As @geennaam stated on amiga-news.de
(but you might have already seen that)

The Mirari is not based on any existing development board. Its design was created entirely from scratch, following the design rules and component library provided by our manufacturing partner in China. Our goal was to deliver a well-specified board at the lowest reasonable cost, so we included features that added value without significantly increasing the bill of materials.

The choice of the T1042 was deliberate: it is much cheaper than the T208x, it uses the proven and supported e5500 core and provides all the interfaces required for a mainboard. Using a more expensive T2080 would have forced us to move to a 10-layer PCB instead of 8 layers, while offering only marginal advantages over the T2081. In addition, the e6500 core was not supported by ExecSG at the time, and likely not by MorphOS either, so software support effort was also a factor.

As for similarities with the Tyche board, I can only speculate. Many component placements are dictated by the processor pinout, and others by the ATX form-factor requirements, so some resemblance is inevitable.

Mirari's compatibility with the T2081 is simply a bonus and was treated as a secondary design objective. The only real T2081 specific design choice was to build the audio interface (I2S) inside the FPGA. Because the T208x does not have a peripheral which can interface directly with an audio codec. The Tyche board took a more risky approach by using an older consumer grade PCIe audio chipset. Another one was to use DDR3 instead of DDR4. Because unlike the T1042, the T208x does not support DDR4. But since the DDR4 interface operated at the same frequence as the DDR3 interface, increased latencies would have resulted in lower poerformance compared to DDR3.
One might argue that DDR4 offers better availability, but we do not see any issues at the moment to aqcuire large SODIMMs at a reasonable price. Given the current DRAM situation, the price and availability of DDR3 might even be better in the near future.

The reduced number of SerDes lanes and lack of SATA on the T2081 are not real drawbacks in this context. The PCIe Gen3 controller in the T208x family is limited to four lanes, so even if you wire eight lanes to a x16 slot, it still operates in PCIe 2.0 mode. Which has no benefit from a bandwidth pov. GPU performance on next-generation PowerPC systems is constrained far more by CPU capability than by PCIe bandwidth.

The same applies to NVMe and M.2 storage: Amiga filesystem performance stay far below 500MB/s, so the available slot bandwidth is not the limiting factor.

The absence of onboard SATA is also not an issue. The board includes two NVMe ports, which provide ample storage options. While I cannot comment on OS4 support yet, both Linux and MorphOS install perfectly well from a USB drive. Modern ATX cases rarely include external 5.25" bays, and optical drives have largely disappeared from laptops as well. For those who still need one, a USB DVD drive or a PCIe SATA card works fine. The Sil3132 SATA2 controller operates correctly under Linux and MorphOS on the Mirari, and the system should also be able to boot from it, though I have not tested that scenario yet.


Hope this helps.