Emulation, FPGA Simulation, or Original Hardware: An Honest Comparison for Retro Gaming on a Modern Display

The single most useful fact for anyone comparing emulation vs fpga vs original hardware retro gaming is this: software emulation using run-ahead can deliver lower total input lag than an original Super Nintendo connected to the same TV, but it achieves that by sacrificing cycle-accurate timing. A SNES game running in RetroArch with two frames of run-ahead removes 33 milliseconds of display lag, while the original SNES with a RetroTINK 5X-Pro adds roughly 0.25ms of scaler lag plus the TV's own processing. The catch is that run-ahead works by saving and rolling back the emulator state, which can break audio synchronisation and introduces glitches in games that poll input during vblank. For a competitive shmup player who needs every millisecond, that trade-off is worth it. For someone who wants the game to behave exactly as it did on a CRT in 1991, it is not.

Every retro gaming display chain on a modern TV is a series of conversions, and each conversion adds either lag or visual change. Original hardware outputs a 15kHz analogue signal at 240p or 480i. A modern flat-panel expects a digital signal at 1080p or 4K. The bridge between them is a scaler, a mod, or an emulator. You cannot skip the conversion; you can only control how much delay and artifact you accept.

Original Hardware Through A Scaler

Original hardware through a good scaler like the RetroTINK 5X-Pro adds roughly 0.25ms of framebuffer lag. The same console through a cheap converter from Sham Shui Po treats 240p as 480i, applies a destructive deinterlace that discards half the vertical resolution, and adds two to four frames of lag. That is the difference between a playable game and a smeared mess. The failure case is buying any generic HDMI adapter labelled 'upscaler' without verifying that it handles 240p correctly. If you already own one and your Super Famicom looks combed, that is why.

FPGA Simulation

FPGA simulation on a MiSTer DE10-Nano board outputs a signal that is bit-for-bit identical to the original console at the logic level, with zero added lag over the analogue I/O board. Over HDMI the MiSTer's ASCAL scaler adds 0 to 1 scanlines of lag. The Analogue Pocket through its dock outputs 1080p, 1440p, or 4K, but the default scaling is not integer, so pixel sizes are uneven until you enable integer scaling in the menu. That is a configuration default, not a hardware limit.

Software Emulation

Software emulation on a PC or Raspberry Pi runs the console's program on a general-purpose CPU. The emulator approximates the original hardware behaviour rather than replicating it at the gate level. The accuracy metric is cycle-accuracy: whether the emulator reproduces every CPU and PPU cycle in the correct order. Bsnes, Mesen2, and Ares are cycle-accurate for their target consoles. DuckStation is cycle-accurate for the PS1. PCSX2 and Dolphin are not fully cycle-accurate but use hardware acceleration to run games at higher resolutions.

The Shmups Forum lag test methodology, using a Time Sleuth lag tester, has measured the MiSTer's HDMI output at under one millisecond when no framebuffer is used. The Analogue Super Nt in zero-delay mode measures under one frame of lag at 1080p. An original SNES 1CHIP model through a RetroTINK 5X-Pro at 1080p measures roughly 0.25ms of scaler lag plus the TV's processing. The difference between the MiSTer and the original console is negligible for most players: both are under one frame. The difference between either of those and a software emulator with run-ahead enabled is that the software emulator can be faster than the original console, but only by simulating the hardware rather than replicating it.

The practical outcome is that a player who perceives milliseconds of input lag for rhythm games or shmups will not see a difference between MiSTer FPGA and original hardware when both are on the same TV. The bigger variable is the TV itself. A Samsung or LG OLED in Hong Kong with Game Mode on still adds 10-15ms at 1080p and 30-40ms at 4K. That is the panel's inherent processing, and no scaler or mod can remove it. Black frame insertion at 120Hz output can reduce perceived motion blur, but it does not reduce lag.

Cycle-accurate emulation reproduces every CPU and PPU cycle in the exact order the original hardware executed them. Bsnes, Mesen2, and Ares achieve this for the SNES, NES, and multiple systems respectively. DuckStation is cycle-accurate for the PS1. The cost is performance: a cycle-accurate SNES emulator requires a modern CPU to run at full speed, while a less accurate emulator can run on a Raspberry Pi. The benefit is that games behave identically to the original hardware, including timing-dependent tricks like the Super Scope light gun or the SNES's sprite priority glitches that games like Super Mario World rely on for correct rendering.

Approximate emulation, used by PCSX2 and Dolphin, runs the game at higher internal resolutions but does not guarantee that every hardware quirk is reproduced. For most PlayStation 2 and GameCube games, the difference is invisible because the original hardware already had a framebuffer and the emulator's approximation matches it. For games that depend on precise GPU timing, like some PS1 racing games, the emulator may introduce visual glitches or timing errors that the original hardware did not have. The 240p test suite ROM is the standard tool for verifying accuracy: run it on the emulator and compare the output to the same test on original hardware or a MiSTer core.

