@Hammer

Quote:

PC's MPC Level 2 (486SX-25, 4MB RAM) PC's MPC Level 3 (Pentium 75, 8MB RAM). https://en.wikipedia.org/wiki/Multimedia_PC

I really like this comparison ! I've been upgrading PCs since 386DX40 (had multiple 486s) till I got Pentium 100 MHz, so I have a very good reference point for each of those machines and what kind of games can they pull off.

Quote:

This topic covers PC's MPC Level 2 and MPC Level 3 equivalent for the Amiga.

That's the question then. So, what is the Amiga equivalent of MPC 2 and MPC3 ?

I'd hazard a guess that Vampire V4 is directly comparable in raw CPU power to Pentium (but at 60 or 75 or 90 or 100 MHz?), it also has an AMMX instruction set that on paper looks very competitive to MMX.

I'm sure the V4 has better RAM bandwidth than Pentium (due to the nature of the FPGA).

Only very few games on Pentium ran in 640x480x256 in a remotely playable framerate. Probably, only MDK 1. Hell, plenty racing games stuttered even at 320x200 (Screamer 2, NFS 1, F1GP) despite the machine having a really fast and expensive 2D card.

Would a PiStorm trump Pentium 200 MHz in raw CPU powah?

@CosmosUnivers

that is nonsense. That might be true for you and some others but not for everyone in the community and for sure not for everyone in the world

@Heimdall

Quote:

I really like this comparison ! I've been upgrading PCs since 386DX40 (had multiple 486s) till I got Pentium 100 MHz, so I have a very good reference point for each of those machines and what kind of games can they pull off.

https://www.youtube.com/watch?v=1B1jKjrRUmk
For Doom baseline, Amiga 1200 AGA with 68030 @ 50Mhz and PC 386DX 40Mhz + ET4000 ISA is about even.

Faster 68K CPU can yield better results.

Quote:

I'd hazard a guess that Vampire V4 is directly comparable in raw CPU power to Pentium (but at 60 or 75 or 90 or 100 MHz?), it also has an AMMX instruction set that on paper looks very competitive to MMX.

MPC Level 3 has Pentium 75 minimum which includes pretty strong FPU performance.

https://www.youtube.com/watch?v=0_dW-21gdkw
https://amitopia.com/first-quake-fps-run-on-warp-1260/
Warp1260 with 68060 rev6 @ 100 Mhz with RTG yields about Pentium 75.

Quake on 68060 has a similar effect as Cyrix 6x86.

https://thandor.net/benchmark/33
Quake demo 3 benchmark.
AMD K5 PR166 @ 116 Mhz = 24.40 fps (66 Mhz fsb)
Pentium 90 Mhz = 24.30 fps (60 Mhz fsb)
AMD K5 PR133 @ 100 Mhz = 22.90 fps (66 Mhz fsb)
Cyrix 6x86 PR166 @ 133 Mhz = 21.40 fps (66 Mhz fsb)
Pentium @ 75 Mhz = 20 fps (50 Mhz fsb)
AMD K5 PR90 @ 90 Mhz = 18.50 fps (60 Mhz fsb)

AMD and Cyrix PR refer to integer performance relative to Pentium.

Cyrix 6x86''s FPU being 3rd best has pushed Cyrix into budget PC builds and its repulation was stained.

Warp1260 with 68060 rev6 @ 100 Mhz = 18.64 fps via demo2

https://www.youtube.com/watch?v=jPAo56J66sw
Warp1260 with 68060 rev 6 with 105 Mhz

Pentium @ 75 Mhz with 50Mhz fsb can be overclocked via fsb jumper e.g. 60 Mhz fsb yields 90 Mhz. Pentium 75, 90 and 100 are the same stepping.

I overclocked my 1996 era Pentium 150 Mhz to 166 Mhz with just fsb 60 Mhz to 66 Mhz jumper.

Pentium 75's 64bit 50 Mhz fsb is equivalent to 68060 @ 100Mhz's 32bit 100Mhz external bus.

