VIOFO High Endurance: two cards, two different factories

Why I bought two of the same card

Every card review you read tests one capacity and then talks about the product line. That has always bothered me, because a 32GB card and a 128GB card of the same name are not the same piece of hardware. They have different numbers of memory chips inside, which changes how fast they can write, and manufacturers are free to change suppliers between capacities without changing a word on the box.

So this time I bought both, ran the full CKBench suite on each, and read the identity register out of both. It is the first time this bench has tested two capacities of one line as a matched pair, and the result was more interesting than I expected. VIOFO's own packaging already tells you something is different: the 128GB claims up to 80 MB/s write and the 32GB claims up to 30 MB/s, on otherwise identical certification marks and an identical 100 MB/s read claim. That is a 2.67 times gap inside one product line, printed by the manufacturer.

Both cards are scored the same way as every other card on the bench, against the same published rubric. Previous runs worth comparing against: the SanDisk High Endurance 128GB, which beat almost every number on its own box, and the Kioxia Exceria G2 128GB that came bundled with a dash cam and quietly keeps four gigabytes for itself, and the bundled Botslab 128GB, whose write speed never sits still for more than three seconds.

What the boxes claim, and what the cards did

Both packages print the same certification marks: C10, U3, V30 and A2. Both claim up to 100 MB/s read. Only the write figure differs. Here is every printed claim on each card against what the bench measured, with the reader named, because a read figure without a reader attached is a meaningless number.

VIOFO High Endurance 128GB

VIOFO High Endurance 32GB

The card says something different from the box

Before any of the numbers, the names, because the packaging cannot keep its story straight even there. You can read this through the blister without opening anything. The packaging calls this line High Endurance. The card itself is printed Industrial next to its capacity: "128 Industrial" on one, "32 Industrial" on the other.

Neither word is defined by anybody. "High endurance" has no certification behind it and no test you can point at. It is a marketing category that usually implies a card built for continuous recording, and when SanDisk uses it they attach 10,000 hours and a bitrate to it, which makes the claim checkable. "Industrial" normally means something more specific again: a part rated for a wider temperature range and a longer service life, sold with published numbers for both. On these cards there are no numbers for either term. Two different unquantified promises, on the two faces of the same product.

Now the numbers, or rather the ones that are missing. Four rows on each table read NOT CLAIMED or NOT A PERFORMANCE CLAIM, and those absences are worth pausing on. Neither card prints an endurance rating of any kind. No hours, no terabytes written, no bitrate condition, on a product line whose name is High Endurance. VIOFO is not alone in this: the Vantrue Endurance PRO 128GB puts the word on the card face and prints no rating on the card or the box either, which makes it two independent brands rather than one brand's quirk. For comparison, the SanDisk High Endurance 128GB prints 10,000 hours and gives its exact benchmark condition in the fine print on the back: Full HD at 26 Mbps. That figure can be checked. It works out to about 117 TB written, roughly 914 complete fills of the card. VIOFO's endurance claim cannot be checked, because there is no claim to check.

The same goes for the temperature icons. Both backs carry a Temperature Proof pictogram beside Water Proof, X-ray Proof and Magnet Proof, with no numbers attached to any of the four. Those are marketing icons, not specifications, and none of them is something a bench can test. Neither package states a warranty term at all.

These two cards were not made in the same factory

Every SD card carries a small read-only register called the CID, written at the factory, which names the manufacturer, an OEM code, a product name, a revision and a manufacturing date. It is the closest thing a memory card has to a birth certificate. I read it out of both cards on a Raspberry Pi rig, because a USB card reader cannot reach it.

The two cards do not match on a single field.

Different manufacturer, different OEM code, different product name, different revision, and seventeen months apart in manufacturing date. They do not even reserve the same proportion of their flash: the 32GB hands over 99.1% of its label and the 128GB hands over 97.4%.

None of that is illegal or even unusual. Brands buy from whoever can supply them, and the badge on the front is the brand's promise rather than a factory's. But it does mean that reading a review of one capacity tells you very little about the other, which is the entire reason this test exists. If you read a glowing review of the VIOFO 128GB and buy the 32GB on the strength of it, you are buying a card from a different company.

Why the 128GB cannot reach its own 80 MB/s

The front of the box says up to 80 MB/s write. Across three separate runs, the card wrote 57.0 MB/s in the sequential test and held between 55.5 and 55.9 MB/s in five-minute sustained tests. The most generous number I can give it is 55.9 against a claimed 80, which is 69.9% of its own claim. There is no test in this set, on either filesystem, in any conditioning state, where it comes close.

