Why a Monitor Goes Black When You Raise the Refresh Rate

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Almost certainly nothing is dying. A panel that blinks black on and off when you raise the refresh rate, and stops when you put it back, is not a fault at all. It is a link retraining, over and over, because you have asked for a mode that does not fit down the wire, and the wire's response to a mode it cannot carry is to drop the link, renegotiate, come back, and fail again.

The useful thing about that is it turns into arithmetic you can do before you buy anything. A display mode is a bit rate. Three separate devices each publish a budget for it. And the one that decides whether the picture appears is whichever of the three is smallest, which is almost never the one on the front of the box.

A mode is a number, and you already have all four factors

Everything on a screen arrives as pixels, in order, at a fixed rate. Multiply what the mode demands:

width, times height, times the refresh rate, times three colour channels, times the bits per channel.

Full HD at 60 hertz and 8 bits per channel is 1920 times 1080 times 60 times 24, which is 2.986 gigabits a second. Ordinary 4K at 60 hertz and 8 bits is 11.944. Those two numbers between them cover most of what most people run, and both are small, which is why nobody ever thinks about this.

Now push on any one of the four factors and watch it escalate, because it is a product and not a sum. 4K at 120 hertz and 12 bit colour is 35.832. The same panel at 144 hertz is 42.998. 5K at 120 hertz is 63.701. 8K at 60 hertz is 71.664. Raising the refresh rate from 120 to 144, which is one menu click and feels like nothing, adds 7.166 gigabits a second, which is more than twice the entire demand of the Full HD mode above.

One correction before the comparisons, and it is the only piece of outside knowledge this page needs. Modern HDMI does not send those bits raw. It uses a fixed rate link with 16 bit to 18 bit line coding, so the channel has to carry 1.125 times the payload. USB4 and Thunderbolt tunnel DisplayPort with 128 bit to 132 bit coding, a lighter 1.03125. Neither factor appears in any listing we carry, and both are in the honest limit at the end.

The cable whose name disagrees with its own specification list

Start with the most extreme claim in this catalogue, because it is also the most usefully documented. A short HDMI 2.2 cable rated at 96 gigabits a second publishes, in its performance text, a specific list of modes: 4K at 144 hertz and at 120 hertz, 5K at 120, and 2560 by 1440 at 240, all of them cited at 12 bit colour, plus 8K at 60 and enhanced audio return. It is 3.3 feet long, braided, and carries dynamic HDR, HDCP 2.2 and 2.3, variable refresh rate and auto low latency mode.

Test that list against its own rating, coded:

4K at 144 hertz and 12 bit is 48.373, which is 50.4 percent of 96. The 2560 by 1440 at 240 hertz is 35.832, or 37.3 percent. 5K at 120 is 71.664, or 74.6 percent. 8K at 60 is 80.622, or 84.0 percent. Every single mode in its own specification list fits inside its own number, and the hardest of them uses 84 percent of it. That is not a lucky accident, it is what a specification list looks like when somebody has done this multiplication.

And one of those figures quietly explains why the cable exists at all. 4K at 144 hertz and 12 bit needs 48.373 gigabits a second, which is 0.8 percent more than the 48 gigabits a second of the previous HDMI generation. Not twice as much. Eight tenths of one percent. A single mode sitting a hair over a ceiling is the whole commercial argument for a new cable standard, and nothing about the picture looks any different.

Then read the product's name. It says 16K. 15360 by 8640 at 60 hertz and 8 bit colour is 191.103 gigabits a second of payload, or 214.991 coded, which is 2.24 times the 96 the same cable publishes. At the 12 bit depth the same record claims for every other mode it lists, it is 286.654 raw and 322.486 coded, 3.36 times the rating. Display Stream Compression is specified to roughly three to one, so the 8 bit case is reachable compressed; the 12 bit case at three to one is still 107.5 gigabits a second, which is 1.12 times the cable. So the one resolution in the name is the one resolution that does not fit, and the specification list beneath it is correct throughout.

Its own drawbacks field gets to the right caution for the wrong reason: "Full spec needs HDMI 2.2 on both ends" and "Resolution claims are maker-stated only". Both true, and neither is the problem. The problem with 16K here is not certification or honesty about testing. It is multiplication. The full record is in our review of the 96 gigabit HDMI 2.2 cable, and the category view, which states the principle in words without doing the sum, is how to choose an HDMI cable for 4K and 8K gaming.

The same test on 40 gigabits, where the headline is the only thing that misses

A right angle Thunderbolt 4 cable rated 40 gigabits a second publishes 240 watts over Power Delivery 3.1, a 90 degree connector, quadruple shielding, a zinc alloy shell, over 49,000 bends and 30,000 plug cycles, and on video: a single 8K display, 5K at 60 hertz, 4K at 144 hertz, or two 4K displays at 60 hertz, with daisy chaining.

Work the list at 8 bit, which is the generous reading. 5K at 60 is 21.234, which is 53.1 percent of 40. Two 4K displays at 60 together are 23.888, or 59.7 percent. 4K at 144 is 28.665, or 71.7 percent. Those all fit comfortably, and they still fit after the 1.03125 tunnelling factor, which moves 4K144 to 29.561.

Now the headline. 8K at 60 hertz and 8 bit colour is 47.776 gigabits a second of payload, which is 1.19 times the cable's entire rating before any coding is counted at all, and 49.269 with it. There is no reading of 40 that contains it uncompressed. What does fit is 8K at 30 hertz, 23.888, which is 59.7 percent of the budget and exactly the same demand as the two 4K displays the record lists beside it. So "a single 8K" on this cable means either half the refresh rate of everything else in the sentence, or compression, and the record does not say which.

