Memory Cards for Video Recording: Speed Classes, Sustained Writes, and Recovery
You press record. The counter climbs to four seconds, then stops. The camera throws an error you've never seen, and the take is gone. The card in the slot…

Research updated Sep 10, 2026
Key topics
You press record. The counter climbs to four seconds, then stops. The camera throws an error you've never seen, and the take is gone. The card in the slot has a big read-speed number printed on it — a number that had nothing to do with what just happened.
That failure can come from several places: a card that can't hold the write rate, a format the camera won't accept, a missing required speed class, a filesystem mismatch, or a card that has simply gone bad. The one thing you can rule out is the marketing number. Peak read speed is not evidence that a card will record reliably in your camera, in your mode, for the length of your clip.
This guide works backward from your camera's documented requirement, not forward from the card's label. It assumes you already know your camera's codec and bitrate. If you don't, find those two numbers first — everything below depends on them.
The Fast Answer: A Three-Tier Decision Snapshot
Before the mechanisms, here's how to self-sort. Then verify against your camera's documentation.
| Tier | What it means | Buy this when | What you're paying for | Skip it when |
|---|---|---|---|---|
| Minimum viable | A card that meets your camera's documented class or VPG rating for the mode you actually shoot, with capacity for one session plus rotation headroom | Your modes are served by a compliant card and your sessions fit your capacity plan | Nothing extra — this is the capability floor | Never skip this tier; it's the floor, not a compromise |
| Recommended | One class or VPG tier above the documented minimum | You record long takes, high frame rates, or intraframe codecs, or you want margin for modes you plan to use | Headroom you will actually grow into | Your worst-case mode never approaches the higher requirement |
| Diminishing returns | The highest sustained-write tier available | Rarely, and only when a specific mode is currently blocked | A number, not a different result | Your camera's modes never approach it |
The rest of this article explains why the middle column is the one that matters, and how to find your camera's actual requirement.
Start From the Camera, Not the Card
The camera's documented media requirement is the governing constraint. Everything else — brand, read speed, review volume — is secondary.
Your camera manual or support page defines four things that decide your purchase:
- Supported card formats. SD versus CFexpress Type A is not a preference; it's a slot. A card in the wrong form factor won't seat.
- Required speed class or VPG rating. This is the number your card must print to unlock specific recording modes.
- Capacity limits. Some cameras cap the capacity they'll address, and some impose per-clip file limits regardless of free space.
- Which recording modes each requirement unlocks. A card that meets the minimum may record standard 4K but refuse a high-bitrate intraframe mode entirely.
A card can be fast and still be rejected. Wrong bus, unsupported capacity, or a missing required class can trigger a camera error even when the read and write numbers on the label look impressive. Sony's own support documentation describes this directly: an SDXC card with high read/write speeds marked but no V60 or V90 logo will not be recognized by a camera that requires that class, and the camera will display an error message rather than fall back to a slower mode.
Manufacturer compatibility lists exist for exactly this reason. Treat them as your first filter, not a suggestion you can out-think.
Sustained Write vs. Advertised Speed
This is the single most consequential mechanism in the article, and the one most buyers get wrong.
The chain runs like this:
Codec bitrate → required write rate per second → the card's guaranteed minimum sustained write → whether the camera can keep recording without buffering out.
Advertised read speed and peak write speed describe burst behavior. They tell you how fast data moves in a short window under favorable conditions. They do not prove the card holds the required rate across a ten-minute take, or after the card has warmed up, or when the buffer is under continuous pressure.
This is why speed classes and VPG ratings exist. They describe a floor, not a ceiling. A V60 rating is a promise about minimum sustained write; a 277 MB/s read figure is a claim about peak transfer. Those are different numbers answering different questions.
The observable consequence of getting this wrong is specific: recording stops after a few seconds, the buffer fills and the camera pauses, or the camera refuses the mode entirely. It rarely fails immediately. It fails once the buffer that was absorbing the mismatch runs dry.
Decision rule: match the card's guaranteed minimum sustained write to your camera's documented requirement for your worst-case mode. Add margin only if you actually shoot that mode regularly.
Speed Class Labels: C, U, V, and VPG
The symbols on a card are not interchangeable. They belong to different systems with different guarantees, and the same card often carries several at once.
- Speed Class (C) and UHS Speed Class (U) are older systems. You'll still see C10 and U3 on cards.
- Video Speed Class (V) is the one that matters most for video. V30, V60, and V90 correspond to progressively higher guaranteed minimum sustained write rates.
- VPG (Video Performance Guarantee) is a separate guarantee aimed at video-oriented media, particularly CFexpress. It's often what a camera manual actually specifies for demanding modes.
Sony's support documentation lays out the minimum write speeds behind these classes: V90 at 90 MB/s, V60 at 60 MB/s, V30 at 30 MB/s, and so on down the ladder. The class you need is set by your camera's mode, not by a general "higher is better" instinct.
The most common mistake here is assuming a high read-speed number substitutes for a missing required class. It doesn't. Some cameras won't recognize the card's speed at all and will show an error rather than degrade gracefully.
Decision rule: find the class or VPG level your camera documents for your mode, then confirm the card prints that class. If the class is missing, the card is wrong regardless of what else is printed on it.
SD vs CFexpress: When the Format Changes the Decision
CFexpress Type A and similar formats exist because some codecs and frame rates exceed what SD buses can sustain. The format boundary is a capability threshold, not a prestige tier.
CFexpress can unlock recording modes unavailable to slower media. Sony's M Series CFexpress Type A cards, for example, are documented as supporting VPG200 — a guarantee of stable recording at 200 MB/s — and Sony connects that specifically to extended high-bitrate XAVC S-I 4K recording on Cinema Line cameras. That's a real capability difference, not a marketing one.