I don't have information on AC68080 V4's Quake results.

https://www.youtube.com/watch?v=FUgGDfYVSjk&t=4s
From 4 years ago, Quake demo1 reached 30.70 fps from AC68080 V2 + RTG with Amiga 2000

That's between Pentium 120 and Pentium 133. Beats Cyrix 6x86MX PR233 @ 187 Mhz's 29.50 fps

From Quake results, AC68080 V2 @ 85 Mhz + RTG accelerated Amiga exceeds 1996 MPC Level 3's Pentium 75 minimum.

I would rate AC68080 V2 @ 85 Mhz Quake Pentium Rated (PR) at 120. AMMX SIMD instructions are not factored in.

Unlike Phase 5 / GVP who focused on minority A1200/A2000/A3000/A4000, both PiStorm/Emu68 and AC68080 based accelerators can upgrade the A500 majority.

Quote:

Would a PiStorm trump Pentium 200 MHz in raw CPU powah?

For PiStorm32/stock RPi 4B/Emu68, Quake demo 1 RTG results are in the Pentium III 733 Mhz range, i.e. blitz pass Celeron 300A's 67.80 fps into 120 fps range.

Quake demo 1 with C= AGA render device is 60 fps with AmiQuake 320x200 demo1, hence C= AGA is a bottleneck.

For RTG, overclocking PRi 4B / CM4 to 2.2 Ghz to 3 Ghz yields higher results.

I have overclocked my PiStorm32/RPi CM4/Emu68 to 2.0 Ghz without voltage increase and the warranty status is still valid.

My PiStorm has RPi 4B for A500 rev6 and PiStorm32 has RPi CM4 for A1200.

Last edited by Hammer on 27-Feb-2025 at 12:44 PM.
Last edited by Hammer on 27-Feb-2025 at 12:37 PM.
Last edited by Hammer on 27-Feb-2025 at 12:21 PM.
Last edited by Hammer on 27-Feb-2025 at 12:15 PM.
Last edited by Hammer on 27-Feb-2025 at 12:12 PM.
Last edited by Hammer on 27-Feb-2025 at 12:02 PM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Heimdall

Quote:

But, Atari is stuck in the past. Amiga is the only retro platform that is marching forward (PPC, PiStorm, Vampire, A600GS, ...)

The Amiga camp still has the MAGA mindset i.e. Make Amiga Great Again.

Historically, Amiga 500's 68000 CPU selection is contemporary with mainstream game consoles like Sega Mega Drive/Genesis' 68000 CPU.

PiStorm with RPi 4B / CM4's ARM Cortex A72 selection is contemporary with current mainstream game consoles like Nintendo Switch's ARM Cortex A57.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Hammer

MAGA is not a bad idea

Perhaps there is a chance to grow to more than just pure retro community, who knows

at least it is fun and keeps developers busy

@Hammer

Thank you so much! That's a lot of great data you gathered there! I'll summarize it into a simple table format (for my own future reference):

Obligatory disclaimer - I'm comparing tractors and oranges here (a SW rasterizer, which - it just so happens- is exactly what I need), but indulge me, pls

---------------------------------------------------------------------------
......PC Model.............Amiga............CPU
---------------------------------------------------------------------------
PC 386DX 40Mhz.....A1200........68030 @ 50Mhz
Pentium 75 MHz.......Warp1260..68060 rev6 @ 100 Mhz
Pentium 120 MHz.....V2............68080 @ 85 Mhz
Pentium 150 MHz.....V4SA..........68080 Latest core *
Pentium III 733 Mhz..PiStorm32/stock RPi 4B/Emu68
---------------------------------------------------------------------------

The fourth entry for V4SA is just a guess on my part, because in last 3 yrs there have been at least 3 official overclocks (via new core) that I happen to remember (likely more). Plus, the MIPS rating was increased on top of clock increase. So, that'd put it somewhere in the range of Pentium 133 MHz - 150 MHz.
That range is precise enough for me to establish where exactly does the Vampire stand compared to PC.

