Part VIIVideo

Cards and Data Rates for Video

October 1, 202620 min readFirmware 2.00

How many megabits a second each video setting writes, how that becomes gigabytes a minute, which card ratings matter for video, and how long the author's cards last.

A school recital runs two hours, and the camera is set to H.265 10-bit at 3840 by 2160 pixels and 30 frames a second. Nikon rates that setting at about 190 megabits per second, which turns into 23.75 megabytes a second, 1.425 gigabytes a minute and 171 gigabytes for the evening. The 512 GB card in the CFexpress slot takes all of it with room to spare. The 128 GB spare holds about 90 minutes by the label, or about 74 by the cautious ratio I derive below.

This chapter builds that sum from bits upward, explains card speed markings and Destination, the menu item that picks the card for video, and ends with minutes per card.

Page numbers in parentheses point to Nikon's Z6III Reference Guide, edition 09, which matches firmware 2.00.

01.Bits, Bytes and Bit Rates

A bit is a single 0 or 1. A byte is eight bits, enough to hold one of 28=2562^8 = 256 values, and files and cards are measured in bytes. Mega means a million, so a megabit is a million bits, a megabyte (MB) is a million bytes, and a gigabyte (GB) is a thousand megabytes. That decimal convention is the card makers' own, and Card Slots and File Handling sets it beside the computer's habit of counting in powers of two.

Video is a flow, so Nikon gives each setting a bit rate in megabits per second (Mbps), the number of megabits written for each second of footage (p. 205). Card speeds come in megabytes per second (MB/s). The two differ by a factor of eight, and mixing them up is the commonest slip in video arithmetic.

Take H.265 10-bit at 3840 by 2160 and 30p, rated at 190 Mbps (p. 207). Nikon's 30p means 29.97 frames per second, written fps (pp. 201-202). Eight bits make a byte, so

190 Mbps÷8=23.75 MB/s.190 \text{ Mbps} \div 8 = 23.75 \text{ MB/s}.

A minute has 60 seconds, so a minute of that footage takes 23.75×60=1,42523.75 \times 60 = 1{,}425 MB, which is 1.425 GB, and an hour takes 1.425×60=85.51.425 \times 60 = 85.5 GB. The two-hour recital takes 1.425×120=1711.425 \times 120 = 171 GB, and a 512 GB card holds 512÷1.425≈359.3512 \div 1.425 \approx 359.3 minutes, just under six hours.

The same steps work for any setting. Write rr for the bit rate in Mbps and CC for the card's capacity in GB. A minute of footage takes 60r/8=7.5r60r/8 = 7.5r MB, which is 0.0075r0.0075r GB, so the card holds

t=C0.0075 r=400 C3 r minutes.t = \frac{C}{0.0075\,r} = \frac{400\,C}{3\,r} \text{ minutes}.

Checking it against the example, 400×512÷(3×190)=204,800÷570≈359.3400 \times 512 \div (3 \times 190) = 204{,}800 \div 570 \approx 359.3 minutes, the same answer.

190 Mbps becomes 23.75 MB/s, 1,425 MB and then 1.425 GB per minute. A 512 GB card divided by 1.425 gives 359 minutes, and times 0.82 gives 295 minutes to plan with.
Figure 1. From Nikon's bit rate to minutes on the author's main card, for H.265 10-bit at 3840 by 2160 and 30p.

The last step in the figure, the factor of 0.82, is my own planning margin, estimated from Nikon's capacity table, and is explained below.

02.What the Camera Writes Each Second

Nikon prints its bit rates as approximate averages, which depend on the Video file type and, for raw video, on Video quality (N-RAW) (p. 205). An average suits capacity sums, because over a long recording a card fills at the average rate. Nikon does not say whether these rates include sound, which MOV files store uncompressed, 48,000 samples a second of 24 bits on each of two channels from the built-in stereo microphone (pp. 1004-1005), about 2.3 Mbps and too little to move these sums. Video Formats and Codecs explains each file type. The one idea needed here is that H.265 and H.264 store mostly the changes from one frame to the next, which makes them far smaller than raw video, which keeps the sensor's own data for every frame.

