1 on an AmigaONE 500. This is the first Amiga I've ever had so I don't know all the ins and outs of AmigaOS yet, so I'm sorry if I'm asking noob questions and stuff.

@KimmoK

I don't see how an emulated chipset on PPC can be the main branch of an Amiga. I don't care as much about the CPU because there are many competing standards there. What I do care about is that there is a large amount of hardware-banging games out there that don't run well on a PPC.

Last edited by Samurai_Crow on 30-Jul-2014 at 11:50 AM.

@Samurai_Crow

"emulated chipset on PPC"

I hope RETRO is not Amiga's future main branch.

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

@KimmoK

PPC is retro too (from a certain point of view). FPGA is used in industry, you cannot call it retro.

68k is a compiler target not the devil

@OlafS25

"You use Benchmarks from 2011? ... So hardware has become much faster since that so your conclusions is not valid anymore."

Those tests did not use y2011 HW.
(and the PPC & 68k HW used has not become slower/faster in meanwhile, also I doubt big improvements have happened in emulation speed either)

But that is not so relevant as:

My point was that native (power UP, no altivec) application on 1.5Ghz PPC can be hundreds of times faster than on any real 68k.

So, I think it would not be realistic for FPGA 68k to come close to PPC in performance.

+taking into account that a lot of people (x64 fans) consider 1.5GHz PPC to be too slow for any modern use, I do not see 68k as the solution for future modernish amigas (including web browsing and games).
(even though superb cool for retro/oldschool perhaps)

From FPGA side of our niche, I would love to see NG Amigalike graphics + audio, rather than AGA compatibility. RadeonHD should be used to max out 3D speed & openCL speed, if we once manage to build support.

Last edited by KimmoK on 30-Jul-2014 at 02:18 PM.

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

@KimmoK

they did not use 2011 hardware, they used older hardware even

my own emulation (2 core system from 2011 with 4 GB RAM) is between SAM 440 and SAM 460, more recent hardware beats SAM 460 and newest hardware is certainly already faster

Last edited by OlafS25 on 30-Jul-2014 at 02:18 PM.

@OlafS25

"my own emulation (2 core system from 2011 with 4 GB RAM) is between SAM 440 and SAM 460, more recent hardware beats SAM 460 and newest hardware is certainly already faster"

Still does not make real 68k or FPGA 68k any faster. (my point)


btw. does your y2011 system run 68k code faster than SAM460 runs native code?
(my secondary point is that even emulated 68k on x86 need to work hard to beat native PPC apps on x1000 and MosG5)

UPDATE:

My points:
-We do not ”need” 30 000MIPS from single core.
-We need some 4000 – 6000 MIPS for normal user (per core)
-We need to start to use multicore (computing world is multicore)
-I doubt FPGA 68k will be able to deliver 4000+MIPS in our lifetime.
-I doubt FPGA 68k will be able to handle gigabytes of fast and “chip” RAM (needed for some modern computer use cases).

Using FPGA graphics + audio:
-would be cool
-might be usefull in minimalistic configurations with PPC cores
-might ease the burden of trying to support alien GPUs and audio chips.
(for some uses modern GPU is anyway needed, FPGA can not compete with RadeonHD in 3D or OpenCL)

I like retro, I like to “play” with 68k HW and I would love to use modernish Amiga system as my main computer (not needing all those other systems).

Last edited by KimmoK on 30-Jul-2014 at 02:51 PM.
Last edited by KimmoK on 30-Jul-2014 at 02:42 PM.

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

@KimmoK

Gunnar told me that next generation of FPGAs (available next year) are fast enough for most purposes (still pricey at the start). We will see when the new accellerators are ready how they compare. FPGA based amigas are certainly more toys, second computer for amigans and technic nerds who like the idea of a programmable hardware. And if they are cheap enough there might even be a bigger market. But they certainly are not for everyone. Because of that I wrote ARM/X86/X64 based devices (that include emulation). I do not think that aging Macs or super-expensive new PPC desktops have a big chance. You get hardware cheap at every corner as long as it is not PPC.

@KimmoK

I used the lame test (because there were more numbers available) and my own system was between SAM 440 and SAM 460, Pavlor did the same on i5 beating 460 and yes running native 68k code.

And how fast FPGA really is cannot be said before workbench is booting. At least for now it will be slower than fast native PPC. But FPGA are a small but important segment, a kind of fun environment and for that it is not important if it beats PPC or not (yet).

"real 68k" What is that? Who cares. I hope that the accellerator will rise the level of "real 68k" considerably so it will become a different discussion.

Last edited by OlafS25 on 30-Jul-2014 at 02:35 PM.
Last edited by OlafS25 on 30-Jul-2014 at 02:30 PM.

@OlafS25

"You get hardware cheap at every corner as long as it is not PPC."

None of that cheap HW run modernish Amiga stuff as standard. (unless someone considers WinUAE+Amikit as modernish Amiga)
With my luck, the only way to find compatible AROS x86 HW is to buy pre-installed system (if that exist with 100% component support), that's not available every corner.

Once AROS becomes as usefull as MOS or AOS4, I welcome it to my desk with fully supported HW.

(if "fully supported" HW means custom HW, the PPC is not more expensive than other ISAs)

@using multiple cores

(from Hyperion MP post)

I read it that we will see some (distant??) form of SMP on AOS4.2 after all (without sandbox in use even).
I assume xkernell will be demonstrated later this year, perhaps x5000 betatesters get it with their HW for christmas.

Last edited by KimmoK on 30-Jul-2014 at 03:05 PM.
Last edited by KimmoK on 30-Jul-2014 at 03:04 PM.
Last edited by KimmoK on 30-Jul-2014 at 02:49 PM.

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

@matthey

Quote:

Moving to 64 bit operations like the other processors that cheat would help the speed some.

for what I have heard there is little or no speed improvement between 32bit vs 64bit programs, if you have a 64bit version of windows installed you should install a 64bit program, if you install a 32bit program it has to be handled by translation layer, that adds 2% speed penalty.

64bit programs can also be slower in some cases, because 64bit programs are fat, and fill the instruction cache quicker, this can result in more instruction caches is needed.

Anyway there are advantages and disadvantages,

so it pointless to talk about 64 or 32bit, this mostly about about Ghz.
and also number of cores does help, programs are primarily serial, AltiVec/VMX/MMX and so on this are more flexible, but require the developer to make an effort to support this, and it might be possible for the CPU emulation to take advantage of this, but I wonder if there is not going to be some overhead, and maybe its not even worth the effort.

So what you need up whit is one CPU core emulation one CPU, while the rest of the core's are ideal, waiting for some thing to do.

Last edited by NutsAboutAmiga on 31-Jul-2014 at 04:32 PM.
Last edited by NutsAboutAmiga on 31-Jul-2014 at 04:31 PM.
Last edited by NutsAboutAmiga on 31-Jul-2014 at 04:28 PM.

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