How Input Lag Works in a Retro Gaming Signal Chain: Frames, Milliseconds, and Scaler Latency

Start with a physical constant: at 60Hz, one frame is 16.67ms. That number is not a setting you can tweak. It is the beat the entire signal chain dances to. Every console that outputs an analog signal sits at the source with zero frames of video latency because the DAC does not buffer. It pushes the picture out. The lag you feel comes from two places you actually control: the device sitting between the console and the TV, and the screen's own processing. That device must convert a 240p or 480i signal into something your 4K OLED understands. How it does that, line multiplication with no framebuffer versus full-frame capture and resampling, is the difference between imperceptible delay and a 30ms smeared mess. That 30ms is two frames. At that point, your Super Famicom is unplayable because the hardware added a felt physical delay between pressing jump and the sprite moving. Know which type of lag lives where, and which of those types a dedicated scaler can fix.

Frame time 16.67ms retro consoles is the baseline from which every measurement here is drawn. A single frame of buffering is the difference between zero perceptible input lag and lag you can feel in a platformer. When you press a button on a SNES, the game state updates at the next vertical blank, and the console's video output latency is zero because the picture is generated live. The scaler can add anywhere from a few microseconds to several frames. That variance is not academic. A device that captures a full frame into memory before scaling, the cheap composite-to-HDMI box from a street stall, adds at minimum 16.67ms before it even starts work. Add the screen's own 16ms of processing if you forgot Game Mode, and you are at 33ms, two full frames of delay. A frame at 60Hz is 16.67ms. That is your yardstick. If a device says it adds half a frame, you know that is 8.3ms. If a TV review quotes 9.4ms in Game Mode, you know that is just over half a frame. The entire subject is a matter of counting how many 16.67ms increments stand between your button press and the photon leaving the screen.

Scaler added latency explained simply: a unit takes a signal in one format and outputs it in another. The 240p signal from a SNES is 240 lines tall. Your 4K TV has far more lines. Somewhere between those two numbers, the device must generate the missing pixels. How it does that determines the lag. The OSSC, designed by hobbyists and documented by the Shmups Forum, uses line multiplication: it repeats each input line the correct number of times to reach the output resolution, and it does this on the fly, scanline by scanline. Because it has no framebuffer, it never holds a full frame, and its latency is measured in microseconds. A single scanline at 60Hz is about 16.7 microseconds. The 5X-Pro uses a framebuffer but processes sub-100-microsecond per line, and Mike Chi's design philosophy explicitly targets lag-free operation. A cheap box takes the entire frame, stores it in a memory buffer, scales it, and only then outputs it. That full-frame capture is the difference between adding 0.1ms and adding 17ms. The job is to make your old console work on a new TV. The method defines whether you feel it.

Display lag vs processing lag is the distinction that separates a fixable problem from a hardware limitation. Processing lag is what the scaler adds. You can reduce it by buying a better unit. Display lag is what the TV adds after the scaler has done its job. That is the see-saw: a 5X-Pro adds 0.25ms to 0.5ms, depending on whether you use framelock or triple-buffer mode, but your LG C3 OLED in Game Mode still adds 9.3ms at 60Hz. The scaler is irrelevant if the TV is not in Game Mode. A modern OLED in Game Mode at 120Hz sits at 5.1ms on the LG C3, but that is after the scaler. The total system lag is source (0ms for original hardware) plus scaler (0.25ms to 8ms) plus screen (5ms to 18ms). A cheap box adds two frames of processing lag, but a good TV in Game Mode adds only half a frame of display lag. The failure mode is when people blame the TV when the scaler is the problem, or vice versa. If you have a good scaler but a TV with motion interpolation on, you are stuck. If you have a cheap box and a TV in Game Mode, the scaler is the bottleneck. Know which one is biting you.

Why cheap adapters add lag frames is a story of design priorities. The generic composite-to-HDMI box from a marketplace seller uses a chip designed for DVD players, not game consoles. That chip's job is to take a 480i video signal, deinterlace it, scale it, and output it over HDMI, and it does all of that by capturing entire frames into a 16MB memory buffer. A frame at 480i is about 0.5MB, so the chip stores multiple frames to make the deinterlacing and scaling easier. Every full-frame capture adds at least 16.67ms. The chip also runs motion-adaptive deinterlacing, which needs to see a full frame before and after the current one to interpolate motion, adding another frame of latency. That is 2 to 4 frames of lag before the signal even leaves the box. A dedicated scaler like the 5X-Pro or OSSC uses a field-programmable gate array (FPGA) that processes lines, not frames, so it never touches a full-frame buffer. The difference is not the price tag. It is the architecture. A line multiplier adds microseconds. A framebuffer scaler adds milliseconds. The cheap one adds both because it was never designed for this job.

