It seems that someone is going on with 68000 advanced core and superaga.

http://www.natami.net/knowledge.php?b=6¬e=50153&order=&x=0

They're talking about 4k resolution and other advanced features

_________________
retired

@AmigaBlitter

I have already the turbo Chamaleon64 on my C64 and is really compatible (much much more compared with euae) with a 68030 33mhz performance only gap it is only Ocs/Ecs

_________________
I love Amiga and new hope by AmigaNG
A 500 + ; CDTV; CD32;
PowerMac G5 Quad 8GB,SSD,SSHD,7800gtx,Radeon R5 230 2GB;
MacBook Pro Retina I7 2.3ghz;
#nomorea-eoninmyhome

@AmigaBlitter

Yeah, right, as if. Where's he found a 1GHz 68000 compatible processor?

Anything on natami.org is pie in the sky fantasy, in my opinion.

@Hattig

I personally doubt it is 68000 compatible.

Quote:

AmigaBlitter wrote: It seems that someone is going on with 68000 advanced core and superaga.

Harold is talking about some kind of VLIW processor which the 68000 is not. Maybe it's some kind of new hybrid VLIW processor that's good at emulating other processors. Most VLIW processors have not been successful. There are some big advantages but also some big disadvantages.

Quote:

They're talking about 4k resolution and other advanced features

I know Marcel has been crunching numbers and playing with the Amiga custom chipset with the intention of enhancing it. He is not planning to recreate a new Natami yet. Harold's plans sound a little far fetched. I wish the Natami had continued as the ThoH plan was realistic and the resulting hardware quite advanced in Amiga terms.

@AmigaBlitter

1 GHz 68k CPU, PCIe 3.0 bus. Wow!

68000 1 GHz would be comparable to 603e 100 MHz.
68060 1 GHz would be comparable to 750CXe 533 MHz.

I hope it will be at least as successful (and not in same price league ) as the X1000.

@matthey

Quote:

matthey wrote: Harold is talking about some kind of VLIW processor which the 68000 is not.

The old 68k has a built in VLWI processor, some people think they could invent something new.


Quote:

AmigaBlitter wrote: It seems that someone is going on with 68000 advanced core and superaga.

Please look again, SAGA is hardly mentioned.

All things planned for Natami are lost and the CPU team(s) are working(?) on a CPU core without any relationship to GFX, Sound or a new main board. It seems they want to use it for an accelerator.

_________________
AmigaONE. Haha. Just because you can put label on it does not make it Amiga.

I borrowed this comments from here (#27 & #28):
http://amigaworld.net/modules/newbb/viewtopic.php?topic_id=38873&forum=2&start=20&order=0

Quote:

pavlor wrote: 68000 1 GHz would be comparable to 603e 100 MHz.

A 1GHz 68000 would be designed nothing like the original 16 bit unpipelined micro-coded 68000 without caches. A modern design would be able to take better advantage of the 32 bit 68000 registers and memory accesses although original 68000 code would no longer be optimal. We are seeing that already with 68000 fpga cores. The shorter length instructions of the 68000 and simpler addressing modes make it a good candidate to clock high. Pipelining and caches would make it fly but the ISA would be missing some modern features. It does have integer 16x16=32 and 32/16=16:16 which was amazing for back then but lacking today. Then again, the ColdFire didn't upgrade it although they were more intent at cutting down the 68k than adding features. It would be a quite capable processors still although I would expect some ISA enhancements if it were to become popular.

Quote:

68060 1 GHz would be comparable to 750CXe 533 MHz.

The 68060 is one of the best processors designs of all time. However, Motorola liked to cut down processors and functionality even then. The 68020 ISA was almost unchanged for the 68060 except they removed the integer 32x32=64 which GCC was already using to turn divide by an immediate into a multiply (invert and multiply) with a huge cycle savings. They could have quite easily made integer 16x16=32 multiply 1 cycle and put it in both integer units. They should have put SWAP in both integer units considering there is only immediate shifts up to 8, it has a longword result for forwarding and it's very simple. Every CPU has the pencil pushers get in the way though. The 68060 is an absolutely amazing processor considering the restrictions. Cut down modern processors to run with the same restrictions and they won't outperform the 68060. Expand the 68060 to run with modern processor features and bandwidth and the 68060 would compete. Enhance and modernize the instruction fetch, cache sizes and ISA and the 68060 would be executing 2+ instructions per cycle without OoO while being easy to schedule instructions due to the early instruction retirement with forwarding. It's already a beast in memory and a cache miser while being one of the easiest processors to program. Too bad the pencil pushers decided PowerPC was the future while closing their eyes to performance and resources used. A modernized 68060 would not be one of the highest clockers but I think it would be one of the strongest performing non-OoO processors for it's clock speed of any CPU.

