Emulation and Original Hardware: What the Neo Geo Teaches Us About Accuracy
The debate around video game emulation is often framed in overly simple terms. On one side, there is original hardware, commonly associated with authenticity, immediacy and technical truth. On the other, there is emulation, often seen as a convenient but necessarily inferior substitute. This distinction can make sense when comparing a well-maintained original machine to a poorly configured emulator running on unsuitable hardware. But as a general principle, it is technically incomplete.
Emulation is not a single, uniform category. Some emulators are designed primarily for speed, convenience or broad compatibility. Others aim to reproduce the behavior of a machine as accurately as possible, down to timing, interrupts, video behavior, sound generation and hardware quirks. Likewise, original hardware is not a single, perfectly stable reference. There are motherboard revisions, regional differences, BIOS versions, video output variations, aging components, repairs, modifications, display chains and controller differences. In practice, the useful question is not simply whether something is “real hardware” or “emulation.” The useful question is: what exact behavior is being reproduced, and how accurately?
The Neo Geo is a particularly good case study because it sits at the intersection of arcade history, home console collecting, technical documentation and modern reproduction. It is also a system surrounded by strong cultural expectations. For many players and collectors, the Neo Geo AES represents one of the most prestigious home consoles ever released. Its cartridges are large, expensive and physically impressive. Its arcade counterpart, the MVS, is equally iconic, associated with cabinets, arcade sticks and the SNK fighting game era. Because of that symbolic weight, discussions around authenticity can quickly become emotional.
Yet technically, the Neo Geo is also a well-documented system. Its architecture is not trivial, but it is relatively clear compared with many more obscure arcade platforms. It uses a Motorola 68000 as its main CPU, a Z80 for sound control, a Yamaha YM2610 sound chip, a sprite-based video system and separate ROM regions for program code, graphics, sound data and BIOS. The MVS and AES share a very close hardware foundation, and much of the same software can run across both environments. This makes the Neo Geo an ideal system for discussing what accuracy actually means.
A video game is not only a cartridge, a board, a shell or a box. Those objects matter historically and culturally, but the game itself is also code, data and behavior. It is a program running within a specific hardware environment. The cartridge is the carrier of that program, not the whole of the experience. If the contents of the cartridge are correctly dumped, verified and executed inside an environment that accurately reproduces the target hardware, the result can be technically very close to the original behavior. The physical object remains important, but it is not the only possible reference point.
This distinction matters because “original hardware” is often treated as though it were a single absolute standard. In reality, even official hardware can vary. The Neo Geo MVS and AES are often described as arcade and home versions of the same system, and broadly speaking that is true. But they are not perfectly identical in every detail. One important example is the master clock. The Neo Geo MVS uses a 24 MHz main clock, while the AES NTSC system uses a slightly different clock, around 24.167829 MHz. This difference is small, but it is real. It also affects derived clocks and video timing.
The difference is not usually something a player will consciously notice during ordinary play. But it is conceptually important. It shows that two legitimate, historical, official versions of the Neo Geo hardware can already behave slightly differently. The documented video refresh rate for MVS is not exactly the same as for NTSC AES. The gap is less than one percent, but it exists. If an emulator introduced that kind of difference without explanation, it would likely be criticized. Yet here the difference belongs to the original hardware family itself.
This does not mean that accuracy is relative to the point of meaninglessness. It means accuracy needs a defined target. Accurate to what? A single-slot MVS? A multi-slot MVS? An AES? A Japanese BIOS? A European BIOS? A specific board revision? A CRT arcade cabinet? A home console through RGB? A modern HDMI output? Once the target is defined, the comparison becomes much more precise. Without that target, “original hardware” becomes more of a cultural label than a technical standard.
This is where emulation can be misunderstood. A bad emulator, or a good emulator configured badly, can certainly produce an inferior experience. It may run at the wrong speed, use an incorrect BIOS, apply excessive filtering, introduce audio errors, mishandle timing or add unnecessary input latency. But these are failures of implementation or configuration, not proof that emulation as a method is inherently unsuitable. A serious emulator is not merely “playing a ROM.” It is attempting to document and reproduce a machine’s behavior.
Projects such as MAME are important in this respect because their purpose is not simply convenience. Their long-term goal is documentation and preservation. Emulation, at its most serious, is a way of preserving how machines worked while original reference hardware still exists and can be measured. This is especially important for arcade systems, which were built for commercial operation, not indefinite survival. PCBs age, capacitors fail, displays disappear, power supplies drift and rare components become harder to replace. If the behavior of those machines is not documented and reproduced, part of their history becomes increasingly fragile.