Frame time 16.67ms constant is the number you use to translate every spec sheet into a feel. When a scaler says it adds less than a frame, you do not know if that is 1ms or 16ms. But you know it is not more than 16.67ms, because a frame is a frame. The constant also anchors the 50Hz vs 60Hz difference: a 50Hz signal has a frame time of 20ms, 3.33ms longer per frame. That is why a PAL-optimized game on a 60Hz screen feels slightly more sluggish. The frame you are waiting for arrives later. The constant also demystifies the measurements you see on display review sites. RTINGS measures the LG C3 OLED at 9.3ms in Game Mode at 60Hz, and 5.1ms at 120Hz. That is 0.56 frames and 0.31 frames, respectively. A frame at 120Hz is 8.33ms, so the screen is adding just over half a frame at 60Hz and just over half a frame at 120Hz. In all cases, the scaler's contribution is either a rounding error (0.25ms is 1.5% of a frame) or a disaster (33.3ms is two full frames). The constant is your sanity check. If someone tells you a scaler adds one frame at 60Hz, you know that is 16.67ms.

A framebuffer full-frame capture is the single most expensive operation in the retro signal chain. The difference between a 5X-Pro and a generic adapter is exactly this. The 5X-Pro, in its default framelock mode, does not capture a full frame at all. It samples lines as they arrive and outputs them scaled, achieving 0.25ms to 0.5ms latency. In triple-buffer mode, used for features like scanline filters and HDR tone mapping, it captures a full frame, and latency rises to 4ms to 8ms. That is still under half a frame. A generic adapter also captures a full frame, but it does so because its chip cannot scale without the entire image in memory, and it does so at 240p, holding the frame for the full 16.67ms before releasing it. The result is that the game moves in discrete steps: press a button, wait a frame, the sprite does not move, wait another frame, it moves. The capture is also why cheap boxes cannot do 480p. They only have memory for 480i frames. The framebuffer is not inherently bad. The 4K model uses one to enable HDR and black frame insertion, but it is the price of those features, and you must know when you are paying it.

Line multiplication microseconds is the other end of the spectrum from the framebuffer. The OSSC is the purest expression of this: it does not capture, buffer, or deinterlace. It counts lines. An SNES 240p signal has 240 lines. The OSSC repeats each line twice for 480p, three times for a higher resolution, and so on, and it does this in real time as each line arrives. There is no point at which the OSSC holds a complete frame, so its latency is measured in the time it takes for a single line to arrive: at 60Hz, one line is about 16.7 microseconds, and the OSSC adds a couple of scanlines at most. That is 0.03ms. The Shmups Forum developers measured this directly with a photodiode taped to a CRT: the video signal triggers the photodiode, and the time to screen luminance change is under 1ms in all modes. The trade-off is that line multiplication cannot create new information. It repeats what is there. You get razor-sharp pixels but no composite or S-Video input, no deinterlacing, and no scanline filters. If you want composite video from a NES, the OSSC is not your device. You need a 5X. But for pure, measurable, sub-millisecond lag, the line multiplier is the gold standard.

Motion-adaptive deinterlacing analysis window is the hidden lag trap in modern scalers. A 480i signal from a PlayStation 2 or GameCube is interlaced: it draws every other line on each pass, alternating between odd and even fields 60 times a second. To display it progressively, the scaler must combine two fields into one frame. A simple bob deinterlacer doubles each line and does not buffer. It adds zero lag but causes flicker and fine details shimmer. A motion-adaptive deinterlacer analyzes each pixel across multiple fields to determine whether the picture is moving, and if so, it weaves the two fields together for static areas but bobs the moving ones. That analysis window is the problem: the scaler must look at the current field and the previous one, which means it holds at least one field in memory. The GBS-Control, a community board, does this and exhibits chroma artifacts on fast-moving red objects. The analysis window is too small to get it right. The 5X-Pro does it with a window of a few milliseconds, and it is nearly imperceptible. A cheap scaler's motion-adaptive deinterlacing adds 16.67ms because it waits for the next field to make a decision. The fix is to know which deinterlacer you have and whether it is worth the lag.