On the other forum, I was attempting to get at least some sort of real-world benchmark from the PiStorm, but this is the first time I have some idea (other than the SysInfo) - about how fast PiStorm actually is (for a SW rasterizer).

@Heimdall

of course you only compare pure processing power here

it also depends on the software, how good it make use of the specific amiga hardware

The apollo team showed a game from around 2004, that needed at least 1 Ghz on PC and worked good on apollo hardware. But of course only after lots of optimizations and adaptions to the hardware

@Heimdall

Quote:

The fourth entry for V4SA is just a guess on my part, because in last 3 yrs there have been at least 3 official overclocks (via new core) that I happen to remember (likely more). Plus, the MIPS rating was increased on top of clock increase. So, that'd put it somewhere in the range of Pentium 133 MHz - 150 MHz. That range is precise enough for me to establish where exactly does the Vampire stand compared to PC.

Ice Drake, Firebird, Manticore, Salamander and Phoenix add-on accelerator SKUs are also AC68080 V4, not just the V4SA. AC68080 V4's clock speed is 92 or 100 MHz. I don't consider official clock speed as overclock since they are OEM's warranty clock speeds.

15% increase on clock speed from 85 Mhz to 100 MHz, hence about Pentium 140 level, hence I agree with your estimation.

My Pentium Rating is based on Quake results that are heavy on the FPU and may differ with fixed point 3D engines.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@OlafS25

https://www.giantbomb.com/robin-hood-the-legend-of-sherwood/3030-6594/
System Requirements
Windows 98/ME/2000/XP
Pentium II 233 MHz
64 MB RAM
900 MB Hard Disk Space
Video Card with 4MB RAM (3D acceleration not required)
DirectX 8.1
4x CD-ROM


https://www.ign.com/articles/2002/11/21/robin-hood-the-legend-of-sherwood
A game review with a posted date of Nov 21, 2002

While the box recommends a minimum spec of a PII 233MHz machine, I would say you'd want at least a 800MHz machine to run the game in 1024x768. It's not a particularly graphic intensive game, but it does chug a bit when you're scrolling on slower systems in higher resolutions, and 1024x768 is indispensable since it gives you a wider view of the particular scene


I have the Y2002 ISO and Windows XP on an IBM T20 laptop, Pentium III 700Mhz speed step down and I plan to double check.

https://www.youtube.com/watch?v=MSCFxmratJg
Vampire version has 848x480 resolution with 16bit color.

I do look at the small details.

Last edited by Hammer on 27-Feb-2025 at 04:21 PM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Hammer

Gunnar wrote it needs 1000 Mhz to run smooth like it now does on the V4 edition so I cannot say how useable 500 Mhz would be

@OlafS25

Vampire version has a resolution of 848x480 with 16-bit color.

What's the PC version running at again? Hint: it's higher resolution.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Hammer

hint hint

I am the wrong person to debate about the differences between the V4 hood version and the pc version and the requirements, I do ont own the game and such old hardware to run it on and I only know the port from the published videos

What I only wrote if comparing amiga with other platforms you cannot simply compare raw processing power because you get much better results when doing something specific

if it is worth the invested time regarding financial income is another question

Last edited by OlafS25 on 27-Feb-2025 at 04:32 PM.

Heimdall Quote:

Thank you so much! That's a lot of great data you gathered there! I'll summarize it into a simple table format (for my own future reference): Obligatory disclaimer - I'm comparing tractors and oranges here (a SW rasterizer, which - it just so happens- is exactly what I need), but indulge me, pls --------------------------------------------------------------------------- ......PC Model.............Amiga............CPU --------------------------------------------------------------------------- PC 386DX 40Mhz.....A1200........68030 @ 50Mhz Pentium 75 MHz.......Warp1260..68060 rev6 @ 100 Mhz Pentium 120 MHz.....V2............68080 @ 85 Mhz Pentium 150 MHz.....V4SA..........68080 Latest core * Pentium III 733 Mhz..PiStorm32/stock RPi 4B/Emu68 --------------------------------------------------------------------------- The fourth entry for V4SA is just a guess on my part, because in last 3 yrs there have been at least 3 official overclocks (via new core) that I happen to remember (likely more). Plus, the MIPS rating was increased on top of clock increase. So, that'd put it somewhere in the range of Pentium 133 MHz - 150 MHz. That range is precise enough for me to establish where exactly does the Vampire stand compared to PC. On the other forum, I was attempting to get at least some sort of real-world benchmark from the PiStorm, but this is the first time I have some idea (other than the SysInfo) - about how fast PiStorm actually is (for a SW rasterizer).

The 68030 has better performance than the 386 at the same clock speed and using the same memory. The 68030 has 256B I+D with cache burst and the 386 has no cache. The 68040 has better performance than the 486 at the same clock speed too although most 486s out clocks 68040s. The 68060 has better integer performance than the Pentium much like the Cyrix 6x86. The ByteMark benchmark showed 40% better integer performance for the 68060 at the same clock speed with one compiler and nearly on par FPU performance with another compiler. The Pentium does have the advantage with a pipelined FPU but the ugly stacked based x86 FPU ISA handicaps performance for compilers. The Quake results are pretty much the perfect benchmark for the Pentium as good FPU performance combined with data bus bandwidth are needed. Much of the data is used once, the streaming data flushes the data cache and data bus bandwidth and memory performance become more important. Newer hardware has better memory bandwidth and performance which also improves performance. Some other 3D programs effectively flush the cache too but maybe not yours with simple flat shading, smaller data sets and a fixed point integer 3D engine. The 68060 has great integer performance if well optimized for the 68060. Most compilers do not have 68060 specific instruction schedulers for the 68060 and the ByteMark benchmark used GCC 3 so the integer performance may actually be better than 40% better than the Pentium. Quake for the Pentium used extensive hand optimized Pentium specific FPU assembly code. I improved the VBCC support code for the 68060 FPU which brought it up to near parity with the Pentium in the ByteMark FPU but the VBCC mediocre at best integer 68k backend limits performance. The Pentium received great compiler support and Quake was highly optimized for it while 68060 compiler support is minimal. If you are coding in assembly language then that is not a problem as 68060 demo coders demonstrate.

Greatest Amiga Demos Ever
https://www.youtube.com/watch?v=RPdB_zdyMbM&list=PLwds84NCmJadTeGeeXBzVuKWsdwi2Y6PB

I suspect you are too high on the PiStorm/Emu68/Pi4 estimation. The OoO Pentium 3 has strong performance and the core design is the basis of modern x86-64 cores. Maybe the PiStorm reaches 68060@733MHz equivalent peak performance but I doubt it can be sustained. Memory bandwidth does wonders for Quake performance though.

@OlafS25

At the start of the game,

https://www.youtube.com/watch?v=MSCFxmratJg
V4's 2:17 of 10:09, the house on the left should have transparency smoke from chimney and missing on V4 version.

IBM T-20 laptop's S3 Savage IX 8MB AGP2X's 800x600p (480,000 pixels) and 1024 x 768p (786,432 pixels) scrolling is smooth as V4. This T-20 laptop has 512MB PC-100 RAM and PATA 40GB.

https://www.youtube.com/watch?v=hVi54w9T6bE
Year 2000 HP OmniBook XE3-GC's S3 Savage IX 8MB AGP2X and Intel Pentium III Coppermine playing games:
1) Aliens vs Predator
2) Atomic Bomberman (it's not 3D)
3) Tactical Ops
4) Quake II
5) Quake III
6) Viper Racing
7) Unreal Tournament

S3 didin't recover from the PR damage from S3 Virge. S3 Savage IX is a reasonable iGPU for its time.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@matthey

Quote:

The 68030 has better performance than the 386 at the same clock speed and using the same memory.