@matthey

So, what performance we can expect from that new 68k core?

@pavlor

I'm guessing none, since it seems even more pie-in-the-sky than 68050, and that also amounted to nothing as of yet.

_________________

@matthey

Quote:

Too bad the pencil pushers decided PowerPC was the future while closing their eyes to performance and resources used.

It was more the fact that RISC was eating Motorola's lunch in workstation and embedded markets and Motorola didn't have the luxury of having a huge PC market to fall on to.
Besides, they got PowerPC from IBM pretty cheap. It was a deal that would be hard to pass on, 68000 was falling, 88000 never even catched on...

_________________

Quote:

pavlor wrote: So, what performance we can expect from that new 68k core?

I don't know what Harold is talking about. I know of no 1 GHz 68k CPU, especially VLIW. I only know about the Apollo fpga core. It will not be competing with any modern processors while it's in fpga. The fpga requires a longer pipeline because timings are sometimes worse than in silicon. This weakens the processor and the maximum clock speeds will be limited by internal fpga clock speeds. Note that I am not the technical guy either so my guess is only a little better than none. The Cyclone II of the Vampire 600 is very cramped but speeds approaching 100MHz with performance between a 68040 and 68060 are my best guess. The cut down Apollo (Phoenix) is still superscalar but will have a low IPC and will be difficult to schedule for. The caches will be the size of the 68040 or smaller which will limit performance. I'm not sure what will be cut and won't compared to the Apollo. The Apollo is targeted at a Cyclone V or higher. I believe the clock speeds were 150-200MHz in simulation but it's been awhile. Altera is working with Intel to double the internal clock speeds of the Cyclone V up so that could be doubled in the next year or so. The Apollo should be able to outperform the 68060 even though some instruction timings will be worse and the pipeline longer (missed branch penalties are bigger). The superscalar should be stronger and maybe easier to schedule for than the 68060 if the primary and secondary integer unit swap idea works. There is more instruction folding/fusion to internal 3 op instructions and loops don't need to be unrolled with this help like the 68060 saving ICache. There is a much larger instruction fetch than the 68060 and ISA improvements should further reduce the average instructions size and hopefully increase code density by 5-15% (includes some ColdFire instructions) . There is early instruction retirement and result forwarding like the 68060 which helps with instruction dependencies and scheduling. Cache performance is one of the strengths with little to no penalty for unaligned reads and writes. The DCache will be writethrough only (good for Amiga game compatibility). There is a design for stream detection for efficient prefetching of memory for streams and to keep the DCache from being trashed. There is 2 bit saturating branch prediction and short forward branches may be automatically turned into conditional instructions. There is a link stack for quick return from subroutines. There has been partial work done on a 68k compatible FPU and an Altivec like SIMD but they are nowhere close to being completed. I can't even give any dates on the Apollo integer core although it should be done by now as planned. This is a big and complex project though. Work is continuing. Whether this will ever all work together I don't know. There is a lot of work that goes into making even a semi-modern processor.

Quote:

WolfToTheMoon wrote: I'm guessing none, since it seems even more pie-in-the-sky than 68050, and that also amounted to nothing as of yet.

I wish Jens would have released it as is instead of cleaning it up. The VHDL code is very neat and well commented in English. I don't know why it wasn't working in the Natami but Jens was working on it. There is some work that goes into mating an fpga CPU to hardware. The TG68 can have issues in the Vampire, MiniMig and fpgaArcade even though that core is solid.

Quote:

WolfToTheMoon wrote: Quote: Too bad the pencil pushers decided PowerPC was the future while closing their eyes to performance and resources used. It was more the fact that RISC was eating Motorola's lunch in workstation and embedded markets and Motorola didn't have the luxury of having a huge PC market to fall on to. Besides, they got PowerPC from IBM pretty cheap. It was a deal that would be hard to pass on, 68000 was falling, 88000 never even catched on...

There is certainly truth to what you say but it was all about the hype of RISC with it's limitless clock speeds and perfect compiler support that would dominate the future. Almost everyone was sold on this flawed misconceptions. The 68060 made more sense with the cost of memory back then. The 68060 clock speeds weren't as high as PowerPC but the 68060 outperformed many of them despite the clock, memory bandwidth and cache advantage of PowerPC. Motorola didn't properly evaluate their own product or try to sell it. They jumped ship based on promises instead of statistics. Now we are stuck with x86_64 and no competition for affordable performance processors.