Table 1. Average bit rates in Mbps for the compressed file types (pp. 207-208)

SizeFrame rateH.265, 10 bitsH.265, 8 bitsH.264
5376 by 302460p, 50p400370None
5376 by 302430p, 25p, 24p340300None
3840 by 2160120p, 100p400370None
3840 by 216060p, 50p340300None
3840 by 216030p, 25p, 24p190150None
1920 by 1080240p, 200p340300None
1920 by 1080120p, 100p190150None
1920 by 108060p, 50p1008050
1920 by 108030p, 25p, 24p504030
1920 by 1080Slow motionNoneNone30

Nikon calls 3840 by 2160 4K UHD and 5376 by 3024 5.4K (p. 202), shortened here to 4K and 5.4K. The 10-bit files cost more than the 8-bit ones at every size, about 27 percent more at 4K and 30p, since 190÷150≈1.27190 \div 150 \approx 1.27. Doubling the frame rate need not double the bit rate. At 1080 it does, 50 Mbps to 100 in H.265 10-bit, but 4K goes from 190 at 30p to 340 at 60p, a factor of 340÷190≈1.79340 \div 190 \approx 1.79.

Raw video sits in another league. Beside each raw clip the camera also records an H.264 8-bit MP4 copy at 1920 by 1080 for playback on the camera (p. 197), and its rate is in the last column.

Table 2. Average N-RAW bit rates in Mbps at 120p, 60p and 30p, with the playback copy (pp. 205-207)

Frame sizeRateHigh qualityNormalMP4 copy
FX 6048 by 340260p3730187050
FX 6048 by 340230p187094030
FX 4032 by 226860p164083050
FX 4032 by 226830p82042030
DX 3984 by 2240120p32401630100
DX 3984 by 224060p162082050
DX 3984 by 224030p81041030

At 100p, 50p, 25p and 24p the raw rates are lower in proportion, 3110 Mbps in place of 3730 at the top setting for instance, and the RG lists every one. FX means the full width of the sensor and DX a smaller crop from its centre, as Image Area and Crop explains. The author does not record raw video, so this table is here for scale. Its top setting writes 3730÷190≈19.63730 \div 190 \approx 19.6 times as many bits as 4K at 30p in H.265 10-bit, and would fill a 512 GB card in 400×512÷(3×3730)≈18400 \times 512 \div (3 \times 3730) \approx 18 minutes, still 18 with the 50 Mbps MP4 copy added.

For ProRes, Nikon prints no figures and points to Apple's published specifications (p. 205). Apple's ProRes white paper gives ProRes 422 HQ target rates at 3840 by 2160 of 707 Mb/s at 24p, 884 at 30p and 1,768 at 60p, Mb/s being Apple's way of writing Mbps. Treating those as the camera's rates is my inference, since Nikon does not say its files meet Apple's targets. Apple has no row for 5376 by 3024 and none above 60p at 1920 by 1080, where the camera offers 120p and 100p (pp. 201-202).

The Main Choices in Card Units

Each row takes the rate in Mbps, divides by 8 for MB/s, then multiplies by 60 and divides by 1,000 for gigabytes a minute.

Table 3. Main video settings converted to card units, in decimal megabytes and gigabytes

SettingMbpsMB/sGB a minute
H.264, 1080 at 30p303.750.225
H.265 10-bit, 4K at 30p19023.751.425
H.265 10-bit, 4K at 60p34042.52.55
H.265 10-bit, 4K at 120p400503.0
ProRes 422 HQ, 4K at 30p (Apple)884110.56.63
ProRes 422 HQ, 4K at 60p (Apple)176822113.26
N-RAW High quality, FX 6048 at 60p3730466.2527.98

The conversion also runs backwards, which is how card ratings meet bit rates. Nikon's 250 MB/s recommendation, which comes up below, is 250×8=2,000250 \times 8 = 2{,}000 Mbps, and the 90 MB/s floor of a V90 card is 90×8=72090 \times 8 = 720 Mbps.