But CFexpress adds cost beyond the card: a compatible reader, often a different offload workflow, and higher per-gigabyte storage expense. Sony's G Series CFexpress Type A cards are documented with read speeds up to 1,800 MB/s and write speeds up to 1,700 MB/s — manufacturer specifications, not independent measurements, and maximum figures rather than sustained ones.
Decision boundary: buy CFexpress when your camera documents it as required for a mode you actually shoot. Skip it when your modes are fully served by a compliant SD card.
One hidden dependency: a CFexpress card is useless without a compatible reader, and reader speed affects your offload time more than the card's peak number suggests. If your reader is the bottleneck, the fast card just waits.
Capacity, Session Length, and Card Rotation
Capacity is a calculation, not a preference.
Your requirement follows from bitrate × recording time × number of takes, plus room for the file system and any per-clip limits the camera imposes. A camera that splits long recordings into separate files still consumes the same total space.
A worked example makes the arithmetic concrete. Suppose your worst-case mode records at roughly 200 Mbps — about 25 MB/s. A 30-minute continuous take consumes roughly 45 GB. Two takes at that length plus a safety margin puts you near 100 GB before you've accounted for retakes, false starts, or a second session. That points toward a 128 GB card as a floor for this mode, not a 64 GB card.
Treat that number as a starting point, not a rule. Your camera's manual usually publishes recording-time tables for each mode and capacity, and those tables override your arithmetic because they account for the camera's actual file behavior. Supported capacity limits also override it: if the camera caps at a certain size, a larger card buys you nothing.
Two mid-capacity cards usually beat one large card — but only when the workflow supports rotation:
- Rotation lets you keep shooting while offloading. One card records while the other transfers.
- Rotation limits how much footage one failure can take with it. A single card failure costs you one card's worth of work, not the whole shoot.
- Large-capacity cards reduce mid-shoot swaps but increase the cost of a single failure and can slow offload if your reader is the bottleneck.
The two-card rule is a risk-management preference, not a universal buying rule. It earns its extra cost when the shoot is paid or irreplaceable, when your camera has dual slots or you can swap cards without stopping, and when you can offload safely during the session. It's the wrong default when you shoot on a single-slot camera, travel with minimal gear, can't offload mid-session, or work in modes where a single large card is the only supported option. In those cases, one correctly sized card plus a verified spare in the bag is a reasonable plan.
Decision rule: size cards to your longest realistic session plus margin. Buy two when rotation fits your workflow and the footage is hard to replace. Buy one large card plus a spare when it doesn't.
Verifying a Card Before It Matters
A bad card or wrong format should be discovered at your desk, not on location.
- Confirm the card is recognized by the camera and that your intended recording mode is actually selectable. Some incompatibilities only surface when you try to engage the mode.
- Record a representative clip at your worst-case settings. Confirm it completes without stopping, then play it back and check the file.
- Format the card in the camera, not the computer. This matches the file system to what the camera expects.
- Check for manufacturer utilities. Some brands offer card-health tools and file-rescue software. Coverage and behavior vary by brand and model — do not assume every card is supported.
One evidence caveat worth stating plainly: independent long-duration and thermal testing of specific cards is limited in publicly available sources. Treat any single test as context-bound rather than universal. A card that held up in one reviewer's camera at one temperature is not proven across all cameras and conditions.
When a Card Fails Mid-Shoot
Prevention beats recovery, and recovery is less reliable than most creators assume.
- Stop writing to the card immediately when footage is missing or the card is behaving abnormally. That's the safest first response, and it keeps your options open. Continued use reduces the chance of recovering anything.
- Some manufacturers provide file-recovery software for their own cards. Sony, for example, offers File Rescue software for its memory card customers. Availability, supported file types, and success rates vary, and recovery is not guaranteed.
- Keep a spare verified card on set. The practical recovery plan for a paid shoot is usually "swap and keep working," not "recover the file."
Recovery software is not a backup. It's a last resort for a specific failure mode — accidental deletion or filesystem corruption — and it does nothing for a card that has physically or electronically failed. The available sources do not establish failure rates or recovery success rates for specific cards, so treat reliability claims from any source cautiously. What you can control is exposure: rotation, verified cards, and a spare that has already been tested in your camera.
Common Mistakes and Overbuying
- Buying by read speed instead of the camera's documented class requirement. Peak read speed is not evidence of recording compatibility or sustained performance. It's the most common way to end up with the wrong card.
- Buying the highest tier "for safety" when your camera's modes never approach it. Headroom you never use is money you never see again.
- Ignoring the reader and offload path. A fast card behind a slow reader feels slow in practice.
- Assuming a card that works in one camera will work in another. Compatibility is per-camera and per-mode.
- Treating brand fame or review volume as evidence of reliability. Popularity is a discovery signal, not a quality score.
Decision Rule: When to Move Up, Down, or Change Format
Move up a class or to CFexpress only when your camera documents a requirement you actually shoot, or when a mode you plan to use is currently blocked.
Stay at the documented minimum when your modes are served and your sessions fit your capacity plan. The cheaper card is the correct card.
Move down when you're paying for sustained-write headroom your codec never touches.
Change format only when the camera's slot and modes demand it — not because a higher number looks safer.
The repeatable procedure: name your worst-case recording mode, find its documented media requirement, confirm the card prints that class or VPG level, size capacity to your session, and decide on rotation based on how replaceable the footage is. The camera's requirement governs the purchase. The card's headline number is just a number until it clears that bar.