That is a real miss, and it is worth being precise about what kind of miss it is. An "up to" figure is a marketing number, and manufacturers are entitled to quote a best case measured on equipment you do not have. What makes this one hard to defend is the size of the gap. A 30% shortfall is not a measurement difference or a reader limitation, and the card's own consistency proves it: this card writes at the same speed no matter what you ask of it. Across the entire 124.6 GB capacity fill the write rate never varied by more than 0.4% from one end of the card to the other. It has one gear.

The good news is that the gear it has is a decent one. Its worst ten seconds of sustained writing was 55.8 MB/s, which holds its V30 rating at 1.86 times the required floor, and it comes third out of seven cards on this bench. It missed a V60 rating by 4.2 MB/s. The card is fine. The number on the box is the problem.

The 32GB holds V30. In July it looked like it did not, and here is what changed

The first run on the 32GB failed its V30 rating badly. A V30 mark certifies that sequential writing never drops below 30 MB/s, and the worst ten seconds of that run measured 14.9 MB/s, less than half the floor. On the bench's standard verdict rule that makes it a V10 card wearing a V30 badge.

But the failure sat entirely in the first fifteen seconds of a five-minute test. After that the card settled at 34.6 MB/s and never dipped again. That is not how a card that simply cannot hold its class behaves, so I ran two more tests in July, and then, in September, the whole suite twice more on two different card readers.

Corrected 15 September 2026. When this review was published, the two July follow-ups seemed to settle it: the slow state appeared only when random small writes ran first, so I wrote that random writes trigger it and that the card needs about 75 seconds to climb back out. The two September runs put the same random writes in front of the sustained test, on two readers, and the card wrote 34.6 MB/s from the first second both times. Not one of the 600 timed samples fell below the floor. The churn explanation was wrong, and the evidence that it was wrong was already in the July run: its sequential write had fallen to 14 MB/s on the second and third passes, before any random writes had run, which the tool flagged as write variance and I read past.

What the card actually does after a burst of random writes is much smaller. On both September runs the untimed warm-up that precedes the sustained test dipped to about 21 MB/s for eleven seconds, then returned to 34.6 and stayed there, and the timed test began more than a minute later. That is real, it is brief, and it is why the warm-up exists.

What the July run was, I cannot prove. It was the card's first-ever write, and a second card built on the same Longsys platform did the same thing on its own first run and never again, which fits a card that behaves differently for its first few gigabytes. It also fits a transient load on my bench machine, which recorded no host telemetry then and has since had background processes found and removed. I am not choosing between them. What I can say is that five later runs on two readers, including two with the full random-write burst in front, all agree, and that is the card's behavior.

One thing the clean runs did show that the slow one had hidden: under a minute of scattered 4K writes, the card's random-write rate falls by 16 to 19% between the first pass and the third, on both readers, the largest fade of any card on this bench. A dash cam never issues that workload, and it does not touch the video rating, but it is a real sensitivity to churn and it is scored where it belongs, in endurance behavior.

Your 128GB card arrives in the wrong format

This one surprised me. The 128GB card came formatted as FAT32 from the factory. Cards above 32 GB are supposed to arrive as exFAT, and that is not a technicality: it is what the SD specification assigns to that capacity class. VIOFO ships this one FAT32 anyway.

The reason is almost certainly sensible. A lot of dash cams read FAT32 and nothing else, and VIOFO sells dash cams. Formatted the way it ships, this card works in the widest possible range of cameras, including old ones.

The cost is that FAT32 cannot store a single file larger than 4 GB. Out of the box, this 128 GB card cannot hold one file bigger than 4 GB. A dash cam never notices, because loop recording writes short clips by design. A phone, a drone or a camera shooting a long 4K clip hits that wall in the middle of recording. The package lists Smartphones, Action Cams, Body Cams, DSLRs and Drones alongside Dash Cams, and for most of those the shipped format is the wrong one. Reformatting to exFAT fixes it in under a minute, and nothing on the packaging tells you to.

I also checked whether the shipped format was costing the card any speed, by running the same test twice and changing only the filesystem. On exFAT the card came out 0.76% faster on its worst ten seconds, which is noise. The format costs you the file size limit and nothing else.

One small bonus if you do reformat: exFAT hands back about 25 MB more usable space than FAT32 on this card. That is much the smaller of the two reasons to do it.