Both 68030 and 80386 doesn't have 1 instruction per clock (IPC). 68030's data cache advantage doesn't translate into significant IPC gains over 68020.

During 1992, Am386-40 has a lower cost advantage over 68030-25 and i386DX-25.

68030-25 Mhz wouldn't be able to fully use 32bit 140 ns (7.1 Mhz) read/write bandwidth FPM DRAM into 7 MIPS let alone the heavy MUL instruction 3D workload.

There's a reason for DSP3210 @ 50Mhz on the 32bit 140 ns (7.1 Mhz) read/write bandwidth FPM DRAM. DSP3210 can saturate 32bit 140 ns (7.1 Mhz) read/write bandwidth FPM DRAM.

Quote:

The 68030 has 256B I+D with cache burst and the 386 has no cache.

The baseline i386 has 32 entry TLB cache, 16 byte prefetch unit, and instruction queue holds three decoded instructions.

386's instruction decode unit breaks apart X86 instruction into its component parts and generates a pointer to the microcode that implements the external instruction.

My 386DX-33 has 64 KB external cache on the motherboard.

For socketed 386 motherboard has options from Cyrix 486DLC which is 386 class CPU core with 1 KB cache. 386 motherboard can be swap out for a standard 486 size motherboard and existing add-on cards can be recycled.

I can reuse my old 386DX's ET4000AX with my Pentium 150 PC, but a no-name OEM S3 Trio 64UV+ PCI card is cheap.

During 1992 year, Apple sold 2.5 million Macs with LC-II with 16 Mhz 030 and 16-bit bus being its best seller.

In 1992, Commodore has unsold A3000 stocks have either 68030-16 or 68030-25 with unwanted pre-VGA class ECS graphics.

During 1992, Apple sold 2.5 million Macs while A1200's production scale is about 44,000 units.

In Jan 1993, Apple released LC-III @ 33 Mhz 68030 and 32-bit external bus. A4000 with 68EC030 @ 25Mhz was released in April 1993. For Xmas Q4 1993, Apple released 68LC040-25 Quadra 605/ LC 475 / Performa 475 with USD1000 price range which is on par with 486SX25 and 486SX33 based PC's price range.

In Australia's Q4 1992 market, 386DX-33 and 386DX-40 based PC clones are approaching the price range of 1989 A500 + 1084S monitor combo.

1992's large production volume 68030 desktop computer offerings during 1992 wasn't great.

CPU alone does not complete the desktop microcomputer.

Remember, Intel's revenue majority is from 486 starting from the year 1992.

If my ex-corporate A3000/030 wasn't purchased in early 1992 and then 386DX-33 PC for Xmas Q4 1992, my PC would be 486DX33 class.

Quote:

The 68060 has better integer performance than the Pentium much like the Cyrix 6x86

68060 has 4 bytes (32-bit) fetch from instruction L1 cache problem, only two 2 byte instructions for sustained dual pipelines. SysInfo benchmark easily exceeds this 68060's design flaw.

Cyrix 6x86's integer performance advantage didn't translate for PC's Tomb Raider which also used FPU.

PC's Tomb Raider (DOS 1996) was designed for Pentium while PS1's Tomb Raider version for its fixed point hardware.

https://www.youtube.com/watch?v=9AV3Dw1rKXg
Cyrix 6x86MX PR200 vs Intel Pentium 200 MMX
3DMarks 99's CPU 3DMarks
Pentium 200 MMX = 1170
6x86MX PR200 = 727

GLQuake.
Pentium 200 MMX = 43.6 fps
6x86MX PR200 = 21.7 fps

Cyrix 6x86MX PR200 was released in May 1997 and it wasn't the best Socket 7 CPU when AMD's K6 MMX 166 Mhz to 233 Mhz was released on April 2, 1997.

Cyrix 6x86's L1 instruction fetch is 16 bytes (128-bit) per cycle.

Quote:

I suspect you are too high on the PiStorm/Emu68/Pi4 estimation.