03.Card Speed Ratings

Headline Read and Sustained Write

A card has two speeds. Read speed is how fast data comes off it, which matters when copying footage to a computer. Write speed is how fast data goes on, and it is the one that matters while recording.

The big number on the packaging is a peak, the best rate the maker measured on its own equipment. Video needs something else. A recording is a stream that never pauses, so what counts is the lowest rate the card holds for minutes on end, its sustained write speed. The SD Association, which writes the SD card standards, explains that video needs a constant minimum write speed so that no frames are dropped.

Table 4. The author's cards as their makers rate them

CardRead, up toWrite, up toFloor for video
Lexar SILVER CFexpress 4.0, 512 GB3,600 MB/s3,000 MB/s2,600 MB/s sustained, Lexar's testing
SanDisk Extreme PRO CFexpress, 128 GB1,700 MB/s1,200 MB/sNone stated
Lexar 2000x SDXC UHS-II, 256 GB300 MB/s260 MB/sV90, so 90 MB/s

The figures come from the makers' own pages: Lexar's for the Professional SILVER CFexpress 4.0 Type B, SanDisk's rating for the 128 GB Extreme PRO CFexpress Type B, since SanDisk rates each capacity separately, and Lexar's for the Professional 2000x SDXC UHS-II GOLD Series, which also marks the card V90. All three base their speeds on internal testing, and the SILVER and SanDisk pages add that real speed depends on the host device. Every figure in the table, V90 included, is a maker's claim, and none was measured for this guide.

CFexpress Versions and VPG

CFexpress is a card standard from the CompactFlash Association (CFA), released as 1.0 in 2017, 2.0 in 2019 and 4.0 in 2023, with no version 3.0. Type B, the size the Z6III takes (p. 1011), uses two PCIe lanes. PCIe is the high-speed link a computer uses for its internal drives, and a lane is two pairs of wires, one sending and one receiving. A small 4 on the CFexpress logo marks a 4.0 card.

A newer card runs only as fast as the slot allows. Nikon's raw-video card list footnotes a CFexpress 4.0 card as running no faster than the 2.0 standard allows (p. 1011), and Card Slots and File Handling explains why the author reads that cap as covering the Lexar SILVER too. For H.265 the cap does not bite: the heaviest H.265 setting averages 50 MB/s, a fifth of Nikon's 250 MB/s bar (p. 1011).

The CFA also certifies cards to its Video Performance Guarantee (VPG), a minimum sustained write speed, which it defines as the rate at which data can be written continuously without dropping frames. The class names carry the number: VPG200, VPG400, VPG800 and VPG1600 guarantee at least 200, 400, 800 and 1,600 MB/s.

The RG states no VPG requirement for the Z6III, and VPG400 appears on p. 1011 only inside one Nextorage card's name. Lexar lists no VPG class for the SILVER, so its only sustained-write promise is Lexar's own 2,600 MB/s, and SanDisk publishes no sustained-write or VPG figure for the 128 GB Extreme PRO.

SD Speed Classes

SD cards carry their write floor as a symbol. The SD Association defines four families of speed mark, three of which appear on the UHS-I and UHS-II cards this camera supports (p. 1011): Speed Class, a C around 2, 4, 6 or 10, UHS Speed Class, a U around 1 or 3, and Video Speed Class, a V followed by 6, 10, 30, 60 or 90. The fourth, SD Express Speed Class, marks cards with a newer, faster connection that Nikon does not list. Each number is a minimum sequential write speed in MB/s, except that U1 means 10 and U3 means 30. Sequential write means writing one long file, which is exactly what a video recording is. V60 and V90 apply only to cards with the faster UHS-II or UHS-III connection, and the slot's connection caps the peak, as Card Slots and File Handling covers.

The older multiplier rating counts 1×1\times as 150 kilobytes a second, as Lexar defines it, which is 0.15 MB/s. The 2000x card's name works out to 2000×0.15=3002000 \times 0.15 = 300 MB/s, its read rating, and Nikon writes its 45 MB/s on the same scale as 300×300\times (p. 1011).