Are they really 32 GB and 128 GB? Every byte, verified

This is the test a counterfeit cannot survive. I write every free byte on the card with data that cannot be guessed, then read all of it back and compare. A card lying about its size fails loudly at its real capacity. Both VIOFO cards passed with zero mismatched bytes and zero read errors, across 156.3 GB written and verified between them.

Before anyone panics about the missing gigabytes: this is normal and it is not the same thing as a fake card. Every memory card sets aside some of its flash for spare blocks and internal bookkeeping, and separately, Windows counts in binary while card labels count in decimal. That is why an honest 128 GB card shows up as roughly 119 GB in Windows. What matters is that the card holds what it says, and both of these do. For contrast, the counterfeit SanDisk I tested delivered 26.8% of its label while passing every speed test thrown at it.

The gap between the two VIOFO cards is still worth reporting. The same brand, bought at the same time, hands over 99.1% of the label at one capacity and 97.4% at the other. On this bench the SanDisk High Endurance delivered a full 100.0% and the Kioxia delivered 96.7%, so the VIOFO pair sit mid-table, in two different places.

I also checked each card's own declared capacity against what it actually delivered, and both matched to within normal filesystem overhead. Neither card overstates itself, which is exactly the check the counterfeit failed.

How hot do they get?

Both cards were shot with a thermal camera the moment the five-minute write stopped, fan off, at an ambient of 76 F or 24.4 C. When this was published in July they were the two hottest cards this bench had measured; several faster cards have since run hotter, and the 32GB read the same 153 F again on both readers in September.

The comparison that matters is the first two rows against the third. The VIOFO 128GB runs 9 F or 5 C hotter than the SanDisk High Endurance while writing 32 MB/s slower. More heat for less work is the part worth knowing about in a card sold for a device that lives on a windscreen in the sun.

That said, do not let the raw number alarm you. Flash tolerates these temperatures, and the bench deliberately generates a thermal load no dash cam will ever impose: it writes flat out for five minutes, where a 4K30 dash cam writes about 6 MB/s, roughly one-ninth of this workload. Neither card throttled. The practical advice is the same as always: give the camera some air and keep it out of direct sun where you can.

What can these cards actually record?

Every row is judged against the worst ten seconds of the run, never the average, because your camera does not get to record over the average. The 32GB column shows its video-duty figure of 34.2 MB/s, which is what a dash cam will actually see.

For the job both cards are sold for, they are comfortable. A 4K30 dash cam needs 6 MB/s and the smaller card delivers nearly six times that under video duty. The caveats in this review are about honesty on the box, not about whether these cards can record your commute.

Two scores, one recommendation

The 32GB now outscores the 128GB, and the 128GB still gets the only buy link below. That is deliberate, and it is worth explaining rather than leaving it to look like an inconsistency. The score measures how well a card does what it claims. The link is about who should actually buy one. The 32GB earns its 3.5 honestly, because every number on its box holds, but a dash cam card's job is retention, and 32 GB is a short loop that a 4K30 dash cam at 6 MB/s overwrites in about an hour and a half; there are better ways to spend that money. The 128GB has a worse claim-accuracy problem and still earns the recommendation, because for the one job it is bought for it does the work without ever wobbling.

Bottom line: buy the big one, for the one job

The VIOFO High Endurance 128GB is a card I would put in a dash cam without hesitating. It writes 55.8 MB/s at its worst, holds its V30 rating with margin to spare, delivers every byte it promises, and never once varied across a full 124.6 GB write. What it does not do is reach the 80 MB/s printed on its own front, and it runs hotter than any card I have measured. Buy it for what it is rather than what the box says, and reformat it to exFAT unless it is going straight into a camera.

The 32GB is a harder card to recommend, not because it is bad but because it is small. It beats its own printed write speed, hands over more of its label than the big one does, and on re-test holds every mark on its face, which its first run in July had made me doubt. For a dash cam it works, for as long as 32 GB of loop lasts. For anything that leans on scattered writes, it is the card here that fades the most under them.

The bigger lesson is the one the identity registers gave up. Two cards, one name, two manufacturers, seventeen months apart. Whatever you read about one capacity of a card line, including this review, do not assume it carries over to another.

Both cards now sit on CKBench alongside the SanDisk High Endurance, the Kioxia Exceria G2, the counterfeit ImageMate and the SanDisk Ultra Fit, all scored the same way.

Disclosure: I bought both cards at retail with my own money; VIOFO had no involvement in this test and no editorial input. This post contains affiliate links. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you. Affiliate links like these and my YouTube channel are what keep the site running.