The Neo Geo benefits from this kind of accumulated work. Over the years, enthusiasts, developers, preservationists and emulator authors have studied its ROM formats, BIOS behavior, graphics system, sound hardware, protections, clock distribution and video timings. The result is that Neo Geo emulation, when done properly, is mature and technically reliable. This does not mean every emulator or every commercial device using emulation is equally good. It means the system itself is now well enough understood that high-quality software emulation can deliver excellent results.
The comparison with FPGA and ASIC-based solutions is useful here. FPGA implementations are often presented as more authentic than software emulation because they reproduce hardware behavior in programmable logic rather than running a software model on a general-purpose CPU. ASIC-based recreations go a step further by implementing logic in dedicated silicon. These technologies can be excellent. They can reduce overhead, provide precise timing and offer a dedicated hardware environment. But they are not automatically accurate by definition.
A modern FPGA core is not the original machine. It is a hardware-level reconstruction of the original machine’s behavior. A modern ASIC recreation is not literally a chip from 1990 brought back into production unchanged. It is a new implementation based on modern manufacturing and on someone’s understanding of the original logic. The quality of such a solution depends on the same essential factors as a good emulator: documentation, reverse engineering, testing, validation, corrections and design choices. A poor FPGA implementation can be inaccurate. A good software emulator can be highly accurate. The method matters, but it does not replace verification.
This is why marketing terms such as “no emulation” should be treated carefully. They may describe a real technical difference, but they do not automatically prove superior fidelity. A device that uses FPGA or ASIC logic may be excellent, but it still needs to be evaluated in terms of timing, sound, video, compatibility, latency, cartridge behavior and default settings. If it adds overclocking, optional display processing, HDMI conversion, region switching or BIOS changes, it is already making choices that affect the experience. Those choices may be useful and welcome, but they should not be confused with unmodified historical behavior.
Overclocking is a good example. Some modern reproductions or emulation setups allow the user to reduce slowdown in demanding games. From a comfort perspective, this can be attractive. Many players prefer smoother gameplay. But from a preservation perspective, removing slowdown is a modification. The slowdown is part of how the game behaved on its target hardware. Improving it may create a more enjoyable version of the game, but not necessarily a more faithful one. Original hardware, FPGA, ASIC and software emulation can all offer enhancements. Once enhancements are enabled, the discussion is no longer purely about authenticity.
Video output is another area where simple categories break down. A Neo Geo MVS in an arcade cabinet was typically experienced through a CRT display. A Neo Geo AES might be connected through composite, RGB or other analog paths depending on the region, cable and setup. Today, many players use scalers, HDMI converters, modern televisions or capture devices. These devices can alter scaling, color, sharpness, deinterlacing behavior, latency and aspect ratio. An original console connected through a poor modern video chain may deliver an experience less faithful to the intended display behavior than a carefully configured emulator using a high-quality display and appropriate scaling.
CRT displays are often treated as a fixed reference, and for good reason: they have extremely low display latency and a visual character that modern flat panels do not naturally reproduce. But even here, the full experience is not reduced to the fact that the source is original hardware. CRT geometry, convergence, phosphor behavior, brightness, wear and calibration all vary. A real CRT is not a mathematical constant. It is an analog display device with its own variations. Software CRT shaders or filters do not replace it physically, but they can be part of a controlled and coherent reproduction when the goal is playability and visual approximation rather than object preservation.
Input latency deserves the same level of precision. It is often assumed that an original machine connected to a CRT has instantaneous input, while emulation necessarily adds delay. This is too simple. A CRT avoids the heavy digital processing that can make modern displays feel sluggish, but the game itself still operates in frames. The software does not usually process input in a continuous, infinitely granular way. It reads controller state at a particular point in its execution loop, generally once per frame or according to a frame-based routine.
At roughly 60 Hz, one frame lasts about 16.67 milliseconds, though the exact duration depends on the system’s real refresh rate. If the player presses a button just after the game has polled the controller state, that input will usually not affect game logic until the next polling point. If the player presses the button just before the poll, it may be registered almost immediately in the next logic update. This creates a natural intra-frame variation in input response even on original hardware connected to a CRT. The difference can amount to several milliseconds, and in the worst case can approach the duration of a frame before the game even begins to process the resulting action.
