Hi dear Amigans,

I know here is a low priority and interesting to few, but took endavour to update Vamp FAQ
with deadline by team.

Its WIP, mostly my answers, and collected info from forums.

DOC version, free to comment, request to edit
https://drive.google.com/open?id=1Gs60MmNrgDUuq-QHHwdtQzwSB9y3y9EE

ODT
https://drive.google.com/open?id=1aYwunmTwB1GzETtQdxJZmUus-foTqU3R

PDF
https://drive.google.com/open?id=1ZVfghiRdd0bGFqCKC9hvpt2cMg8C9vFp

Needs:

- Proofreading by native English speakers and sentence rework :)
- Looking for more questions and improved answers
- More updates to links as resources, as I would love to show the world what is left
of OS3 AROS arena

- Suggestions?

Mind that its WIP and NOT official until given to team for final edit and publishing (early Feb 2020)

Thank you, all contributors will be signed.

[ Edited by vox 06.01.2020 - 09:14 ]

vampire cpu is not evolution of original 68k, it was made from scratch.
68060 with fast RAM is still faster than vampire cpu.
vampire cpu is not 100% compatible with 68k.
mmu tools stil are not public available.

> vampire cpu is not evolution of original 68k, it was made from scratch.

...as was the 68060, for instance.

> 68060 with fast RAM is still faster than vampire cpu.

Benchmarks and user reports indicate otherwise.

I have Vampire2 in my A500 but seeing how little software were done for it i think that Mediator+PPC on PCI card is better option for big Amigas. Everything may change when Warp060 accelerators will be ready.

This is a MorphOS forum. While off-topic discussions have traditionally been tolerated, this stuff is really getting out of hand.

If this is now being used as a platform for unpaid Vampire ads, which cannot run MorphOS whatsoever, I think some changes might be in order.

Quote:

Your mental issues have been adressed in last part of FAQ
under "pancakes".

Please read them first before sounding like broken record.

FAQ was made both for customers, potential customers and mumbo-jumbo makers
hot air sellers.

I wonder how would AmigaOne (e.g. my PA Semi) pass the same test

- Not evolution of m68l defenetely
- Slower RAM and mem controller then V4
- Not compatibile at all, Petunia does not emulate chipset
- No MMU tools publicly avail

Quote:

I have made a clear FAQ aid request. There is no intention to turn it to war.

No more public posts, those who want to help can PM me or e-mail me jahdeliverus@gmail.com

Hope that is acceptable, please close the thread, without deletions.

P.S.
MorphOS is mentioned in FAQ

Quote:


There is a FAQ setion that has 080-MMX m68k native s/w.

Au contraire, I have seen more m68k production for 060 RTG level since Vamp appeared.
And I dont see WarpUp, PowerUp or even OS4 library growing.

[ Edited by vox 06.01.2020 - 12:24 ]

Quote:

Prove it.

68060 was made by Motorola by the same team that made former Motorola 68k cpu.

> 68060 with fast RAM is still faster than vampire cpu.

Quote:

Apollo/Natami/Vampire Team benchmarks?

On my new better Amiga ppc works everything what I used on my 1200. So it is compatible.
As I wrote on pm on aw vox your problems was your lack of experience with good old ModePro.
gdb is also available for years.


[ Edited by ppcamiga1 06.01.2020 - 11:47 ]

Quote:

And you used only WB RTG games I pressume?

No, screen promotion does not always help. I tried every advice I could get on
Amiga Intution, but m68k compatibility has not increased. Contrary it has decreased
from OS 4.0 Classic to OS 4.1 FE u1 Enhancer 1.3

Simply, AEON or Hype did nothing to increase Classic compatibility,
community did OS4 JIT. And its one of few PA Semi Altivec optimized pieces of code.
In the beginning non JIT OS4 UAE could run just A500 titles.

In LW, for example, you cant see preview since its chipset based.

I have prepaid OS 4.2 license and paid x1000 dearly to use Libre office and Timberwolf, promised at x1000 launch.

x1000 came with copied CD (no original), in case generic box with b/w photocopied manual.
In long time I have not seen such un-pro package. AmigaKit did not even know which mem modules
were included, had to open to see that. And so on ...

Also, x1000 AEON and AmigaKit break EU customer laws by providing only 1 year return to base warranty, which is lower then law requires.

What you forget is that FPGA is not carved in stone: situation improves as we speak,
while badly designed and supported boards, remain so.

In the end, market will say it word, as well as community.

End of all comms, phasers down, warp back to FAQ :)

[ Edited by vox 06.01.2020 - 13:35 ]

dear vox on Amiga Os 4.x you can use this patch:

http://os4depot.net/index.php?function=showfile&file=graphics/raytrace/lightwave_rtg_patch.lha


[ Edited by ppcamiga1 06.01.2020 - 12:46 ]

>>> vampire cpu is not evolution of original 68k, it was made from scratch.

