68080/68060 discussion, comparisons etc

[FREEMILK]

kind of pointless a 100 GHz 68060 with all the add ins will not change the fact there is no software . nothing that can use more power. games need the old slow 68000 mostly. Workbench even 4.1 Forever Edition is lacking hardware support and no modern browser , so what would you do with that kind of power in an Amiga.

dont get me wrong , I want to see better things , these accelerators TG68 at low cost are perfect. I was a X5000 owner and there is just nothing to do with it . I ended up selling it. took a big loss, and bout a Ryzen thread ripper 1950x , system with the money

[Chucky]

oh too my tl;dr here..

but.. oen thing is nice with the cult of vampire. many thinks that just if people add new instructions, 64bit etc it automatically gets better! same as some seems to think that just add 16bit audio into vampire and protracker etc automatically will support it...

guess what: it will not.

[AmigaHope]

Originally Posted by Chucky

oh too my tl;dr here..

but.. oen thing is nice with the cult of vampire. many thinks that just if people add new instructions, 64bit etc it automatically gets better! same as some seems to think that just add 16bit audio into vampire and protracker etc automatically will support it...

guess what: it will not.

Add 16-bit audio into vampire with an AHI driver and anything that supports AHI (including many trackers) will automatically support it.

Add new instructions and replacement for commonly used libraries and everything that uses those libraries will support it.

Yes it won't make Shadow of the Beast any better but it will makes tons of stuff that actually uses the OS better. This kind of stuff is not for hardware banging games coded strictly for A500/A1200.

[Chucky]

well where are those libs?

if you talk to vampirecult-members many seems to think that it is something magic that "just will work" but this is also the same people who thinks it is hard to copy 4 files to libs:

[Promilus]

Oh, everything just works out of the box?

Add new instructions and replacement for commonly used libraries and everything that uses those libraries will support it

Yes... so few years after V2 - what libs are rewritten to support those features from AC68080? Yeah, exactly. Suddenly jpeg decoding done on STM32 or ARM inside ZynQ (@ARM as well) makes sense. And with Cyclone V SE V4 could also use ARM inside FPGA to do things faster the same way now use new 080 features to do things faster. The only difference would be... well, you could run that mpeg2 with higher resolution perhaps? Gunnar knew that. Gunnar chose AMMX anyway.

[kipper2k]

as an original member of the apollo team i was excited to see the development of the core. when it first was released it was at a budget price of $199 USD and was good value foe the money. The core was not yet finished but plans were under way to make it more compatible for (Existing) software, as after all there is very unlikely to be anything that is new for the Amiga that can be enjoyed by the masses.

With the advent of the Raspberry Pi, Mister etc there are now ways to have an emulated Amiga inside an Amiga case running much faster than an 68080. I have always wanted to use my Amiga for amiga things under the expectation that i can load all my favourite games quickly, without too much added expense and use my Amiga for Amiga-esque things and use tablets etc for the other things.

There is no denying that the current core has power but ultimately it comes down to whether someone will pay all that extra dosh to play something a basic or regular accelerated Amiga can do just as well

PiAmiga speed is 3 times that of the Vampire, and has more options for compatibility (Calculate the dhystone/$ ratio to see what the value is like

[Promilus]

That's why I'm opponent when it comes to speed focused accelerator. It doesn't have to be as fast as V4 because there's little advantage of that. Everything you need along with jpeg and mpeg2 decoding can be reliably offloaded to different hardware, there are things where is room for improvement like mass storage and it's speed. Or network and it's speed. 080 or not - there's no way for classic amiga to erase the gap between 68k and the rest of the world. Ironically - even cheap ARM emulating 68k is faster than 080 implemented inside FPGA. So going into faster 68k isn't direction I find as important. Decent CPU speed along with great SDK and easy-of-use features like maybe SATA, USB, WiFi - that's what Amiga needs.

[hammer]

Originally Posted by FREEMILK

kind of pointless a 100 GHz 68060 with all the add ins will not change the fact there is no software . nothing that can use more power. games need the old slow 68000 mostly. Workbench even 4.1 Forever Edition is lacking hardware support and no modern browser , so what would you do with that kind of power in an Amiga.

dont get me wrong , I want to see better things , these accelerators TG68 at low cost are perfect. I was a X5000 owner and there is just nothing to do with it . I ended up selling it. took a big loss, and bout a Ryzen thread ripper 1950x , system with the money

I have ASUS ROG Strix X570 platform ready for Ryzen 9 5000 series. Currently on Ryzen 9 3900X. Ryzen TR 1xx0 is a dead end platform, inferior AVX2 and PC gaming performance.

