@Mobileconnect

I believe SysSpeed and AIBB give more accurate synthetic benchmarks, however given the plethora of hardware options nowadays, it would be cool to have a more modern, accurate HWInfo and Benchmarking tool all on one

_________________
Live Long and keep Amigaing!

A1200, Power Tower, TF1260 128MB RAM, 68060 Rev 6, OS3.9 BB2, HD-Floppy, Mediator TX+ PCI, Voodoo 3 3000, Soundblaster 4.1, TV Card, Spider USB, 100MBit Ethernet, 16GB CF HD, 52xCDRom.

@Amiboy

I suspect all these these older benchmarks are utterly destroyed by the cache sizes of the cpus used to emulate 68K these days. The performance offered by JIT emulation on x64 and ARM these days, heck even the faster PPC systems likely exceeds the wildest imaginings their authors originally had.

I think the only benchmarks that matter are how well your preferred software works.

_________________
Doing stupid things for fun...

@Karlos

Ahh yeah didn't think of that :-/

_________________
Live Long and keep Amigaing!

A1200, Power Tower, TF1260 128MB RAM, 68060 Rev 6, OS3.9 BB2, HD-Floppy, Mediator TX+ PCI, Voodoo 3 3000, Soundblaster 4.1, TV Card, Spider USB, 100MBit Ethernet, 16GB CF HD, 52xCDRom.

@cdimauro

Quote:

cdimauro wrote: @bhabbott Quote: bhabbott wrote: The Amiga stretches this further than most other vintage computers because it was designed to be extended with virtually no limit. In the 80's and 90's we were shoving the fastest 68k we could get into them, and then PPC in the late 90's. Also higher resolution graphics and 16-bit sound chips, integrated into the OS. What about PCs, then?

PCs not so much. Unlike the A1000, A500 and A2000, the PC/XT and AT did not have a CPU expansion slot, so installing a later generation CPU was tricky.

Orchid made a 286 card that plugged into an XT bus slot, with a ribbon cable going to the 8088 CPU socket. The 8088 was transferred onto the card so the machine could be used with either CPU. It didn't have any onboard RAM or 16-bit bus connector so you were stuck with the motherboard RAM (no extended memory beyond 1MB) and 8-bit cards. A later version did have onboard cache RAM though, which mitigated for the slow 8-bit motherboard. I don't think they were very popular (I have never seen one in person).

In general if you wanted to upgrade your PC to a later CPU and/or wider bus slots etc. it was time for a new motherboard, and in many cases a new case, power supply and keyboard too. IBM expected you to buy a whole new machine.

@OneTimer1

Quote:

OneTimer1 wrote: Ever tried AIBB (read the full story here https://www.edsa.uk/blog/the-emu-and-the-storm and stay away from AOS3.2)? It is a bit like SysInfo but gives more precise information:

Interesting.

The V2 Vampire in my A600 does the Beachball test 872 times faster than a stock A600, which is 85% of EMU68's speed. TGTest is 3.27 times faster, 78% faster than EMU68. SysInfo says chip speed is 3.6 times faster than a stock A600. This makes my A600 quite snappy in OCS screenmodes, especially on a CRT (there is a noticeable delay of ~2 frames in composite on my Samsung 32" LED TV, caused by the TV's upscaler).

I think EMU68/PiStorm may have been updated since 2022 for better ChipRAM speed? Still not as good as the Vampire though. Of course V4 Vampires are many times faster when running SAGA, and don't have the upscaler delay either when outputting on HDMI.

@bhabbott

The beach ball test, if I recall, renders the ball, pixel by pixel, to the chip ram display. Single pixel rendering from the CPU to chip ram involves reading words from the chip ram, setting and clearing individual bits in each word and writing them back. There's no question that this is going to dramatically impact peak performance, as you get faster and faster at computation, it'll approach the read/write limit for the chip ram access.

_________________
Doing stupid things for fun...

@bhabbott

Quote:

bhabbott wrote: I think EMU68/PiStorm may have been updated since 2022 for better ChipRAM speed? Still not as good as the Vampire though. Of course V4 Vampires are many times faster when running SAGA, and don't have the upscaler delay either when outputting on HDMI.