This point does not deny that bad emulation, slow displays or poor controllers can add real and perceptible latency. They can, and they often do. But it does mean that “zero latency” is not how frame-based video games actually work. Even in an ideal original setup, input response is tied to the timing of polling, logic updates, rendering and display scanout. The original machine is not an abstract zero point. It is a real-time system with its own timing structure. A good emulation setup must minimize additional latency, but the baseline itself already contains frame-dependent variation.
For the Neo Geo, this matters because the discussion often treats arcade hardware as an unquestionable reference while holding emulation to a much stricter standard. If a software emulator adds avoidable latency, that should be criticized. But if a small timing variation is inherent to the way the original machine, game loop or display works, then it should be understood rather than used as a blanket argument against emulation. The same applies to refresh rate differences between MVS and AES. A serious discussion must distinguish between errors, implementation choices and legitimate hardware variation.
A well-configured Neo Geo emulation setup should therefore be judged by concrete criteria. Does it use verified ROM dumps? Does it select the correct BIOS for the intended experience? Does it reproduce MVS or AES behavior consistently rather than mixing them casually? Does it handle audio correctly? Does it preserve slowdown and timing unless the user intentionally enables enhancements? Does it output video at an appropriate rate? Does the display chain add avoidable lag? Does the controller introduce unnecessary delay? These are more meaningful questions than whether the word “emulation” appears in the description.
The same criteria should be applied to modern hardware recreations. If a new device claims to reproduce the Neo Geo without software emulation, that may be technically interesting. But the claim itself is only the beginning of the analysis. How are the ASICs implemented? Which hardware revision is targeted? How are AES and MVS differences handled? How is video output generated? Does HDMI output preserve timing accurately? Are analog outputs faithful? What BIOS behavior is used? Are optional enhancements enabled by default? How is input handled? These questions are not hostile. They are simply the technical questions required by any serious claim of accuracy.
Original hardware remains essential. Without it, there is no reference to measure against. Original boards, consoles, cartridges and displays are necessary for preservation, comparison and historical understanding. They also carry a material and emotional value that emulation cannot reproduce. The feel of an arcade cabinet, the weight of an AES cartridge, the stick, the CRT, the physical ritual of inserting a cartridge or powering on a board: these are real aspects of the historical experience. They matter, especially to collectors.
But they are not the same as behavioral accuracy. A cartridge can be original while the display chain is modern and laggy. A console can be original while its video output is modified. A board can be original while components have aged or been replaced. Conversely, an emulator can be non-original as an object but highly accurate as a behavioral model. These are different categories. Confusing them leads to poor arguments.
The Neo Geo shows that demanding accuracy is not the problem. Accuracy is a valid and important goal. The problem is when accuracy becomes a selective standard. If small variations are accepted without concern on original hardware but treated as unacceptable when associated with emulation, the discussion is no longer purely technical. It also reflects cultural preference, nostalgia and attachment to physical media. Those preferences are legitimate, but they should be named honestly.
A balanced position is therefore possible. Emulation should not be defended blindly. Poor emulation exists, and it can damage the experience of a game. But original hardware should not be idealized blindly either. It is not always a single, stable or automatically superior reference. FPGA and ASIC solutions should not be dismissed, but neither should they be treated as magic guarantees. What matters is the quality of the implementation and the coherence of the target.
In the case of the Neo Geo, a carefully configured emulator can provide an excellent experience. For a system as well documented as this one, software emulation can reproduce the essential behavior of the machine with a high degree of confidence. It can also offer stability, accessibility, reproducibility and preservation value. In some situations, a good MVS-focused emulation setup may be more coherent with the behavior of an arcade MVS than a modern hardware solution that blends AES assumptions, alternative BIOS options, display processing and convenience features.
This does not make emulation “better” than original hardware in every context. It simply makes it a serious option. For collecting, original hardware is irreplaceable. For preservation, original hardware and emulation need each other. For playing, the best solution depends on the user’s goal, equipment and tolerance for configuration. For technical accuracy, every solution should be evaluated against a defined target.
The debate should therefore move away from slogans. “Original hardware” is not automatically perfect. “No emulation” is not automatically accurate. “Software emulation” is not automatically inferior. The meaningful distinction is between coherent and incoherent reproductions, measured behavior and assumed authenticity, documented targets and vague nostalgia.
The Neo Geo is a useful reminder that fidelity is not a magical property of an object. It is a technical result. It depends on timing, video, sound, input, BIOS behavior, hardware variation and the entire chain between the software and the player. When those elements are understood and controlled, emulation is not merely a substitute. It is one of the most important tools we have for preserving, studying and experiencing classic video games with seriousness.