@ppcamiga1

Vampire did deliver promised, no matter how much I dislike fuhrer stylee.

Unlike OS4.

So you should really be Vamp preacher

_________________
OS 3.x AROS and MOS supporter, fi di good, nothing fi di unprofessionalism. Learn it harder way!
SinclairQL and WII U lover :D
YT http://www.youtube.com/user/rasvoja

@tlosm

Thats good real Amiga! Mine is QL with ICE module and mouse and mem expansion!

_________________
OS 3.x AROS and MOS supporter, fi di good, nothing fi di unprofessionalism. Learn it harder way!
SinclairQL and WII U lover :D
YT http://www.youtube.com/user/rasvoja

@agami

Aha agami speaks SRPSKOHRVATSKI ili JUZNOSLOVENSKI, good
Velki pozdrav za Majstorovica iz Bgda!

_________________
OS 3.x AROS and MOS supporter, fi di good, nothing fi di unprofessionalism. Learn it harder way!
SinclairQL and WII U lover :D
YT http://www.youtube.com/user/rasvoja

@agami

Whats funny, on x1000 only way to reach compatibility is UAE JIT x1000 and RunInUAE.
SAM460 and likely A1222 can only bear to emulate OCS/ECS

_________________
OS 3.x AROS and MOS supporter, fi di good, nothing fi di unprofessionalism. Learn it harder way!
SinclairQL and WII U lover :D
YT http://www.youtube.com/user/rasvoja

agami Quote:

The Apollo V4 SAGA+Maggie is as powerful as a PS1, and a little bit more.

Probably but difficult to compare. The AC68080@~100MHz CPU is much more powerful than the PS1 R3000@~33Hz CPU but there is not much FPGA space for the V4 3D. The PS1 used most of its transistor budget for the 3D hardware. Wiki states that one PS1 chip used 1,000,000 logic gates which was more than Sony could fab at the time. Assuming a rough 4 transistors/gate, that would be 4,000,000 transistors for one chip. The R3000 CPU core is tiny and likely smaller than 500,000 transistors with caches. The AC68080 was influenced by the 68060 design using 2,530,000 transistors but likely has larger caches. I believe the PS1 uses fixed point integer fixed pipeline 3D rendering where the V4 3D is likely using floating point for better quality rendering and a more flexible and parallel rendering pipeline to support newer 3D features so the V4 3D is using more transistors for quality 3D. The PS1 invested millions of transistors into fast but low quality 3D rendering which I'm not so sure the V4 3D could keep up with with whatever transistor budget was left in the FPGA after beefing up the CPU and enhancing the chipset. The V4 3D no doubt wins on quality and features while Gunnar may say AMMX can do the 3D heavy lifting which is likely partially true, 68k+AMMX assembly required. The V4 3D may actually be easier to program than the PS1 but the cramped FPGA no doubt results in even semi-modern 3D features being left out and a lack of parallel 3D performance. Certainly the V4 3D features and performance are not currently translating to easily compiled programs using Warp3D but maybe the AMMX optimized V4 3D Warp3D driver/library is not done yet. The 68k Warp3D that is unmaintained, older than PPC, buggy and not open source could be blamed but A-Eon/Trevor has priorities like Warp 3D Nova and all the software being developed for it with SMP around the corner to support 2005+ software.

agami Quote:

It would've been hard for Commodore/AT to make a PowerAmiga in the '90s with the spec of even the SAM440. PowerPC in the late '90s is between 200MHz and 450MHz. Emu68 on a PiStorm with Raspberry Pi 4+ easily beats that.

Steve Jobs was unhappy when he announced a Mac PPC G3/G4@500MHz and then had to backtrack the announcement, perhaps resulting in the switch to x86 a few years later. This doesn't mean PPC was not capable of higher clock speeds in the late 1990s. The Alpha 21264A released in 1999 reached 833MHz with a 7-stage pipeline and 250nm process. At a higher clock speed, there is less time to for the electricity to travel through the logic circuit of the CPU core. Pipelining divides up the logic into stages with each stage having less logic and a shorter distance for the electricity to travel in one cycle. A smaller chip fab process reduces the physical distance the electricity has to travel.
A deep pipeline and smaller chip fab process are well understood ways for architects to increase clock speeds although there are other ways.

year | CPU | pipeline | process
1993 PPC601 4-stage 600nm
1993 PentiumP5 5-stage 800nm
1994 68060 8-stage 500nm
1994 PPC603 4-stage 500nm
1994 PPC604 6-stage 500nm
1995 PPC603e 4-stage 350nm
1995 PentiumP6 14-stage 500nm (OoO high end)
1996 PPC604e 6-stage 350nm (clock speed lower than PPC604 limited by large 32kiB-I+32kiB-D?)
1997 PentiumP55C 6-stage 350nm (in-order low end)
1997 PPC-X704 6-stage 500nm (Exponential Technology, bipolar/dynamic logic, ~500 MHz)
1997 PPC750 4-stage 260nm (G3, based on PPC603)
1998 PPC755 4-stage 220nm (G3)
1999 PPC7400 4-stage 200nm (G4)
2001 PPC7450 7-stage 180nm (G4)
2002 PPC7455 7-stage 180nm (G4, reached 1 GHz, Amiga1 XE G4)
2002 PPC970 16-stage 130nm (G5, reached 2 GHz)

