What's this "420 MHz Cyberstorm" thing?

https://kato.kuvat.fi/kuvat/Amiga34/DSC06834.JPG

604e only went up to 266 MHz. Is it a typo of 240MHz?

[ Edited by KennyR 12.10.2019 - 20:50 ]

The way I understood it (aka sure way to be wrong ;) )

Quark and basic drivers compiled for AMD64

Everything else still running as 32(31)bit PPC code with all it limitations.

Bootscreen stated 1GB of RAM (quite sure there was more on the mobo).

System is booted via network as no massstorage was available (there was an NVM module on the board)

Basic PCIe GFX card as the onboard GFX isn't supported yet.

AMD64 code can access the ABox API, but must do so in bigendian.


So it is indeed in alot of way like Amithlon only that instead of using a Linux kernel it build on top of a ported Quark.
Just like with Amithlon all limitations of Amiga/MorphOS do apply but one can run native code if it deal the endians in the correct. For Amithlon there was a special GCC build that created bigendian x86 code, but that was to limited/incomplete to be a way forward.

Next step(s) should be writing more drivers to make a full "native" boot possible.

Once that has been done there will be the question how to go for multi-core and full 64bit and here I see 2 scenarios:

1: Throw away most of our API and start fresh resulting in a massive rift, but also in a "modern" system within a reasonable timeframe

2: Keep the ABox and expand/enable the QBox so that ABox apps can spawn x64 tasks to be scheduled over all core.
This pretty much what happened with PowerUP or the original way to run x86 code on Amithlon.
Later on one could add more APIs to QBox till it becomes a full OS and ABox just a legacy box.


Still not holding my breath (or fa##s).

604ev supports 400Mhz, most of PowerPC board here in Poland are upgraded by stachu (including my 603 to 333mhz), some people have 400 604 and it works without problems.

Wow - this is awesome! I was close to giving up hopes on MorphOS x64.

Hopefully we will get a bit more info about what the design of this actually is. I still favour the idea I lined out a few years ago: New MorphOS x64 im LE - similar to current MorphOS witm Ambient etc., but modernized. Plus an underlying BE ABOX that can execute 68k code and x64 flip code (like done with the saif Amithlon GCC). PowerPC emulation would run also in this box. And all that nicely integrated into MorphOSx64 LE (shared clipboard, filesystem, pipes etc)

We'll see. And I am rather confident that the team eventually will implent something nice and useful. Their track record in this regard is preatty good and clean - thumbs up!

Shame I am too tight on time to come to Neuss myself - it's not far away, but I need to do some work tomorrow that should be done on Monday...

Quote:

Although it is off-topic in this topic, it isn't a typo. The card in question had a dead PowerPC chip, and then a few years ago it was sent to Stanislaw Sedlak, a well-known "PPC repair guy" in Poland who specializes in repairing these cards with dead chips, and he brought it back to life by replacing the CPU on it. (Among other repairs he did on the card.) It sports a later, 400Mhz version of the PowerPC 604 chip now (the 604ev a.k.a "Mach5"? I'm unsure.), which is pin and voltage compatible with the CyberstormPPC card, and it was then overclocked to 420Mhz.

This chip was actually manufactured as a smaller process, therefore it dissipates less heat, so at 420Mhz this thing actually runs much cooler than the original 200-233Mhz chips, which really helps with the longevity of these old CPU cards. It is a benchmark-beast when it comes to number crunching, but in real life it's sadly still somewhat handicapped, as the RAM interface on the CSPPC really keeps it back at these speeds.

Here is a closeup photo of the CPU on this very CSPPC: https://twitter.com/chainq/status/811276581226344449/photo/1

(BTW, this system was already presented in this form, with the same card, at Amiga32.)

> 604e only went up to 266 MHz.

604e on stock Phase5/DCE boards only went up to 233 MHz.

> a later, 400Mhz version of the PowerPC 604 chip [...] (the 604ev a.k.a "Mach5"? I'm unsure.)

Correct.

https://en.wikipedia.org/wiki/PowerPC_600#PowerPC_604ev_"Mach5"

> AMD64 code can access the ABox API, but must do so in bigendian. So it is indeed in alot
> of way like Amithlon only that instead of using a Linux kernel it build on top of a ported
> Quark. Just like with Amithlon all limitations of Amiga/MorphOS do apply but one can run
> native code if it deal the endians in the correct. For Amithlon there was a special GCC build
> that created bigendian x86 code, but that was to limited/incomplete to be a way forward.