HDMI handshake negotiation time is the lag that exists even when the scaler is not processing anything. Every time you turn on your console, the scaler and the TV must agree on a resolution, refresh rate, and color space. This happens over the HDMI control data channel, and it involves an EDID metadata exchange. The TV sends its capabilities, the scaler sends back a video format. On a modern scaler with a fast EDID, this is a one-time event that takes a few milliseconds. But on a cheap adapter, the handshake can fail, and the two devices enter a re-negotiation loop. The symptom is a blinking black screen every few seconds, and the cause is often a capture card in the chain that reports an incorrect EDID. The 5X-Pro and OSSC let you disable HDCP and choose a fixed output resolution, which eliminates the most common handshake failures. The failure mode is not the handshake itself. It is the loop. If your screen blinks, do not blame the scaler. Check the EDID metadata from the display and the capture card. This is also why the mClassic, a dongle that sits between the console and the TV, can cause issues: it has no power of its own and must negotiate with both ends, and if the TV's EDID is quirky, the dongle forwards the wrong data.

Time Sleuth latency measurement device from Leo Bodnar is the gold standard for scaler testing, and its numbers settle arguments. The Time Sleuth sends a test pattern over HDMI, detects when the video signal changes, and measures the time to a luminance change on the screen. It is a photodiode-based system, and it is how RTINGS measures the LG C3 at 9.3ms in Game Mode and how the Shmups Forum measured the 5X-Pro at 0.25ms in framelock mode. The mClassic's claimed 2ms latency comes from a Time Sleuth test published by independent testers. It is not a marketing figure, but it is also not zero, despite what the box implies. The OSSC's sub-millisecond latency was confirmed by the same method. The takeaway: when a scaler vendor says zero lag, ask for the Time Sleuth graph. If they cannot show one, they are guessing. A Leo Bodnar unit costs more than most scalers, but it is the only way to know for sure. The 5X-Pro has a built-in test pattern generator that works with an oscilloscope, but for a quick consumer check, a Leo Bodnar is the reference.

The 5X-Pro sub-millisecond processing is the number that makes it the default recommendation. In framelock mode, the 5X-Pro adds 0.25ms to 0.5ms, less than 1/60th of a frame, imperceptible to any human. In triple-buffer mode, it adds 4ms to 8ms, which is still under half a frame and still fine for 99% of players. The 5X-Pro accepts composite, S-Video, component, and RGB SCART, and its menu system lets you save profiles for each console. It also has a scanline filter and a CRT simulation mode that is genuinely good on an OLED. The key is that Mike Chi designed it with a specific philosophy: no processing step may ever add more than a frame of lag. The 4K model goes further, adding HDR, 120Hz black frame insertion, and a higher resolution, but at 4K output, the framebuffer cost is higher, and it sits at 4-8ms in triple-buffer mode. The 5X-Pro is the sweet spot: it handles every analog signal type, and its lag is a rounding error. For the competitive player, the 5X-Pro in framelock mode is indistinguishable from a direct connection. Check the maker's site for current pricing.

OSSC scanline-level latency is what you get when you refuse to touch a framebuffer. The OSSC is a line multiplier. It has no memory for a full frame, so its latency is bounded by the time it takes to process a single scanline. At 60Hz, a line lasts about 16.7 microseconds, and the OSSC adds a few lines of pipeline delay for a total under 1ms in all output modes. This is not a marketing claim. It is a measurement taken by the Shmups Forum developers using a photodiode on a CRT. The trade-off is that the OSSC accepts no composite or S-Video. You must use component or RGB SCART. It also has no scanline filters, no HDR, and no motion-adaptive deinterlacing. It is a pure, uncompromising signal path. If you are a hardcore speedrunner or a rhythm game player, the OSSC is the lowest-latency device you can buy. But if you need composite video from a NES or you want to play PS2 games with deinterlacing, the OSSC is the wrong tool. The OSSC Pro adds a framebuffer for those features, but it also adds lag when you use them. The OSSC is the scaler for people who count microseconds, and the Shmups Forum is where they publish their measurements.