PPC suffered from shallow pipelines limiting clock speeds and they tried to compensate by using more expensive fab processes. The PPC604 was an exception but the PPC604e appears to have hit a clock increase wall with the large 32kiB-I+32kiB-D caches as it did not clock as high as the PPC604 with half the caches. A die shrink of the PPC604e likely could have solved the problem but the combination of the large core and caches using an expensive fab process may have cause the PPC603e design to be beefed up into the G3 design instead. The PPC603e cache was doubled to the same as the 604e, a 2nd integer unit was added which makes scheduling much easier, dynamic branch prediction was added because the PPC603 static prediction mispredicted 25%-35% of the time and an external L2 cache was added but the 4 stage pipeline was kept. The success of the G3 led to the cancellation of deeper 6-stage 604e design development which already had most of the G3 much needed enhancements except for the external L2 cache support and the die shrink (using the minimal PPC603 design saved transistors and retained bus/socket compatibility of a more common CPU to upgrade a turd). The PPC7450 finally increased the pipeline to 7-stages and the PPC970 to a less practical 16-stages but not until 2002. The PPC outlier Exponential X704 had a 6 stage pipeline in 1997 that could reach about 500MHz but rumor is that Freescale and IBM convinced Steve Jobs that their designs were better and would soon reach similar frequencies. This is partially true but the X704 was using an older more affordable 500nm process for a cost advantage. A newer fab process likely could have increased clock speeds considerably and reduced the 85W max power (PPC603e was 6W). The X704@500MHz and 68060@50MHz used a similar sized process and the 68060 had a 2 stage deeper pipeline but the 68060 uses all static logic which is easier to develop and maintain instead of dynamic logic. Most CPU designs today have gone back to full static logic CMOS designs to save power too.

So why did PPC architects like the shallow pipeline designs? Shallow pipeline designs use fewer transistors for smaller cheaper cores so the same PPC cores were used for embedded and desktop use to improve economies of scale. Longer pipelines need better branch prediction where the PPC603(e) only had static branch prediction despite a high mispredict rate as mentioned above. RISC load-to-use latency/stalls often grows with a longer pipeline and the limited OoO PPC designs may not have been able to avoid load-to-use stalls. Much of the PPC design philosophy was designed around this shallow pipeline limited OoO idea and deeper pipelined PPC cores had disappointing performance, perhaps because the limited OoO could no longer hide the L1 cache load-to-use latency (the more aggressively OoO PPC970 had other issues but also had disappointing performance). Most CISC designs avoid L1 cache load-to-use stalls which provides a huge performance advantage and was later demonstrated by the simple SiFive U74 core with a CISC like in-order design that outperforms the complex superpipelined OoO PPC970 at the same clock speed on some benchmarks even though it doesn't gain as much performance as a CISC ISA. It kind of makes me wonder how much a 68060 would have to be upgraded to outperform the PPC970 per clock too. The 16-stage PPC970 would allow a higher clock speed giving an overall performance advantage but mid-depth pipelines are the most practical which even embedded CPU cores moved to like the Cortex-A53, Cortex-A55 and SiFive U74 which are all 8-stage in-order cores like the 68060. Another thing that makes me wonder is whether the 68060 would still be viable with incremental upgrades over time kind of like the P6 Pentium which is the predecessor of modern x86-64 cores. The x86 ISA lacked orthogonality for a good in-order core like the 68060 despite a couple of attempts with the early Bonnell Atom and Larrabee GPGPU projects which limits how far x86(-64) can scale down. One thing is for sure. CISC has the performance and RISC had to give up most RISC philosophies and adopt CISC advantages to compete. It's unfortunate the 68060 was locked in the basement with 16 GP registers and industry leading code density as x86 with 8 GP registers and inferior code density was killing shallow pipeline and fat PPC with 32 GP registers. PPC CPU designs looked better when caches were increased and die shrinks were applied at the same time but this was expensive and could be applied to any CPU.

agami Quote:

Good enough graphics? For what exactly? AmigaOS 4 machines don't even have all the games that were available to Windows gamers between 1995 and 2000. Let alone what was available during 2000s, when 1GHz+ PowerPC Macs were in their prime. So from a gaming aspect, AmigaOS 4 and MorphOS PPC gaming is crap compared to PC from late '90s. So it's plenty worse than cheap PC made in '90s.

The x86 PC became a gaming platform starting in the early to mid 1990s that surpassed and killed the 68k Amiga gaming platform with thousands of games. Even the too little too late 68EC020+AGA Amiga spec in 2 years (1992-1994) has far more games than PPC AmigaOS 4 and MorphOS combined in 22 years (2002-2024).

Last edited by matthey on 09-Nov-2024 at 02:55 PM.
Last edited by matthey on 09-Nov-2024 at 10:09 AM.
Last edited by matthey on 09-Nov-2024 at 09:49 AM.
Last edited by matthey on 09-Nov-2024 at 09:12 AM.