This reminds me of a statement by the father of MorphOS:

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=7183&start=261 (bottom of #262)

;-)

Quote:

Amiga Os X (MOS X) amiga gui and graphics on top of unix.
Everything else end in vb as worse uae.

@Kronos

Yeah, something like Amithlon sounds plausible. At least at this (“so for now it's just a fun demo, please don't interpret it as more than that”) stage.


@ppcamiga1

*Nobody* (except you perhaps) wants Linux re-branded as MorphOS. Everyone is here for the “Amiga”, and even if would turn out that both binary compatibility and source compatibility would be sacrificed, I am absolutely confident that changes to the API would be kept to a necessary minimum, because the OS developers are dedicated Amigans as well. POSIX has already been discarded countless of times during MorphOS’s two decades of evolution. It won’t happen.

Quote:Quote:

I totally agree!

This was very much needed, it’s now clear (shown, not just sparingly discussed) there is a commitment to MorphOS beyond the dead PPC-platform. Invigorating! Almost makes me want to get myself a new system (970mp maybe?), download the new SDK and finally start learning C or C++ programming!

Quote:

Why do they not release this for classic? I would take this, even with a reduced function set.

Quote:

Because it's not ready. Very few of us have working Classic PPC setups left.

Quote:

I know that statement. And hence, back then, I thought and wrote about the box approach reloaded: On the one hand the legacy compability Abox with 68k and ppc emulation and with that "butt ugly" x64 byte reordering (for fast ports/recompiles) and on the other hand the LE MorphOS NG box to eventually overcome this reordering for new developments. We'll see what it finaly will be.
And as said, because of the good track record, I am pretty confindent that it will be something good.

MOS on x86 will be good enough to justify resignation from win/lnx/osx or it will end as AROS as worse uae.
Amiga gui and graphics on top of unix is one and only chance for MOS.

-Edit-
I see others covered this in more depth than me.

Quote:

There is a guy over on the Amiga Facebook group who offers a service of replacing the stock CPU with a faster one. I never paid close attention to how, but he is doing it and the upgraded boards are out in the wild

[ Edited by redrumloa 13.10.2019 - 11:43 ]

Quote:

This is quite the reveal!! Total game changer!

Quote:

I'll sit here drooling until it is more than a fun demo :-p

All hail Bigfoot!

Funny thing: 68k emulation is usally faster than PPC emulation, right? So app XYZ compiled for 68k would be faster than app XZY compiled for PPC. So you may wish to throw away all PPC software that also exists in 68k version, and for software that exists on PPC only - port it back to 68k (if porting to x64 native is not possible yet, or too much work, or too annoying).

Quote:
Unless the 68K emulator is identical to what is used on actual PowerPC hardware and runs on top of a PowerPC emulator.

Quote:
This is the crux right there: "PPC software that also exists in 68k version" are usually not titles that would need additional performance.

For instance, there are no 68K versions of Odyssey v1.26, Iris, Flow Studio, etc.

Yeah, that's why I said for those you would have to start making 68k versions if you want best possible speed of emulated software XYZ. In case native x64 port of software XYZ is not possible yet, too much work, too annoying, ...

Quote:

IIRC there were some surprising benchmarks posted over at amigaworld that showed 68k emulation on WinUAE is not faster than ppc emulation on WInUAE. I was pretty surprised at least.

In theory there should be little difference in speed between PPC and 68k emulation on x64 if both emulators were properly optimised.

It's only in FPGA where you see the difference. FPGA is not different - it can emulate a PPC as fast as it emulates a 68k, it's that no FPGA yet can quite reach the speeds of any of the PPC used on Amiga/MOS systems. That would need an ASIC solution.

This is the main barrier to dropping all x64 and PPC and using 68k as a "universal virtual instruction set" for everyone in Amigaland. "True" hardware is magnitudes slower than software.

@Georg

I think the beauty of JIT is that it actually compiles the code into native binaries, but using a binary stream as source instead of source code in text format. The end result depends on the quality of the JIT compiler of course. Here it seems like it's the normal 68k->PPC Trance first, making everything PPC, then a PPC->AMD64 compiler translating that to AMD64.

Regarding subject: "MorphOS x86 I want this badly"

https://sintonen.fi/cloud/index.php/s/dksFJsyFdCQWy38

Yeah I'm kinda on that picture too.


Lets say I had my fun.....