I prefer something better than 68030 at 50 Mhz since I use to have A3000/030 at 25Mhz and my Wicher 508's 68000 at 50Mhz is close to it. My A1200 has 8 MB fast ram + 68882 40Mhz card which improves performance, but it's about 68% of my old A3000.

TG68K needs to be better e.g. reach 040 at 50 Mhz performance levels.

[hammer]

Originally Posted by Promilus

That's why I'm opponent when it comes to speed focused accelerator. It doesn't have to be as fast as V4 because there's little advantage of that. Everything you need along with jpeg and mpeg2 decoding can be reliably offloaded to different hardware, there are things where is room for improvement like mass storage and it's speed. Or network and it's speed. 080 or not - there's no way for classic amiga to erase the gap between 68k and the rest of the world. Ironically - even cheap ARM emulating 68k is faster than 080 implemented inside FPGA. So going into faster 68k isn't direction I find as important. Decent CPU speed along with great SDK and easy-of-use features like maybe SATA, USB, WiFi - that's what Amiga needs.

SATA is not a major problem if the Amiga config has IDE interface and KS ROM supporting beyond the 4GB limit. My A1200 has a PCMCIA wifi card which I haven't configured. IDE can be converted into SATA.

100Mbit NIC has 12.5 MB/s speed, hence it may need some CPU and fast ram upgrades.

I'm using a USB-to-Amiga mouse port adapter for USB PC mouse.

Both modern Arm8 and X64 beats 68060 and 68080 at their fastest clock speeds.

I ready have an Xbox Series X killer gaming PC rig and Surface Pro 5 tablet for WinUAE, hence I don't need a 1080p H264 limited Pi4.

[hammer]

Originally Posted by Promilus

@hammer - your whole post just confirms you missed the point. You can have SIMD as wide as you like and it has nothing to do with CPU architecture. It's just yet another peripheral unit which was integrated into CPU along with MMU, FPU, memory controller and pci (now pcie) controller. You've chose SIMD along with 4g+ world... it has nothing to do with it. That was the point of my post. Address calculation is AGU related (which is part of integer pipeline of CPU alone) and addressing capabilities are related to how many address lines are out there. 32bit chips had in reality 36 address lines - that's how PAE was created but mostly only server line systems used more than 4GB RAM and the 2GB limit of RAM per process (per process, not per application) still remains. Doesn't mean you actually need continuous 4g+ of RAM to make bigger game. You don't. It just MAKE THINGS EASIER!

Your 2048 bit SIMD argument is flawed when it's a theoretical distraction.

Offer ARMv8 CPU with 2048 bit SIMD + motherboard at Ryzen 9 3900X+B550 motherboard's price range.

I'm tired of the promises made by ARM, IBM and Motorola/Freescale/NXP. Where's your product offerings at reasonable price range?

64bit CPU is required for Xbox One/PS4 era games, your distraction wouldn't change that fact. PAE memory segmentation model has performance penalty.

IF heavy vector math is needed, I have GpGPU with RTX 2080 Ti hardware raytracing acceleration. AMD's Big NAVI is about RTX 3080 level. Your A64FX example can go jump the lake.

[kipper2k]

it appears this thread seems to be descending into anarchy and flaming, there seems to be a pattern and definately heading off topic. shame

[Promilus]

IDE can be converted into SATA

Yes, but inherent flaws of IDE PIO remains so what good it does?

100Mbit NIC has 12.5 MB/s speed, hence it may need some CPU and fast ram upgrades

Well vampire has ENC28J60 running @ SPI which already limits it pretty bad. Warp1260 has ESP32 which runs through ~1mbit UART - that's even more limited. Since there's no really fast peripheral bus @ classic amiga and it's newest turbo cards there won't be any really fast ethernet and storage. And as I said that's area one can focus more than new ISA.

Both modern Arm8 and X64 beats 68060 and 68080 at their fastest clock speeds

No way!

Offer ARMv8 CPU with 2048 bit SIMD + motherboard at Ryzen 9 3900X+B550 motherboard's price range

You're going way out of the context here.

I'm tired of the promises made by ARM, IBM and Motorola/Freescale/NXP. Where's your product offerings at reasonable price range?