Last edited by matthey on 21-Feb-2014 at 02:20 AM.

@matthey

Quote:

Almost everyone was sold on this flawed misconceptions. The 68060 made more sense with the cost of memory back then. The 68060 clock speeds weren't as high as PowerPC but the 68060 outperformed many of them despite the clock, memory bandwidth and cache advantage of PowerPC.

Hi!

Thanks for the interesting discussion! Could you explain why the idea about RISC are misconceptions... I have them too

Thanks!
Tygre

_________________
Tygre
Scientific Progress Goes Boing!

@matthey

a 20% difference in code size will be COMPLETELY dominated by a 100% increase in clock frequency, which is what happened when 68k was replaced by PPC.

That's NOT a misconception. Freescale has had plenty of reduced 68k cores on the market after the 68060, but that core simply CANNOT scale as high as PPC without resorting to a LOT of strange tricks (in the i7 class of tricks, like we're discussing in that other thread).

Obviously, if Motorola HAD put all eggs in the 68k basket like Intel did with their x86, the story might have been different, but looking back at this point is not really productive...

_________________
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

Quote:

if Motorola HAD put all eggs in the 68k basket like Intel did with their x86

Intel never did that... Like AMD, they also developed numerous RISC CPUs and intended for IA64 to replace x86 CPUs.

The biggest difference between Motorola and Intel was that in the early 90s PC market really boomed, while on the other hand the 68K personal computers entered in a death spiral. That allowed Intel for a generous cash inflow and gave them a reason to continue developing x86 even while they were designing IA64 which was meant to completely replace x86.

Motorola hardly had any market left for high end 68K by 94', even without Apple going PowerPC - Amiga was gone, so was Atari and even Apple wasn't doing that good. The only viable market was embedded, which Coldfire took care off.

_________________

@WolfToTheMoon

Decisions are always done in a certain situation also influenced by emotions and trends and not only facts. I can remember how everyone claimed that the future would be RISC replacing CISC and everyone not convinced only oldfashioned. Now CISC beats RISC. I do not know if the decision of Motorola to end development of its own processor was a good one, today the company more or less has vanished so it cannot be seen as a "success".

@OlafS25

If we're honest, Motorola lost the battle against Intel the minute IBM decided for Intel's CPU in the PC.
That decision was so final that even IBM lost again, when they tried with the PowerPC/POWER to disrupt the Wintel monopoly.

Commodore's biggest chip mistake was not investing in their own designs past 8 bits. 6516 was ready, but was cancelled. So was the 32 bit project.

_________________

@WolfToTheMoon

Did not know CBM designed any CPU. (links to learn would be nice)

They designed chips like AGA chip set and had them manufactured elsewhere (by HP, IIRC).

For PPC... I did not see the co-operation with apple+moto+IBM to be very successfull. To me it seemed that they did not agree much. It seems more that the co-operation slowed things down. IBM should have made some effort to support consumer PPC systems and not to let Apple go around killing other developments (they killed PPC clone computers, they killed PA Semi, ...).

Now they seem to support Linux for PPC, but it might be too late...

And I think late 90's IBM+Motorolla thought MIPS and ARM is already dead, so they became lazy.

@Olaf
"today the company more or less has vanished so it cannot be seen as a "success"."

Freescale at least is one of the few bigger players still. PPC has been success in embedded as it is only now starting to face real competition.

Last edited by KimmoK on 21-Feb-2014 at 11:19 AM.

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

@KimmoK

I more think of "Power" processors used for desktops (what we need, "Amiga" is desktop). Embedded market consists of partly very special small applications (in industry, machines and so on) or is used in large scale in cars. You can still live there even with exotic processors as long as you offer special solutions to the customer but there it is more important to have high reliability and low consumption than high performance. Hardly the market Motorola was about and more or less useless for AmigaOS or MorphOS. I am collecting benchmarks right now (lame) and even emulated 68k can already outperform SAM 460, that shows how far ahead Intel/AMD performance wise already are.

Regarding Freescale I only read that they are investing in ARM. Do they still invest in high-performance PPC processors?

Last edited by OlafS25 on 21-Feb-2014 at 11:35 AM.
Last edited by OlafS25 on 21-Feb-2014 at 11:34 AM.