04.What Nikon Asks For

Nikon's card advice for video sits on one page (p. 1011). For ordinary recording and playback it asks for at least 45 MB/s of maximum transfer on a CFexpress or XQD card, or a U3 rating or better on an SD card. Where the frame size or rate is high, the figure rises to 250 MB/s on both kinds, with the SD card still rated U3 or better. Slower cards, it warns, may break off recording or playback, and it never defines high.

Set against the bit rates, the lower bar leaves a gap. U3 promises 30 MB/s, or 240 Mbps, which covers 4K at 30p in H.265 10-bit at 190 Mbps but not 4K at 60p at 340 Mbps. The 45 MB/s CFexpress figure, 360 Mbps, falls short of the 370 and 400 Mbps settings. And Nikon's 45 and 250 MB/s figures are maximum transfer rates, peaks rather than floors. My reading, which is inference, is that any setting above U3's 240 Mbps belongs in Nikon's high group: all of 5.4K, 4K at 60p and above, 1080 at 240p and 200p, and every ProRes and raw setting.

The author's cards clear both bars for H.265. The SD card's V90 floor of 90 MB/s is 90÷50=1.890 \div 50 = 1.8 times the heaviest H.265 average, and its rated 260 MB/s write passes 250 MB/s. Both CFexpress cards are rated far above either bar. ProRes 422 HQ is where the SD card's guarantee runs out, since Apple's 221 MB/s for 4K at 60p is well over V90's 90 MB/s.

For raw video Nikon names five CFexpress cards and card series (p. 1011), none of them the author's, and raw video cannot be recorded to SD cards or to XQD cards of 32 GB or less (p. 212), both points Card Slots and File Handling already makes.

When a card cannot keep up, the camera shows Recording interrupted. Please wait., blames a card that does not support the required video write speed, and suggests a faster card or a different Frame size/frame rate (p. 927). Nikon also warns that a slow card may stop a recording short of the maximum length (p. 212).

05.Destination

Destination picks the slot that receives video when two cards are inserted (p. 583). It chooses between the slots the menus call CFexpress/XQD card slot, slot 1 in this guide, and SD card slot, slot 2 (p. 456), and defaults to the first (p. 578). It applies to video alone: the photo shooting menu's Primary slot selection does not move video, and Nikon's tip on that page points to Destination instead (p. 456).

Each slot in the menu carries an estimate of the recording time left on its card (p. 583). Nikon does not say how it reaches that figure, but the camera can see each card's actual free space and a table cannot, so once a card is in I trust its number over mine. While recording, the shooting display keeps a similar running estimate (p. 108).

Destination does not spill over. Recording ends automatically when the current card has no time left (p. 583), and Nikon describes no rollover to the other slot, as First-Day Setup already warns. Nikon describes the item only for two inserted cards and says nothing about an empty chosen slot, so I keep both cards in.

Destination sits in slot 12 of the default video i menu (p. 84), and the author leaves it there, as The i Menu records, so one press of the i button confirms the card before a performance. Time-lapse video, set from the photo shooting menu, has its own Destination, which also defaults to the CFexpress/XQD card slot (p. 448).

My choice is the default, CFexpress/XQD card slot. The Lexar SILVER holds twice what the SD card does, 512 GB against 256, and its rated write speed leaves the most margin for any file type. The SD card stays free for stills, in the second-slot role Card Slots and File Handling gives it.

06.Minutes on the Author's Cards

Nikon's capacity table, itself an estimate rather than a recording, gives fewer minutes than label arithmetic. It gives approximate minutes for one card and one file type, Nikon's 660 GB MC-CF660G CFexpress card with H.265 8-bit (MOV) (pp. 1014-1015). At 4K and 30p, which is 150 Mbps in that file type, it gives about 482 minutes. The label gives 400×660÷(3×150)=264,000÷450≈586.7400 \times 660 \div (3 \times 150) = 264{,}000 \div 450 \approx 586.7 minutes, so Nikon's figure is 482÷586.7≈0.82482 \div 586.7 \approx 0.82 of it. The other rows land between 0.82 and 0.87, for instance 952 minutes against 1,100 at 80 Mbps. Nor does one bit rate always get the same minutes: 4K at 60p and 5.4K at 30p are both 300 Mbps in this file type, yet Nikon estimates 242 minutes for the first and 249 for the second (p. 1014).