The Analogue Pocket uses an Altera Cyclone V FPGA with 49,000 logic elements, the same chip used in the Analogue Super Nt and Mega Sg. On its built-in 3.5-inch LTPS LCD at 1600x1440, input lag measures under one frame. Through the Analogue Dock at 1080p, the same measurement holds. The original Game Boy Advance SP with an IPS mod adds roughly one frame of lag from the mod itself, plus the screen's response time. The Pocket's FPGA replication is deterministic: it does not suffer from capacitor aging or voltage regulator drift that can shift timing on a twenty-year-old original console. The trade-off is that the Pocket cannot play original Game Boy cartridges that rely on the link port for multiplayer without a separate adapter, and its docked output defaults to non-integer scaling, which you must manually correct in the menu.

For a Hong Kong apartment gamer who wants one device for portable and TV play, the Pocket plus Dock replaces an original Game Boy, Game Gear, and several other handhelds. The failure case is assuming the Pocket's docked output is lag-free by default: it is, but only after you enable integer scaling and disable any TV processing. The Analogue DAC accessory is required if you want to output to a CRT monitor, adding RGB, component, S-Video, or composite at 240p, 480i, or 480p.

The cost of running original retro hardware on a modern TV in Hong Kong is not the console. It is the signal chain. An original SNES costs roughly a few hundred Hong Kong dollars in Sham Shui Po. The RGB SCART cable from Insurrection Industries or Retro Gaming Cables adds a similar amount. The RetroTINK 5X-Pro is the largest single expense. If the console is a PAL model, you also need a step-down transformer for Japanese 100V consoles at 220V, adding further cost. If the console lacks RGB output, like a North American N64, you need an RGB mod kit and installation. If you want to avoid the scaler entirely, an internal HDMI mod like the PS1Digital or N64Digital costs more again, plus installation.

The total for a single console on a modern TV runs into several thousand Hong Kong dollars before you buy a single game. The MiSTer DE10-Nano board covers thirty consoles and arcade systems and outputs directly to HDMI, at a fraction of that total. The Analogue Pocket covers handhelds. The cost argument for original hardware only works if you already own the console, the cables, and a CRT, or if you specifically need the original cartridge reading experience for light gun games or accessories that FPGA and emulation do not support. Check current prices from the official retailers before committing; they change.

Original hardware is not static. The capacitors that smooth the power supply and the voltage regulators that maintain stable voltage age over time. A 1990 SNES with original capacitors outputs a slightly different sync signal than it did in 1990. The difference can cause sync drop on white flashes, where the scaler loses lock for two to three seconds. The Shmups Forum has documented this failure mode extensively: the SNES's sync signal dips during full-white frames, and a scaler with tight sync tolerance drops the signal. Replacing capacitors restores the original timing, but that is a maintenance step that FPGA simulation and software emulation do not require.

FPGA logic gate replication is deterministic. The MiSTer's SNES core, developed by srg320, configures the DE10-Nano's Intel Cyclone V logic gates to match the Ricoh 5A22 CPU and the 5C77 and 5C78 PPUs. The core does not age. The Analogue Pocket's Altera Cyclone V does not drift. Software emulation is also deterministic in the sense that the same emulator version on the same OS produces the same output every time, but the host hardware introduces variables: GPU driver version, operating system scheduling, and background processes all affect timing. The MiSTer's HDMI output via the ASCAL scaler has a fixed latency of 0 to 1 scanlines regardless of what else the board is doing.

The TV is the largest source of lag in any chain. A Hong Kong apartment with a 4K OLED from LG or Samsung adds 10-15ms with Game Mode on at 1080p and 30-40ms at 4K. That is the panel's own scaler and image processing. No external scaler, mod, or emulator can remove it. The only way to reduce it is to use a CRT or a gaming monitor designed for low latency. A Sony PVM-20L5 professional monitor adds zero processing lag, but it is a 15kHz CRT that accepts 240p to 1080i and commands a high price in Hong Kong second-hand markets, if you can find one.

The RetroTINK 5X-Pro and RetroTINK 4K include black frame insertion at 60Hz or 120Hz output, which reduces sample-and-hold blur on OLEDs. BFI does not reduce lag; it reduces perceived motion blur by inserting a black frame between each game frame. At 120Hz BFI on a 60Hz source, the effect is a clear motion image without visible flicker. The failure case is assuming that BFI solves lag. It does not. The only way to reduce total lag is to reduce the TV's processing, and that means using a display with a measured lag under 5ms at the resolution you play at.

Under Hong Kong law, the legality of software emulation depends entirely on how the BIOS and game files are obtained. Emulators themselves are legal software. Dumping your own BIOS from a console you own and your own game cartridges is generally accepted as lawful. Downloading BIOS files or ROMs from the internet without owning the original hardware is not. This is not legal advice. If you are uncertain, the Hong Kong Retro Game Association organises meetups where members can discuss the practical and legal aspects of emulation with people who have verified their hardware claims physically.

The single thing that most often goes wrong is that a buyer spends heavily on a RetroTINK 5X-Pro and a full RGB SCART cable chain, plugs it into their LG OLED, and still feels lag. The scaler is not the problem. The TV's game mode at 4K adds 30-40ms, and that dwarfs the scaler's 0.25ms. Check the TV's input lag at the resolution you actually play at. Use the 240p test suite's lag test to measure total lag from button press to on-screen response. If the total is over 50ms, the TV is the bottleneck, not the console, not the scaler, not the emulator. Set the TV to 1080p output if it allows it, enable game mode, and test again. If the lag is still unacceptable, a gaming monitor or a CRT is the only solution.