Well this is topic about 68080 so most likely about what softcore use in future fpga-based turbo cards for classic amiga. During that discussion there was opinion that AMMX and 64bit instructions are pretty much useless in classic amiga. Then you went with that silly AMD64 comparison. I rectified some things but still stayed within the topic. You went nuts with all those xbox, ps5 etc. comparison and misunderstood a simple fact - simd width has NOTHING TO DO with addressing capabilities NOR internal architecture of CPU alone. And that's the context behind my example with SVE.
That's a direct response to this BS:

PC's SSE2/AVX/AVX2 hardware has support for full-speed/double speed/quad speed SIMD INT64 and FP64 to support very large scale game worlds that exceed 4GB ram

- addressing capabilities has basically NOTHING TO DO WITH IT!
So please either stay on the topic or ask moderator to cut off all that x86/aarch64 OT to another thread (thrashbin perhaps). One fact remains - there were FPGA with PPC HPS (but cut down and not very powerful), there are FPGA with ARM HPS (decent and fairly powerful). There are no FPGA with x86 HPS (despite Altera being in intel hands for few years now).

[Tigerskunk]

Originally Posted by FREEMILK

kind of pointless a 100 GHz 68060 with all the add ins will not change the fact there is no software . nothing that can use more power. games need the old slow 68000 mostly. Workbench even 4.1 Forever Edition is lacking hardware support and no modern browser , so what would you do with that kind of power in an Amiga.

This post is the best of the whole thread.

And that's the problem, folks. You can make your cards as fast as you want, but in the end all you do is making your Amigas more PC like, without the benefit of better software using that.

In the beginning I hoped that there would be a slew of new software (or even coders coming back?) benefitting from the new possibilities of the Apollo core. Looked good in the beginning, with that Cannonball port and other stuff.

Now, a few years later we know that this didn't really happen, unfortunately.

@Promilus: That's "Hammer" for you. That has been his shtick since like forever on other forums like Amigaworld, etc. Just don't mind him, and god forbid starting a discussion with him...

[Promilus]

Well now I learned my lesson. If I only had known that sooner.

but in the end all you do is making your Amigas more PC like

That's true if all you get is being able to run PC port (faster). But there aren't any native amiga games which require faster CPU or benefit greatly from it. Even if pushed performance even further what will you get? OpenRA games? GemRB games? VCMI? That's not even funny.

[grond]

Well, what did you expect? The Vampire has moved the Amiga to the capabilities of a mid-90s PC. Did you expect people to write new and Amiga-exclusive mid-90s games? We can be glad that some people made some ports. Nonetheless a lot of Amiga software does really nicely with more processor speed. It certainly isn't about OCS A500 games.

[kamelito]

Originally Posted by Chucky

oh too my tl;dr here..

but.. oen thing is nice with the cult of vampire. many thinks that just if people add new instructions, 64bit etc it automatically gets better! same as some seems to think that just add 16bit audio into vampire and protracker etc automatically will support it...

guess what: it will not.

If I’m not mistaken Marlon Milky tracker port added V4SA support.

[Chucky]

well not "automatically" somone needed to do the coding to handle it.
but when I read posts from cultmembers you can almost read that they actually believes that it will atomatically fix it. no clue whatsoever that you need to do coding etc

[8bitbubsy]

Yeah, adding 16-bit sample support to ProTracker would be a big project. There's a huge amount of hardcoded Paula + 8-bit sample code in there. This probably goes for most of the Amiga trackers...

[NovaCoder]

Originally Posted by FREEMILK

kind of pointless a 100 GHz 68060 with all the add ins will not change the fact there is no software . nothing that can use more power. games need the old slow 68000 mostly. Workbench even 4.1 Forever Edition is lacking hardware support and no modern browser , so what would you do with that kind of power in an Amiga.

Yes that's what I have said a few times now, a super powerful 68k like CPU is almost pointless to have these days.

Of course you could go and compile specific software to use this new super powerful 68k like CPU but then you have to ask yourself why you don't just code on the Pi which would be a more powerful (hobby/niche) computer to develop for

For the record my 'dream' FPGA 68K CPU would be something that can behave like a 100% compatible 68000/020/030/040/060 by just switching out the core. Something that could behave like a combined 040/060 CPU would also be great to improve compatibility in some games. Speed is important to me but 100 MIPS is plenty fast enough to run pretty much everything I'd want to run.

When I'm talking about compatibility and behavior I'm referring to how the CPU appears to the software, not how the CPU operates architecturally.

[Chucky]

yeah and if I would put me into do that project. I would do it AHI..

(so there we comes to my point of vamp where they fuck around with chipset registers adding a risc of incompability) but THAT is anohter thread...