mClassic Time Sleuth 2ms measurement is the number that separates marketing from reality. The mClassic is a dongle from Marseille that plugs into an HDMI port, applies context-adaptive anti-aliasing, and upscales to 1440p. The marketing says zero lag, but the box is wrong. Independent testers using a Time Sleuth measured 1-2ms of processing latency. Not zero, but close enough that you will never feel it. The mClassic is a passthrough device: if you do not want the processing, it has a physical button to disable it, and in passthrough mode, the lag is 0ms because the signal is not touched. The 2ms figure is for when the anti-aliasing is active, and it is the cost of the analysis window that decides which edges to smooth. The mClassic is not a scaler for 240p. It is a scaler for 480p and 720p signals, and it is best used on modern consoles. For retro gaming, the mClassic's 2ms is excellent, but it is not a 5X-Pro. The 5X-Pro is 0.5ms and accepts analog signals. The mClassic only accepts HDMI. If you have a console with native HDMI, the mClassic is a good low-cost option. If you have a SNES, you need a 5X-Pro.

Generic composite-to-HDMI 2-4 frame buffer is the failure mode that sends people to forums asking why their games feel like molasses. A generic adapter takes a composite or component signal, deinterlaces it, scales it, and outputs HDMI, and it does so by storing 2 to 4 frames in a buffer. At 60Hz, that is 33ms to 66ms of added latency. The chip inside is designed for DVD playback, where delay does not matter because the video is pre-recorded. In a game, the lag is catastrophic: you press jump, and the character jumps a quarter of a second later. The fix is to throw the adapter away. A 5X-Pro or OSSC costs more, but they solve the problem at the architectural level. A cheaper fix is to use a CRT for 240p content, which has near-zero lag, but that is not practical for most people. The generic adapter is the reason the retro gaming community buys a dedicated scaler. It is not a luxury. It is a necessity. If you are using a cheap box and wondering why your rhythm game is unplayable, this is why.

Sample-and-hold pixel response is the display-side lag that no scaler can fix. Modern OLED and LED panels are sample-and-hold devices: they show each frame for the full refresh period, then hold it until the next frame arrives. A CRT draws each line and then lets it decay, which is why motion looks different. The sample-and-hold effect is why a 60Hz game on an OLED looks blurrier than on a CRT, even when the input lag is low. The fix is black frame insertion (BFI), which inserts a black frame between each game frame to simulate a CRT's flicker. The 4K model supports 120Hz BFI, which eliminates the sample-and-hold blur without visible flicker. A scaler that supports BFI can reduce perceived motion blur, but it cannot reduce the screen's inherent latency. A TV with a 120Hz panel and BFI at 60Hz input shows each game frame for two refresh cycles, but the black frame in between means the image is off for 8.33ms, which the eye perceives as crispness. You cannot fix sample-and-hold by buying a scaler. You fix it by buying a TV with BFI or a high refresh rate. The scaler enables it, but the panel is the final arbiter.

Black frame insertion persistence reduction is the scaler feature that proves the screen matters more than the scaler. BFI works by flashing each frame once, then turning the backlight off for the rest of the refresh period. On a 120Hz display, a 60Hz source shows each game frame twice, and BFI inserts a black frame between the two. The result is that each game frame is only visible for 8.33ms, which is the same persistence as a CRT. The 4K model supports 120Hz BFI, and the result on an OLED is remarkable. The trade-off is that BFI can cause flicker at 60Hz, and it reduces brightness by half. Some people are sensitive to the flicker and prefer to leave BFI off. The key is that BFI reduces persistence, not input lag. The input lag is still the scaler's 4ms plus the screen's 5ms. BFI only affects how long each frame stays on screen. If you are playing a rhythm game, BFI will not make it more responsive. It will make the motion look cleaner. The scaler's BFI feature is a display enhancement, not a latency feature. Do not confuse the two.

Shmups Forum lag testing methodology is the community standard that turned a hardware hobby into a science. The forum's members, including Mike Chi, the inventor of the 5X-Pro, developed the photodiode testing method that measures the time from a video signal trigger to a luminance change on the screen. They published their results in a dedicated thread, and those numbers became the reference for every scaler review. The methodology is simple: tape a photodiode to the center of the screen, send a test signal, and measure the delay with an oscilloscope. This is why the OSSC's sub-1ms figure is trusted. It was measured, not claimed. The forum also developed the 240p test suite, a collection of test patterns for calibrating screens and scalers. If you want to test your own setup, the test suite is free to download and runs on a Raspberry Pi or a flash cart. The Shmups Forum is the reason we know the mClassic adds 2ms and the 5X-Pro adds 0.25ms. Without them, we would be at the mercy of marketing departments.