I'm game for another around Quake benchmark.

https://framebuffer.io/project/quake
Quake demo 1 640x480p
Pentium III EB @ 800Mhz = 68.2 fps
Pentium III @ 600Mhz = 52.2 fps

AmiQuake demo 1 640x480p
PiStorm32 / RPi CM4 / RTG = 59.43 fps

Quake demo 3 640x480p
Pentium III EB @ 800Mhz = 65.8 fps
Pentium III @ 600Mhz = 50.4 fps

AmiQuake demo 3 640x480p
PiStorm32 / RPi CM4 / RTG = 58.85 fps

PiStorm32 / RPi CM4 / RTG is around Pentium III 700 Mhz.



Quote:

The OoO Pentium 3 has strong performance and the core design is the basis of modern x86-64 cores.

Intel Core 2's 64-bit ALU and 128-bit SIMD units are new.

Intel Core 2:
Port 0 has ALU, SSE Shuffle ALU, 128bit FMUL/ FDIV,
Port 1 has ALU, SSE Shuffle MUL, 128bit FADD,
Port 2 has Load AGU,
Port 3 has Store AGU,
Port 4 has Store Data,
Port 5 has ALU branch, SSE ALU,
Reorder buffer size: 96 entries
Decoder flat layout: 3 simple decoders, 1 complex decoder (4 uops), 1 microcode engine,
Basic ALUs are 64-bit wide. SSE SIMD units are 128-bit wide.
AMD K10 would need additional X86 decoder to match.


Intel Pentium III:
Port 0 has ALU, LEA, X87 FADD, X87/64bit FMUL, FDIV,
Port 1 has ALU, MMX ALU, 64bit FADD, Shuffle,
Port 2 has Load AGU,
Port 3 has Store AGU,
Port 4 has Store Data,
Reorder buffer: 40 entries
Decoder flat layout: 2 simple decoders, 1 complex decoder(4 uops), 1 microcode engine,
Basic ALUs are 32-bit wide.


Intel Pentium IV:
Port 0 has ALU+ALU, branch, store data (ALU double pumped at twice advertised clock speed)
Port 1 has ALU+ALU, 32bit MUL, Shift (ALU double pumped at twice advertised clock speed)
Port 2 has 64bit MUL, FADD, FMUL
Port 3 has Load AGU
Port 4 has Store AGU
Reorder buffer: 126 entries
Decoder layout stack:
1-way complex decoder, linked to microcode engine,
12K entry 8-way trace cache, linked to microcode engine,
Three micro-op queue.

Pentium IV bottlenecked by 1-way complex decoder entry point and heavily reliant on trace cache (decode cache) pathway. Trace cache miss will result in 486 level IPC 1-way complex decoder. It's aggressive in elimination of multiple X86 hardware decoders and heavily use trace cache (decode cache). Intel wanted to kill X86 and high-clock speed marketing took over R&D.

Pentium IV's reorder buffer (OoO) is larger than Core 2's. The major bottleneck for Pentium IV is the front end i.e. 1-way complex decoder.

Intel Core i series 1st gen Nehalem,
Port 0 has ALU/MMX ALU, FP MUL, 2x AGU, SSE MUL/DIV/move,
Port 1 has ALU/MMX ALU, FP ADD, SSE ADD/move,
Port 2 has AGU, Load address,
Port 3 has AGU, Store address,
Port 4 has Store data,
Port 5 has ALU/MMX ALU, branch, SSE ADD/move,
Reorder buffer size: 128 entries with fusion,
Decoder flat layout: 3 simple decoders, 1 complex decoder (4 uops). 1 microcode engine is attached 18 entry X86 instruction allignment macro-op fusion.
28 entry decoded instruction queue.
Basic ALUs are 64-bit wide. SSE SIMD units are 128-bit wide.

My point, Pentium III wasn't reused for Core 2.

