Four months ago, on November 22, 2025, the MARS FPGA project was declared dead. The promising MARS FPGA console is dead, according to one of the project’s developers. Todd Gill broke the news on Discord, revealing that the person in charge of the MARS board development “just disappeared,” putting the project on ice. No launch event. No refunds to process, because there was nothing to refund. Just silence from the one person who held the keys to the circuit board, and a team left standing in an empty room.

The timing is worth sitting with. November 22 is also the anniversary of the Sega Saturn’s Japanese launch in 1994. Whether that’s coincidence or cosmic irony, it lands.

The Machine That Was Going to Change Everything

MARS stood for Multi Arcade and Retro System. It was pitched as “an all-new open-source FPGA-based system that dials everything up to eleven,” boasting custom-designed hardware that eclipses the DE-10 Nano, poised to dethrone the MiSTer in terms of compatibility and versatility. The spec sheet was intoxicating. It ran on a 400MHz Ti180 M484 FPGA with significant advantages over the current DE10-Nano board, at least in terms of on-paper specs. That meant 180,000 logic elements on a 16nm process, compared to MiSTer’s roughly 113,000 on a 28nm chip. More room to breathe. More room to build.

MARS would output HDMI up to 4K, as well as RGB via a D-sub connector supporting YPbPr, VGA, composite, and S-Video. There was even MIDI input and output, bound to excite the retro-diehards who kept their MT-32 after all these years. JAMMA support was built in. Snap-in controller modules offered zero-latency SNAC connections for original NES, SNES, Genesis, and Neo Geo pads. The release date was early 2024, with a $699 price point.

And then Mike Chi signed on.

The RetroTINK Seal of Approval

Chi is the electrical engineer behind RetroTINK, whose line of video scalers has quietly become the gold standard for anyone trying to get a clean signal from vintage hardware onto a modern panel. His involvement with MARS wasn’t ceremonial. The MARS team announced they’d be integrating his display processing expertise directly into the hardware, bringing auto-scaling, shadow masks, and scanline filters from the RetroTINK-4K into the MARS platform. Chi himself said he was “deeply honored and humbled to be a part of this amazing project.”

For the FPGA faithful, this was validation. With hardware experts like Mike Chi, atrac17, and Wizzo lending their expertise, many had high hopes for this new contender. The MiSTer, for all its brilliance, is a hobbyist’s device. You buy a DE10-Nano development board, bolt on I/O and RAM expansion boards, 3D-print or buy a case, and hope you’ve soldered everything right. MARS promised a polished, integrated product. A console, not a kit.

The Cracks

Early 2024 came and went. No MARS. In March, Todd Gill insisted the project was “proceeding as planned” and promised a showing at the TooManyGames convention in June. That showing never materialized in the form anyone expected. Skeptics questioned atrac17 and “his strange choice of publicly crapping on the MiSTer project,” noting that this wouldn’t attract customers to MARS.

In April 2024, Wizzo, the MiSTer Extensions developer who’d joined the MARS team with considerable fanfare, quietly departed. He said he left “on good terms” but couldn’t elaborate on the reasons without revealing private information. Then came the blocking. The official MARS Twitter account began blocking journalists and community members, including Time Extension, the outlet that had given the project some of its most favorable coverage.

On the MiSTer FPGA forum, one user summarized the mood: “So far, all MARS has produced is theories and promises, not one of which has been kept.”

By November 2025, it was over. The team had included Flxel, developer of 3DO and Dreamcast emulators, and RetroTINK creator Mike Chi. But the person in charge of the MARS board development “just disappeared,” and the rest of the team was effectively “ghosted.” Community members identified the vanished party as Flxel. No one was scammed. No preorders were taken. The project simply evaporated, like a signal with no receiver.

Meanwhile, in Ukraine

Here is where the story pivots from disappointment to something closer to awe. While MARS was being announced, delayed, and buried, a Ukrainian developer named Sergiy Dvodnenko was doing something that many in the FPGA community had said was impossible.

He was building a Saturn core for MiSTer. On the old hardware. The cheap hardware. The 113,000-logic-element Cyclone V that MARS was supposed to make obsolete.

