The ZORRAM

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

@matthey

Quote:

I expect the ARM SoC will still use in-order ARM cores because of the lower price and passive cooling. It may have small upgrades like a Cortex-A53 to Cortex-A55 and higher performance memory which should provide a small boost.

The A55 based Allwinner T536 SoC then? That would be ok! Keep praying!

A bit of a jump in price though! I still think a small overclock of a A53 based SoC is more likely.

Source Forlinx Embedded via Google AI
Quote:

The Allwinner T536 is a high-grade, industrial-tier SoC that costs about \\$55, primarily targeting automotive and ADAS solutions. In contrast, the Allwinner H6 (aimed at home media) is a consumer-grade chip that costs between \\$3 and \\$20 depending on the variant.

Last edited by BigD on 22-May-2026 at 11:03 AM.

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

@matthey

They might get a A55 based Rockchip RK3566 for $30! Still a massive increase in cost not reflected in THEA1200's RRP!

Or more probable a MediaTek MT9666.

Last edited by BigD on 22-May-2026 at 11:09 AM.

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

@matthey

The jump in performance required is about right though. They just need a cheap low end SoC for a low end media centre or similar!

Quote:

ARM Cortex-A55 provides roughly 18% higher performance and 15% better power efficiency compared to its older A53 predecessor.

Last edited by BigD on 22-May-2026 at 11:17 AM.

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

@matthey

Ok, here's a front runner! If it had this and was optimised I'd be happy!


Quote:

The Rockchip RK3566 (which features a quad-core ARM Cortex-A55 processor and a Mali-G52 GPU) has a baseline SoC component cost ranging from $10 to $15 per unit depending on order volumes.

Source: https://www.lcsc.com

Quote:

The Amiga A500 Mini uses an Allwinner H6 or similar low-power ARM SoC, not a Rockchip RK3566. However, the RK3566 is a capable, widely used chip for retro-gaming. Devices running it can run many Amiga games and emulators or similar microcomputer emulation platforms

Source: Reddit via Google AI

Last edited by BigD on 22-May-2026 at 11:21 AM.

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

@BigD
Customers are expecting a performance upgrade from THEA500 Mini to THEA1200. There are plenty of ARM SoCs to choose from.

Most likely options to increase performance from THEA500 Mini to THEA1200
1, in-order Cortex-A53 with better memory - minor performance upgrade
2. in-order Cortex-A55 with better memory - moderate performance upgrade
3. OoO Cortex-A72 with better memory - large performance upgrade

The Cortex-A55 core is a mildly enhanced and better tuned Cortex-A53 core and is a worthwhile upgrade with minimal cost increase (load-to-use penalty drops from 3 cycles to 2 cycles, better power efficiency). SoCs in mass production quantity are likely $3-$7 USD. The RPi 3 continues to use an ancient SoC with Cortex-A53 and old LPDDR2 memory because of the RPi standard VideoCore GPU found in Broadcom chips limiting their choices. Despite the RPi 3 outselling the C64, RPI is losing business to competitors using Cortex-A55 and better memory. For example, the A600GS did not use a RPi 3 (or PiZ 2 with same SoC) and chose the mildly upgraded Orange Pi Zero 3 equivalent of #1 above to have a little better performance.

Raspberry Pi 3 / Pi Zero 2
SoC: Broadcom BCM2837
CPU: 4xCortex-A53
mem: LPDDR2

Orange Pi Zero 3
SoC: All Winner H618
CPU: 4xCortex-A53
mem: LPDDR4

The reduced 2 cycle load-to-use penalty of the Cortex-A55 over Cortex-A53 3 cycle penalty reduces stalls with emulation of 68k code where 68k CPUs have no load-to-use stalls so code is not scheduled to avoid them. The reduced load-to-use penalty not only improves general performance but improves the power efficiency (performance/W). Power efficiency is very important even for small SBCs or option #3 above, the OoO Cortex-A72 would be used.

Raspberry Pi 4
SoC: Broadcom BCM2711B0
CPU: 4xCortex-A72
mem: LPDDR4

The OoO Cortex-A72 reduces or avoids load-to-use stalls and has much better performance than the in-order cores but the core size is many times larger increasing costs and more power is dissipated increasing heat requiring a smaller chip fab process that also increases costs. There may be increased power supply and heat related costs under some conditions which increases with more aggressive OoO CPU cores. Notice there is no RPi Zero based on any OoO CPU core. The RPi 4 is a nice upgrade over the RPi 3 because there is no RPi 3 with a Cortex-A55 and LPDDR4. The problem for RPI is that there is no Broadcom SoC with Cortex-A55 CPU cores, LPDDR4 mem and VideoCore GPU.

https://en.wikipedia.org/wiki/VideoCore#Table_of_SoCs_adopting_VideoCore_SIP_blocks

I doubt THEA1200 would move up to an ARM SoC with OoO CPU cores as a SoC with Cortex-A55 cores and LPDDR4 is between a RPi 3 and RPi 4 in performance with a SBC price closer to the RPi 3 than RPi 4.

An ASIC SoC with 68060 CPU core and AA+ chipset, the one after AGA, would use fewer transistors than the less than $1 USD RPI RP2040 and RP2350 microcontrollers (MCUs) which are like SoCs but use on die SRAM memory instead of SRAM caches and external memory. The 68060 is roughly 10 to 20 times the performance of 68030 performance with clock speeds limited to 1990s silicon where semi-modern silicon may allow 10 times the clock speed or more. With such cheap transistor costs, quadrupling the L1 caches is cheap and even a L2 cache would be cheaper than much larger ARM SoCs with 4xCortex-A53 CPU cores and similar L1 and L2 caches. These caches could double performance again. The old in-order Intel Atom vs 386 CPU performance is a 100 to 200 times CPU performance improvement according to Google AI which is likely roughly where a modernized 68060 could be. Also, the memory footprint of the hardware could be much smaller or much more 68k Amiga system memory available as JIT emulation wastes a lot of memory and increases SBC costs. A modernized ASIC 68k SoC would not give the 68030 performance level of a RPi 3 using emulation, 68040 performance level of a RPi 4 using emulation, 68060@100MHz performance level of upgraded original Amiga hardware or AC68080@100MHz performance level of FPGA hardware. It would take the 68k Amiga to a whole new level of performance that is at least somewhat competitive outside the 68k/Amiga market as in-order CPU cores can be valuable as I have explained above. CISC in-order CPU cores have an innate performance advantage over RISC in-order cores where they can compete with limited OoO RISC cores. The 68060 outperformed the limited OoO PPC603 and the in-order Intel Atom outperformed the limited OoO Cortex-A9. It may be possible for an ASIC 68k SoC with enhanced 68060 cores to have the performance of a RPi 4 with price closer to a RPi 3.

Last edited by matthey on 22-May-2026 at 02:31 PM.