Last edited by Hammer on 28-Feb-2025 at 04:20 AM.
Last edited by Hammer on 28-Feb-2025 at 04:14 AM.
Last edited by Hammer on 28-Feb-2025 at 03:16 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@Hammer

Quote:

Hammer wrote: During 1992, Am386-40 has a lower cost advantage over 68030-25 and i386DX-25.

There you go again, bringing up irrelevant information. Nobody cares whether AMD DX40 did or didn't have a 'lower cost advantage' over 68030 in the 90's.

Quote:

My 386DX-33 has 64 KB external cache on the motherboard.

And we don't care about that either.

If you don't have anything useful to add to the conversation then what are you going here?

@bhabbott

Quote:

There you go again, bringing up irrelevant information. Nobody cares whether AMD DX40 did or didn't have a 'lower cost advantage' over 68030 in the 90's.

Price matters.

https://www.youtube.com/watch?v=YTFH9TAmwhY

When it comes to 3D games, Matt's claim doesn't hold water.

Quote:

And we don't care about that either. If you don't have anything useful to add to the conversation then what are you going here?

Look in the mirror.

Last edited by Hammer on 01-Mar-2025 at 06:38 AM.
Last edited by Hammer on 01-Mar-2025 at 06:24 AM.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@OlafS25

Quote:

of course you only compare pure processing power here

Yeah. That's because for my SW rasterizer, the only thing that matters is how fast the CPU is. Secondarily, the RAM bandwidth is of high importance (each pixel is 16 or 32 bits).

Until I get the benchmark numbers from each system for my own benchmark, the Sysinfo MIPS is the only reference point I have.
Quake is certainly the second best, despite being floating point heavy (which I am not).


Quote:

it also depends on the software, how good it make use of the specific amiga hardware

And that is especially true when talking about Vampire, because:
- AMMX has potential to increase performance
- Maggie certainly should increase performance

AMMX is a deep rabbit hole. I would have to rearchitect each engine component around that. Not happening this year.

Maggie - last I checked, it couldn't do entire triangles, only scanlines. So, I am not sure how much of a speedup it is for flat shading. I'm sure Maggie will be faster for texturing than CPU.
I will spend one weekend with Maggie to try it out, because especially at higher resolutions (1280*720), filling the scanlines takes , basically, 90% of CPU time.

Quote:

The apollo team showed a game from around 2004, that needed at least 1 Ghz on PC and worked good on apollo hardware. But of course only after lots of optimizations and adaptions to the hardware

I forgot about that. Which game are we talking about?

@Hammer

Quote:

The Amiga camp still has the MAGA mindset i.e. Make Amiga Great Again.

I want a full merch with that 😄 But at the very least, hats with a boing ball
However, it might get me killed in North Dakota, once the random folks would realize the gravity of the situation upon chatting me up

@Hammer

Quote:

Ice Drake, Firebird, Manticore, Salamander and Phoenix add-on accelerator SKUs are also AC68080 V4, not just the V4SA. AC68080 V4's clock speed is 92 or 100 MHz. I don't consider official clock speed as overclock since they are OEM's warranty clock speeds.

I guess I just keep using the term from Apollo forums. It's an FPGA board, after all, so it's technically not an overclock, despite the fact they're using the term multiplier (x11, x13, x16).
But I meant the speed increase compared to the original "clock" of about 80 MHz.


Quote:

15% increase on clock speed from 85 Mhz to 100 MHz, hence about Pentium 140 level, hence I agree with your estimation.

I just checked the forums and found a thread about 205 MIPS/ x16 / 113 MHz!

That's a significant speed-up from the original ~80 MHz even for my integer purposes alone! Nice �
Once I start testing on V4SA around April,I should update the core, as they made some rasterizer updates to Maggie recently.

Quote:

My Pentium Rating is based on Quake results that are heavy on the FPU and may differ with fixed point 3D engines.

Fully agree! But it's still order of magnitude more expressive compared to just a Sysinfo MIPS value as it's a SW rasterizer.
Well, for me, anyway...
Hopefully, the relative performance differences translate also to Integer pipeline performance