Mike Chi lag-free design philosophy is the guiding principle behind every 5X-Pro product, and it is a stance, not a marketing slogan. Chi's stated rule is that no processing step may add more than one frame of lag, and he publishes the latency numbers for each mode in the product manual. The 5X-Pro's framelock mode is designed to be under 0.5ms. The triple-buffer mode is for when you need scanline filters, and it is still under 8ms. The philosophy extends to the video path: the unit is designed to pass through the audio and video without re-clocking, and it uses a single clock domain for the input and output to avoid frame rate conversion. This is why the 5X-Pro is the reference, not the OSSC or the mClassic. The OSSC is also lag-free but lacks features. The mClassic is a good product but is hamstrung by its HDMI-only input. Chi's philosophy is that lag is a bug, not a feature, and he is willing to sacrifice convenience to preserve it. When you buy a 5X-Pro, you are buying a person's conviction, not just a circuit board.

Choose the Right Scaler for Your Console

If you have a 4K OLED and a SNES, buy a 5X-Pro and set it to framelock mode. Leave it there. Do not touch the scanline filters unless you are playing a game that you know benefits from them. The triple-buffer mode adds 4ms, which is fine, but framelock is better. If you are on a budget, buy an OSSC for your component or RGB cables and accept that you cannot play composite video through it. If you are playing a PS2, use a 5X-Pro with motion-adaptive deinterlacing. The OSSC's bob deinterlacing is too flickery.

Set Up Your Capture Chain Correctly

If you are streaming, split the signal with a powered HDMI splitter, not a passive one, and set your capture card to 1080p to avoid the EDID mismatch that causes black-screen loops.

Fix the Most Common Failure

The failure case is 1am when your screen starts blinking from an EDID loop: unplug the HDMI cable from the capture card, turn off the TV, and turn it back on. If that does not fix it, reset the scaler to factory defaults. The 5X-Pro's menu has a Reset EDID option under the HDMI settings. Use it.

Device	Type	Input Lag (measured)	Method	Best For
5X-Pro	Framebuffer scaler	0.25-0.5 ms (framelock), 4-8 ms (triple buffer)	Time Sleuth	All analog consoles, composite to RGB
4K	Framebuffer scaler	4-8 ms (triple buffer)	Time Sleuth	4K OLED with HDR and 120Hz BFI
OSSC	Line multiplier	<1 ms (all modes)	Photodiode	Component/RGB SCART, zero-lag purists
OSSC Pro	Line multiplier + framebuffer	<1 ms (line mult), 4-8 ms (framebuffer features)	Photodiode	Modern and retro HDMI, adaptive line mult
mClassic	HDMI dongle	2 ms (processing), 0 ms (passthrough)	Time Sleuth	480p/720p HDMI sources, cheap anti-aliasing
GBS-Control	Budget scaler	~1 frame (variable)	Shmups Forum	DIY projects, composite/S-Video, low cost
Generic composite-to-HDMI	Framebuffer scaler	33-66 ms (2-4 frames)	Not recommended	Nothing; replace it

Frequently Asked Questions

Q: Does a scaler add input lag if I use a CRT?

A: Yes, but it is negligible. A line multiplier like the OSSC adds under 1ms on a CRT, which is imperceptible. A framebuffer scaler adds 0.25-8ms, which is still under half a frame. The CRT itself has near-zero lag, so the total is fine.

Q: Can I reduce input lag by turning off the scaler's scanline filter?

A: On the 5X-Pro, yes. Scanline filters require triple-buffer mode, which adds 4-8ms. In framelock mode, scanlines are not available, and the lag drops to 0.25ms. On the OSSC, scanlines are generated on the fly and add no lag.

Q: Is the mClassic worth it for a SNES over a 5X-Pro?

A: No. The mClassic only accepts HDMI, so you would need a separate analog-to-HDMI converter first, which adds lag. The 5X-Pro accepts the SNES's native RGB output and adds less lag. The mClassic is for modern consoles with HDMI out.

Q: Why does my screen blink when I enable 1440p on the scaler?

A: The TV is likely not reporting a 1440p EDID correctly, or the HDMI cable is too old. Use a high-speed cable rated for 18Gbps, and set the output to 1080p or 4K instead. The blinking is a handshake loop, not a scaler fault.