>> ...as was the 68060, for instance.

> Prove it.

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=7675&start=702

>>> 68060 with fast RAM is still faster than vampire cpu.

>> Benchmarks and user reports indicate otherwise.

> Apollo/Natami/Vampire Team benchmarks?

No, user benchmarks.

Quote:

Sysinfo on x15 v4 core
http://www.apollo-core.com/gfx/V4-SysInfo-SA_6619df_x15r.png

PC Task x86 emu
http://www.apollo-core.com/gfx/x86power.gif

P96 speed
http://www.apollo-core.com/knowledge.php?b=2¬e=14103

Mem speed (I benchmarked x1000 against)

VAMPIRE-2 reaches with LONG memory access now
530 MB/sec READ
660 MB/sec WRITE speed.

Using AMIGA program BUSTEST
Download it from AMINET here: EXTERNAL LINK
AFAIK VAMPIRE outclasses here every other 68K and PPC AMIGA system,
including latest 2 GHz X5000 and 2.7 GHZ G5 MACs.







AMIGA BUSTEST - READL Score
------------------------------------------

VAMPIRE 2 529.5 MB/sec (world record!)

x1000 AMIGA-1 PA6T 1800 MHz 285.6 MB/sec
MOS IBook PPC G4 1400 MHz 185.4 MB/sec
MOS MacMini PPC G4 1500 MHz 184.3 MB/sec
OS4 SAM PPC 460 1100 MHz 109.7 MB/sec
OS4 SAM PPC 440 667 MHz 91.1 MB/sec
OS4 Peg2 PPC G4 1000 MHz 80.6 MB/sec

CyberStormPPC 68060@50 55.1 MB/sec
Apollo 1240 68040@40 47.8 MB/sec
FPGA ARCADE 29.0 MB/sec
Blizzard 1230 68030@50 24.3 MB/sec
Typhoon MK2 68030@40 18.8 MB/sec
AMIGA 4000 68040@25 11.5 MB/sec
AMIGA 1200 ACA 1220 10.9 MB/sec
Zeus 68000 50 Mhz 9.1 MB/sec


Thanks for reminding me to add performance section PPC Amiga :)

> I have Vampire2 in my A500 but seeing how little software
> were done for it i think that Mediator+PPC on PCI card is
> better option for big Amigas.

I think there is more m68k software using AMMX than WarpOS/WarpUP software using AltiVec ;-)

> Everything may change when Warp060 accelerators will be ready.

How so?

> I wonder how would [...] my PA Semi [...] pass the same test
> [...]
> - Slower RAM and mem controller then V4

PA6T on X1000 has actual memory performance of 2.7...4.0 GB/s (depending on chunk size and access type) according to RAGEMEM benchmark. What's the actual memory performance on Vampire V4? You quoted ≈0.5...≈0.7 GB/s for Vampire V2 in comment #13, so less than a fifth of PA6T performance. Is Vampire V4 really that faster?
(Btw, your quoted ≈0.3 GB/s for PA6T is measured with an m68k program, so that's the memory performance of the m68k emulation running on PA6T, not that of the PA6T itself.)

> I dont see [...] OS4 library growing.

A look at os4depot.net indicates otherwise.

> Hype did nothing to increase Classic compatibility, community did OS4 JIT.

Petunia has been part of OS4 since AmigaOS 4.0 Prerelease Update #4 in 2006.

> And its one of few PA Semi Altivec optimized pieces of code.

At least with regard to Petunia, I don't think this claim makes any sense.

> I [...] paid x1000 dearly to use Libre office and Timberwolf,
> promised at x1000 launch.

Timberwolf you got, and LibreOffice was first announced by A-Eon 9 months after commercial X1000 launch.

Quote:

You do not prove that 68060 was made from scratch by different people than original 68k team that know every detail of 68k.
What You wrote is that there was optimisation of part of 68060, this is very different from made everything from scratch by different people.

Quote:

Do they compare vampire with 68060 with 100 MHz RAM or only poor Amiga cards with 15 MHz RAM?

> Using AMIGA program BUSTEST [...]
> AFAIK VAMPIRE outclasses here every [...] PPC AMIGA system,
> including latest 2 GHz X5000 and 2.7 GHZ G5 MACs.
>
> AMIGA BUSTEST - READL Score
> ------------------------------------------
> [...]
> x1000 AMIGA-1 PA6T 1800 MHz 285.6 MB/sec
> MOS IBook PPC G4 1400 MHz 185.4 MB/sec
> MOS MacMini PPC G4 1500 MHz 184.3 MB/sec
> OS4 SAM PPC 460 1100 MHz 109.7 MB/sec
> OS4 SAM PPC 440 667 MHz 91.1 MB/sec
> OS4 Peg2 PPC G4 1000 MHz 80.6 MB/sec

Please keep in mind that these are m68k emulation results, not PPC-native results.