What its drawbacks field does say is the set of conditions that genuinely apply: "Full 40Gbps speed needs a Thunderbolt 4 device", "Video output requires DP Alt Mode support", "Dual-monitor setups need two cables". All three are real and all three are about the devices at the ends rather than the pipe. Our review of the right angle Thunderbolt 4 cable has the rest, and how to choose a Thunderbolt 4 cable for your setup covers the compatibility side.

The dongle that explains its own 30 hertz and gets no credit for it

This is the best of the three records and the one nobody would guess was doing arithmetic. A three in one USB-C multiport adapter with 4K HDMI, USB 3.0 and passthrough charging volunteers, in its own performance text, that "single-cable adapters commonly output 4K at a 30Hz refresh rate. That is fine for video but can feel less smooth for desktop mouse work." It is right, and the reason is sitting two sentences away.

A USB-C connector has four high speed lane pairs and no more. The adapter's own feature list needs some of them for a USB 3.0 port it rates at 5 gigabits a second. Give two lanes to USB and two to DisplayPort at the common HBR2 rate, and the video budget is two times 5.4 gigabits a second times the 0.8 that 8 bit to 10 bit coding leaves, which is 8.64 gigabits a second.

4K at 30 hertz and 8 bit is 5.972, which is 69.1 percent of 8.64. It fits, with room. 4K at 60 hertz is 11.944, which is 1.38 times 8.64. It does not, and no firmware or certification changes that. So the 30 hertz is not a compromise the maker chose. It is the USB port sold on the same dongle taking half the lanes, and the record describes the symptom one paragraph after describing the cause.

There is a pleasant symmetry worth noticing. Four lanes of HBR2 give 17.28 gigabits a second, and 4K at 60 hertz against that is also 69.1 percent. The mode you can have on two lanes at half the refresh rate, and the mode you can have on four lanes at full refresh rate, sit at exactly the same fraction of their budgets, because halving the refresh rate and halving the lanes are the same division. Details in our review of the USB-C multiport adapter, and the wider version of this trap is USB-C is a plug, not a promise, which is about the same four lanes being asked to do several jobs at once.

Why the symptom is a blink and not an error message

The behaviour in the question follows from all of this without any extra physics. When a source offers a mode, the sink answers with what it can accept, the link trains at a rate, and the picture appears. Ask for a rate above what the weakest element in the chain can carry and the training either fails or succeeds marginally and then loses lock. A failed training is a black screen. A marginal link that loses lock and retrains is a black screen that comes and goes, which is the exact complaint, and it is also why the problem is worse when the cable warms up, worse on a longer run, and gone the moment you step the refresh rate back down.

It also explains the thing that confuses people most, which is that the panel, the graphics card and the cable can each be individually capable of the mode and the combination still fail. The budget is not the average of the three. It is the minimum, and a dongle in the middle of an otherwise excellent chain sets it at 8.64.

The practical order to check, then, is the order of the three budgets above, smallest first. Take the adapter out of the path. Shorten the run. Drop the colour depth from 12 bit to 10 or 8, which cuts the demand by a quarter or a third for a difference most people cannot see on a desktop. Or drop the refresh rate, which is the lever the question already found by accident.

And the projector version of the same confusion has its own page, because projectors publish two resolutions on purpose: how to choose a projector is right that "a native 1080p projector that supports 4K does not show a 4K image", and the arithmetic here is what that sentence is made of. A device's supported resolution is a statement about the bit rate it will accept at the input. Its native resolution is a statement about the pixels it owns. Those are different quantities, and only one of them is on the front of a cable.

The honest limit, and what this page refuses to compute

What is exact: every payload figure in the table. Each is one multiplication of four published integers, and the ratios between them, including the 2.24 and the 3.36 and the 1.38, are exact in the same way. The percentages against 96, 40 and 8.64 are exact given those three budgets, and all three budgets are published by the products themselves.

What is representative rather than published, and it is a short list that matters. The 1.125 coding factor for fixed rate HDMI links and the 1.03125 for USB4 tunnelling are standard but appear in no record we carry; drop both and every coded figure falls by those amounts while every ratio between modes is untouched. Blanking intervals are ignored entirely, which is the biggest soft edge here: real timings add roughly a tenth to a quarter on top for standard formats and much less with reduced blanking, so every demand above is a floor and the true margins are tighter than stated, never looser. The three to one ceiling on Display Stream Compression is the specification's. And the two lane split on the dongle is inferred from the fact that its own USB 3.0 port must be fed from somewhere, not stated anywhere in its listing.

Three things this page deliberately does not compute. It does not predict which mode a given link will actually negotiate, because that depends on the identification data the monitor publishes to the source and on whether both ends implement compression, and not one record we carry states either. It does not use chroma subsampling, which cuts the payload by a third at 4:2:2 and a half at 4:2:0, because no listing says whether a mode it advertises assumes it, and presenting a subsampled mode as if it were full colour is precisely the move this page is about; naming the gap is more useful than filling it with an assumption that flatters the product. And it does not attempt the audio side of the same budget, which the first cable's enhanced return channel also has to fit, because lossless surround formats are published as names rather than rates.

Three numbers to carry. A display mode is width times height times refresh times three times the bits, so 4K at 60 and 8 bit is 11.944 gigabits a second and 8K at 60 and 12 bit is 71.664. A link budget is the minimum of the source, the cable and the sink, never the average, which is why one dongle at 8.64 decides a chain full of 40s. And when a resolution appears in a product's name but not in its own specification list, do the multiplication, because on the cable above the gap between the two was a factor of 2.24.