Nikon gives no reason for the gap, saying only that capacity varies with shooting conditions and card type (p. 1014). I expect some of it is space the card never offers for footage. Both Lexar pages say usable storage may be less than stated because of formatting and other functions, and SanDisk's page says actual user storage is less. None of them gives a figure, and neither does the RG, so I do not invent one.

I plan with the most cautious of these ratios, 0.82. That is my estimate, not a specification, because it assumes the gap in Nikon's estimates for one 660 GB card and 8-bit files carries over in proportion to other cards and to 10-bit files, which nothing in the RG confirms.

Table 5. Minutes of H.265 10-bit 4K on the author's cards, from the label and with the 0.82 planning ratio

Card30p, label30p, plan60p, label60p, plan
Lexar SILVER 512 GB, slot 1359295201165
Lexar 2000x 256 GB, slot 218014710082
SanDisk Extreme PRO 128 GB, spare90745041

For the recital, a two-hour show uses about a third of the 512 GB card by the label and 120÷295≈41120 \div 295 \approx 41 percent by my margin. The spare's 90 label minutes, 74 by my margin, fall short of the camera's 125-minute limit for one clip (p. 212), so on that card a long take ends at the card. Nikon rates one EN-EL15c battery at about 100 minutes of video (p. 1016), so on the big card the battery runs out first. Clip length, file splitting and power belong to Heat, Power and Recording Limits.

Table 6. Recommended card and Destination settings for the author's video

SettingChoiceWhy
DestinationCFexpress/XQD card slotLargest and fastest card
Video i menu slot 12Destination (default)Confirms the card before a show
Card in slot 1Lexar SILVER 512 GB4K at 30p: 359 minutes by the label, 295 to plan with
Spare cardSanDisk Extreme PRO 128 GB4K at 30p: 90 minutes by the label, 74 to plan with
Check before an eventTime shown in DestinationThe camera's own estimate
Planning marginLabel minutes times 0.82My cautious estimate from Nikon's capacity table

Stage dance. A two-hour show at 4K and 30p needs about 171 GB by the label and fits the 512 GB card with room for a second evening. I check the time on the Destination page before the house lights go down and keep the SanDisk spare in a pocket, a stopgap of 74 to 90 minutes rather than a second show.

Gimbal dance. The frame rate for passages that may be slowed later is set in Video Formats and Codecs, and at 60p, 2.55 GB a minute, the 512 GB card holds about 201 minutes by the label or about 165 by my margin.

08.Beginner Mistakes

Reading Mbps as MB/s. Nikon's bit rates are in megabits and card speeds in megabytes, eight times larger. The 190 Mbps of 4K at 30p is 23.75 MB/s, and mixing the units throws every card sum out eightfold.

Trusting the headline speed. The big number on the packaging is a peak, often a read peak. For video the figure that counts is a sustained write floor: V90 on an SD card, a VPG class on a certified CFexpress card, or a maker's sustained figure, read as the maker's claim.

Expecting video to roll over to the second card. It does not. Recording stops when the Destination card is full (p. 583), whatever the other slot holds.

Planning with the label capacity. Nikon's own capacity estimates come to between 82 and 87 percent of what label arithmetic predicts (pp. 1014-1015). I take the lower figure as my margin, and trust the camera's estimate on the day.

Pointing raw video at the SD card. Raw video cannot be recorded to SD cards at all (p. 212), whatever their speed class, so N-RAW and ProRes RAW need Destination set to the CFexpress/XQD card slot.

Sources and verification

Checked against Z6III firmware 2.00, last verified October 1, 2026.

FeedbackBook mode
z6iiivideomemory-cardscfexpressbit-ratesd-cards