If you are switching to faster accelerators, the ChipRAM speed is becoming the major bottle neck, a full emulation or a 'non original AGA' system like the Vampire V4SL, might have faster ChipRAM access.

You might experience the improvement in 3D FPS games using AGA or RTG.

Some 030 cards where faster than 060, because their flexible bus access was better suited for ChipRAM than the always full 32Bit access of a 040/060.

I was told that AB3D ran better on a Blizzard 030/50 than on a Blizzard 060/50, I don't know where Emu68 / Amiberry or other systems might be and exactly this is a problem when testing with benchmark programs.

@OneTimer1

Quote:

I was told that AB3D ran better on a Blizzard 030/50 than on a Blizzard 060/50,

Given the tiny logical resolution of just 96x80 pixels, I find it hard to believe this would be the product of slower chip ram access. It's more likely to be down to a trap and emulate issue for some unimplemented instruction, or possibly something to do with setting up some interesting chip ram attributes in the MMU, e.g. marking it as imprecise rather than straight up uncached.

_________________
Doing stupid things for fun...

@bhabbott

On Emu68 version with PiStorm16 which I have now (CM4, 1.8 GHz) the beachball benchmark is 2293 times faster than stock A600. I obtain the very same result even with CM4 overclocked to 2.2 GHz since writing pixels to chip memory is the limiting factor. TGTest score is 2.57 versus stock A600.

Regarding memory speed I will not discuss SysInfo results since these are not reliable and the score varies from version to version. Instead I used bustest tool. On the same PIStorm16 it gives write speed of 3.5 MB/s, which is physical top limit for 16-bit amiga. Reads are still slower, but not as bad as on classic PiStorm: word reads at 2.3 MB/s, long/multiple at 2.8 MB/s.

@Karlos

Quote:

Karlos wrote: @bhabbott The beach ball test, if I recall, renders the ball, pixel by pixel, to the chip ram display. Single pixel rendering from the CPU to chip ram involves reading words from the chip ram, setting and clearing individual bits in each word and writing them back. There's no question that this is going to dramatically impact peak performance, as you get faster and faster at computation, it'll approach the read/write limit for the chip ram access.

11.29 seconds to render the Beachball. Pixel writing is only a small part of it.

As michalsc just pointed out, PiStorm16 with CM4 at 1.8 GHz is twice as fast due to both more compute power and faster ChipRAM access.

In the future we can expect even faster operation as more powerful chips and better interfaces are developed. Even that 3.5MB/s ECS ChipRAM limit could easily be smashed with a bit of extra hardware - if anyone can be bothered when much faster video hardware is available. In their current states both PiStorm and Vampire are plenty fast enough to do anything you would reasonably want on ECS.

Quote:

On Emu68 version with PiStorm16 which I have now (CM4, 1.8 GHz) the beachball benchmark is 2293 times faster than stock A600. I obtain the very same result even with CM4 overclocked to 2.2 GHz since writing pixels to chip memory is the limiting factor. TGTest score is 2.57 versus stock A600. On the same PIStorm16 it gives write speed of 3.5 MB/s, which is physical top limit for 16-bit amiga. Reads are still slower, but not as bad as on classic PiStorm: word reads at 2.3 MB/s, long/multiple at 2.8 MB/s.

.....Maaaaaaaaaaaaatttttttheeeeeyyyyyyyyyyyyyyyyy!!!!!!......

Last edited by MEGA_RJ_MICAL on 26-Nov-2025 at 12:50 PM.

_________________
I HAVE ABS OF STEEL
--
CAN YOU SEE ME? CAN YOU HEAR ME? OK FOR WORK

@bhabbott

I think you misunderstood my point. Yes, writing the pixels is a small part, but it's also the only part that's fundamentally limited by the existing hardware. As computation gets faster, that small part will become a larger proportion of the total because it's already as fast as it can be.

Last edited by Karlos on 26-Nov-2025 at 12:56 PM.

_________________
Doing stupid things for fun...