>>>>> vampire cpu is not evolution of original 68k, it was made from scratch.

>>>> ...as was the 68060, for instance.

>>> Prove it.

>> https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=7675&start=702

> You do not prove that 68060 was made from scratch by different people
> than original 68k team

I didn't claim it was made by different people. There likely were some overlappings in the 68040 and 68060 design teams. One and the same people can very well do different from-scratch implementations of the same specification.

> What You wrote is that there was optimisation of part of 68060

You do not seem to understand the difference between a microcoded microarchitecture and a hardwired microarchitecture. This is not really about optimization but about a different approach to microarchitectural design.

>> No, user benchmarks.

> Do they compare vampire with 68060 with 100 MHz RAM
> or only poor Amiga cards with 15 MHz RAM?

They compare with what's available on different 68060 cards.

Quote:

So You have no proof that 68k team at Motorola made everything from scratch instead of use parts of older design to save time and money.

Quote:

This is your opinion based on nothing. You exaggerate change of part of cpu.
It is just change of part of cpu nothing more.

vampire has not any single bit of original 68k cpu. vampire do not share history with 68k.
vampire was made from scratch by different people than original 68k team at Motorola.

> So You have no proof that 68k team at Motorola made everything from
> scratch instead of use parts of older design to save time and money.
> This is your opinion based on nothing. You exaggerate change of part
> of cpu. It is just change of part of cpu nothing more.

Please educate yourself to see that the difference is on the gate level, i.e. on the actual microarchitectural level:
https://en.wikipedia.org/wiki/Control_unit#Hardwired_control_unit
https://en.wikipedia.org/wiki/Microcode#Justification

The claim was on Wikipedia from 2004 to 2009 and deletion was justified not by incorrectness but by unsourcedness (as per Wikipedia standards), also deleting "unsourced" facts like "a second integer pipeline, a two cycle integer multiplication unit, a faster FPU, and branch prediction logic" in the process:

https://en.wikipedia.org/w/index.php?title=Motorola_68060&diff=4592540&oldid=4462952
https://en.wikipedia.org/w/index.php?title=Motorola_68060&diff=291144973&oldid=291144208

> vampire has not any single bit of original 68k cpu.

Its CPU implements the m68k ISA, or rather a union of the different m68k ISA incarnations (AFAIK minus some rarely used instructions not even supported by all m68k CPUs).

> vampire do not share history with 68k.

It's just an ISA, nothing more. Anything else is religion.

Quote:

True, but either compile same test to OS4 (and you will get similar results, since its mem test not CPU test) or provide at least 040-FPU binary of ragemem.

I am sorry, Petunia is very fast JIT and does not have a lot of CPU overhead. That part of OS4 is nicely done, like its MOS counterpart.

Memory controller depends on paired memory and V4 has increased results due to DDR3 RAM, as well as Igors V1200.

Quote:

No, I have been daily fun at it since its creation to 2018.
and in my exp, its been down since at least 2016.

Its again 3rd party works and thanks to effort of few dedicated
contractors and PPC developers (Huno, Daytona ...). But it is not real
OS team work.

That way lookinet, Aminet is flourishing equals OS3 is strong.

AmigaOS 4 free updates and baselines are rare nowadays. MOS is greater there.

Nuff talk, I have deadline to finish FAQ

If you have any ideas for as-is FAQ, let me know.

Above performance will be included.

> Hype did nothing to increase Classic compatibility, community did OS4 JIT.

Petunia has been part of OS4 since AmigaOS 4.0 Prerelease Update #4 in 2006.

I am speaking of UAE OS4 since it is real Amiga compatibility layer in OS4, not Petunia.

>vampire was made from scratch by different people than original 68k team at Motorola.

True, but so were AEON boards, Phase 5 boards, Freescale CPUs ...



[ Edited by vox 06.01.2020 - 18:05 ]

>> your quoted ≈0.3 GB/s for PA6T is measured with an m68k program,
>> so that's the memory performance of the m68k emulation running
>> on PA6T, not that of the PA6T itself.)

> compile same test to OS4

The bustest archive does not come with source. You may try to run the stream benchmark on your Vampire, which is available natively for m68k, MorphOS and OS4. Stream results for PPC boards running MorphOS or OS4 are widely available.

> and you will get similar results, since its mem test not CPU test

No, memory read/write performance very much scales with CPU speed/clock if done by the CPU (which is the case with both bustest, RAGEMEM and stream).

>> You quoted ≈0.5...≈0.7 GB/s for Vampire V2 in comment #13, so less
>> than a fifth of PA6T performance. Is Vampire V4 really that faster?

> Memory controller depends on paired memory and V4 has increased
> results due to DDR3 RAM, as well as Igors V1200.

I still doubt that Vampire V4 memory performance sextuples that of Vampire V2.