The Sega Saturn has historically been one of the harder consoles to emulate due to its infamously difficult architecture consisting of 8 processors overall and use of quadrilaterals rather than triangles as polygons. The Saturn has a dual-CPU architecture and eight processors. Two Hitachi SH-2 CPUs sharing a single bus, two video display processors with completely different rendering philosophies, a Motorola 68EC000 running the sound subsystem, a Yamaha DSP, and a Saturn Control Unit tying it all together with its own coprocessor. Rodrigo Copetti’s architectural analysis notes that the Saturn houses “a significant number of processors (eight!)” and that engineers grouped the circuitry into four separate subsystems just to keep them from fighting over resources.

Yu Suzuki, the legendary Sega developer, once observed that most programmers could only extract about one-and-a-half times the performance of a single SH-2 from the dual-CPU setup, and that “only 1 in 100 programmers are good enough to get this kind of speed [nearly double] out of the Saturn.” If it was hard to program in 1995 with full documentation and Sega’s internal tools, imagine recreating it in programmable logic on a $200 board, from a country at war.

After nearly two years of iterations, core updates, code updates, a whole new RAM component rewrite, the dev Srg320 living in and then evacuating Ukraine, the Sega Saturn MiSTer FPGA core hit a big accuracy milestone: a 100% “pass” on the SCSPtest. The core now passes every test in the sound chip testing tool, and “the core is the first system after the original hardware to pass all the SCSP tests,” according to MiSTer community member Zet-sensei. No software emulator had achieved that.

By mid-2025, Dvodnenko updated the repo and added the Saturn FPGA core to the main distribution branch with 100% of the Saturn library working, with just a few tiny loose ends like some wonky dual stick support. The impossible core was possible. And it was free.

Just two days ago, on March 20, 2026, an update for the MiSTer FPGA’s Saturn core improved the performance of many 3D fighting games when using just one memory module, stretching the DE-10 Nano to its limits and beyond. Users initially got an error stating the core was trying to use more resources than the DE-10 Nano has. Zet-sensei then got to work to make the new core fit in the FPGA’s silicon, and within a few hours, he generated a working build. They are literally filling every available logic cell on the chip. There is no room left. And they are still finding ways to make it better.

What the Siege Teaches

The MARS saga and the Saturn core’s triumph are not separate stories. They are the same story told from opposite ends.

MARS believed the path forward required bigger silicon, more logic elements, a fresh start on new hardware. That belief wasn’t wrong, exactly. A more powerful FPGA would allow for more cores than MiSTer can support, with potential for hybrid emulation where the FPGA handles the board and an ARM chip handles the DSP audio. Dreamcast, NAOMI, 3DO: these were the promises that justified the $699 ask and the years of waiting. None of them shipped.

Dvodnenko’s Saturn core, by contrast, is an argument for constraint as a creative force. As Zet-sensei put it: “From one of the least well-documented consoles a few years ago, we now have one of the most precise cores in the MiSTer ecosystem.” He did it on hardware that the MARS team considered insufficient. He did it while evacuating a war zone. And he did it in the open, with every commit visible on GitHub.

Dvodnenko himself remains characteristically measured about his achievement. When publications declared the Saturn core “100% hardware accurate,” he clarified that “the core is still far from being 100% accurate,” though most people playing it on the MiSTer won’t be able to tell the difference. That kind of honesty is its own form of credibility, the polar opposite of hype cycles and convention demos that never quite arrive.

What Remains

The FPGA retro gaming scene is fractured. Analogue makes beautiful, closed hardware. MiSTer remains open but aging. The Replay2 project is still in development, promising an Intel 7nm Agilex 5 FPGA, but it carries the same vapor risk that sank MARS. As one commenter noted when MARS died: “Well, this was a surprise to… no one. Every time, something like this is announced, I’m already prepared for it to be dead, within the next couple of years.”

But here is what I keep coming back to. As one forum member wrote years ago, when the Saturn core was still a fantasy: “Having the Saturn in FPGA will open it up to a far larger audience, and ensure it survives as a platform; the hardware itself is both limited in availability, dying and difficult to emulate.” That survival has now happened. Not because a $699 box arrived in a slick injection-molded case with 4K output and MIDI ports. It happened because one developer, working alone with a Patreon and a cheap development board, decided the Saturn was worth saving.

MARS never landed. But the Saturn, after thirty-one years, has finally been caught. And the net that caught it was woven from open-source code, community testing, and the stubborn conviction that you don’t always need bigger hardware. Sometimes you just need a better engineer.