@Karlos

Who is ppcamiga1? This guy's girlfriend's dad!

Last edited by jorkany on 09-Nov-2024 at 12:16 AM.

_________________
Here for the whimpering end

@jorkany

Lol. The thing is, be does t really criticise x86, he instead trashes 68K.

_________________
Doing stupid things for fun...

@Karlos

Well x86 is PowerPC to PPCAmiga1

Bet he doesnt have Intel Outside backdrop

Last edited by vox on 09-Nov-2024 at 12:15 PM.

_________________
OS 3.x AROS and MOS supporter, fi di good, nothing fi di unprofessionalism. Learn it harder way!
SinclairQL and WII U lover :D
YT http://www.youtube.com/user/rasvoja

@vox

I use LCD monitors
I play old games on uae
On winuae I may setup display filters for any game
so it looks almost like on 1084s
even today I try to play my beloved Agony on my vampire but it looks ugly
I turn it off it was not worth playing

@vox

no vampire is still to slow in graphcis
it is still 1/3 performance of PS 1
(without clipping etc)

still nothing that Commodore made if not bankrupt in 1994

amiga ppc reach at least this level

Karlos Quote:

Lol. The thing is, be does t really criticise x86, he instead trashes 68K.

I don't know any other so called Amiga fans who hate the 68k more than ppcamiga1 and Trevor. Neither of them have obviously ever programmed a 68k Amiga and a PPC AmigaNOne or understand anything technical about CPUs or the Amiga for that matter. Most of the remaining Amiga fans love or at least respect the 68k, the compatibility it provides, the small footprint with tiny fast loading programs which allowed the low end Amiga market, the ease of programming and debugging at a low level, the large flat address space which allows a 1985 68k Amiga to be expanded to 2GiB of memory, etc. Surely two Amiga users this ignorant would be an anomaly beyond comprehension in what is left of the Amiga market after Trevor has devastated it. This is your best argument yet that they are the same person but it doesn't rule out that Trevor, aka ppcamiga1, is secretly Bill Gates sabotaging the 68k Amiga.

Last edited by matthey on 09-Nov-2024 at 07:52 PM.

@matthey

I don't think Trevor is a 68K hater. He has 68K machines in his collection AFAIK and in order to be such a proponent of the "next gen" platform he must've been a fan of the original in the first place.

_________________
Doing stupid things for fun...

Karlos Quote:

I don't think Trevor is a 68K hater. He has 68K machines in his collection AFAIK and in order to be such a proponent of the "next gen" platform he must've been a fan of the original in the first place.

But can Trevor explain what made the Amiga special besides a big expensive desktop box using AmigaOS with lots of slots and a graphics card?

Last edited by matthey on 09-Nov-2024 at 10:36 PM.

@vox

Quote:

vox wrote: Aha agami speaks SRPSKOHRVATSKI ili JUZNOSLOVENSKI, good Velki pozdrav za Majstorovica iz Bgda!

When I was in school during the '80s in SFRJ it was Srpsko-Hrvatski, but given the layout of the Balkans today, I like this term Južnoslovenski.

Unfortunately for us, our friend Majsta is out of the hardware game and is focusing on community projects. I will pass on your pozdrav next time I correspond with him.

_________________
All the way, with 68k

@ppcamiga1

Stop spreading lies about 68k+
Apollo 080+SAGA+AMMX+Maggie 3D is overall better than PS1
Emu68 on Raspberry Pi 5 is just as fast as X1000

Stop trolling, and start working on a low-cost 3GHz+ PPC AmigaOS 4 compatible motherboard without drawbacks.

_________________
All the way, with 68k

@ppcamiga1

Quote:

amiga ppc reach at least this level

what a joke, which gfx board was made for amiga ppc? It uses off the shelf gfx boards, if you have a 68k with raspberry 4 emu68 levels of speed accessing the very same off the shelf gfx boards what difference you think you would find?

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@agami

Quote:

Emu68 on Raspberry Pi 5 is just as fast as X1000

Raspberry pi 5 be used on pistorm due to latency, and afaik there's no draco like system being used on standalone. How do you get those values? I would love to see it though, and running games on instances of uae.

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@agami

Quote:

Emu68 on Raspberry Pi 5

Do tell more.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@pixie

Quote:

pixie wrote: @ppcamiga1 Quote: amiga ppc reach at least this level what a joke, which gfx board was made for amiga ppc? It uses off the shelf gfx boards, if you have a 68k with raspberry 4 emu68 levels of speed accessing the very same off the shelf gfx boards what difference you think you would find?

The BlizzardVision. That's literally the only graphics card ever created specifically for an Amiga equipped with a PPC as far as I can tell. Even the CVision worked with the 68K only CSMK3.

Nothing about it actually required PPC, of course, but it was the only combination it fit.

_________________
Doing stupid things for fun...

@agami

stop spreading lies about 68k+
Apollo Maggie 3D is two and half times slower than PS1

stop spreading lies about emu68
it is not 68k it is emulator
and this emulator is still slower than many including sam580

remeber that other people have rpi and vampire
and may easily check your lies

so agami dont lie !!!