VIOFO High Endurance 64GB: the small claim is the fast card

Why the third capacity got its own review

I have already tested this line twice. The 32GB and 128GB went through the bench as a pair in July, and they turned out to be two different cards from two different factories wearing one name, with a write-claim ladder that read 30, 40 and 80 MB/s across the three capacities. The 64GB is the one in the middle, and it arrived in September alongside a VIOFO A119M Pro dash cam that VIOFO sent for review. That makes it the second manufacturer-supplied card on this bench after the Vantrue Endurance PRO, and it gets the same treatment: one unit, the full suite, and a second card reader to check the first. It joins CKBench, the running index of every card on the bench, alongside those two and the SanDisk MAX Endurance 128GB, which is the endurance card that actually prints an endurance rating.

The lesson the pair taught me is the reason this is a separate review rather than a paragraph added to that one: whatever is true of one capacity of a card line does not carry over to another. This card proves the point in the friendly direction. It is the 128GB's silicon, it prints half the 128GB's write claim, and it beat the 128GB anyway.

What the box claims, and what the card did

Everything below is read off the retail packaging and the card face, which agree with each other. Two rows carry a printed number, four carry a class mark, and the rest are empty on purpose: a claim the card does not make still gets a row, because the absence is information.

The read verdict needs a sentence. 96 MB/s is where the standard UHS-I bus tops out in practice, and this card reached it on both readers, one of which pulls over 200 MB/s from cards that speak a faster mode. So the printed 100 is the bus's number rather than the card's, the card delivers all the bus allows, and I call that met. A2 is not a failure and I never score it as one: the A-classes assume command queuing that USB card readers do not perform, so A2 is untestable on any reader bench, for any card. The A1 thresholds of 1,500 read and 500 write IOPS are at least measurable at queue depth 1, and this card cleared both, on both readers.

The box prints a write figure, and undersells the card

Unlike the Vantrue, this card prints numbers: 100 MB/s read and 40 MB/s write on the front of the pack, with C10, U3, V30 and A2 on the back and on the card face. The write figure is the conservative one. Its worst ten seconds across five minutes of committed writing was 59.5 MB/s and its steady state 60.2, so the number on the box is about two thirds of what the card sustains. That is accurate, and it is the same direction SanDisk goes with its High Endurance line, whose 128GB measured more than twice its printed write on this bench.

What the box does not print is the same thing the whole VIOFO line does not print. No endurance rating of any kind, on a product called High Endurance whose card face reads Industrial: no hours, no terabytes written, no bitrate condition. No warranty term either. Temperature, water, X-ray and magnet resistance are ticked as proof claims with no range or standard attached. Three capacities of this line have now been through the bench and all three carry exactly this posture, so it is the line, not the unit. The contrast is SanDisk again, whose endurance boxes print a capacity-indexed hours ladder and the 26 Mbps Full HD condition that ladder assumes. One thing this pack does print that its siblings' packs do not show is a model number, SD-64G, on the label in the lower right, which is the kind of thing that lets a reader tell this hardware from a future revision sold under the same name.

How fast is it when the writing does not stop?

This is the number that decides a video speed class, and it is always the worst ten-second window rather than an average, because a camera does not get to record over the mean. Five minutes of continuous committed writing with the cooling fan off, sampled once a second, after an untimed warm-up so the clock starts on a controller that has finished reacting to the previous test.

On the V60 line, without holding it

Ten consecutive thirty-second blocks land between 60.0 and 60.5 MB/s. The worst single second in the whole five minutes is 55.6 MB/s. There is no cache cliff, and I checked that twice: the knee scan came back with no knee, and the whole-card fill later in this review held 60.5 MB/s across 62.4 GB with not one of its 1,062 samples dropping below 53.

Against the V30 the card prints, 59.5 MB/s is twice the floor. Against V60, which it does not print, it is 0.55 MB/s short on this reader and 1.6 short on the second, with a steady state that sits at 60.2 on both. So this is a V30 card whose steady speed is V60-class and whose worst window is not, on either reader, and I am reporting it exactly that way rather than rounding it up. In proportion: the one dash cam duty cycle I have measured by reading a four-channel camera's own recordings off its card came to about 4.8 MB/s aggregate. This card's worst moment is roughly twelve times that.

Does it slow down under the writes loop recording actually makes?

Loop recording is not one long clean write. It is scattered small writes: event saves, file table updates, overwrites of the oldest clip. Three passes of random 4K writes in a row show whether a card's controller degrades as that churn accumulates. This is a directional indicator, not an endurance test, and nothing done in an hour tells you about year three.

It gets faster, not slower

1,722, 1,738 and 1,744 write IOPS across three consecutive passes. The third pass is the best one, 1.3% above the first, which is the controller warming into the workload rather than degrading under it. For comparison, the reference card in that panel is a cheap Class 4 unit that loses 71% of its write IOPS across the same three passes, and that collapse is what this test exists to find. The second reader told the same story in its own numbers: 1,525, then 2,054, then 2,070.

The warm-up before the sustained test says the same thing from another angle. On both readers this card came out of the random 4K leg already writing at its full media rate on the first thirty-second window, 60.1 and 59.7 MB/s, so the timed test started at the earliest moment the analyzer allows. It does not need to recover from churn, because the churn did not cost it anything. That is worth noting on this line specifically, because its 32GB sibling's first run looked very different on this exact test, and the pair review carries the correction.

Is all the space really there?

Every free byte gets filled with data generated from a seeded random stream, then every byte gets read back and compared. It is slow and it is the only test that catches a fake, because a fraudulent card looks perfectly normal until you use the far end of it. On a 64 GB card it is also cheap, which is no reason to skip it.

62.4 GB, verified, zero errors

62,356,619,264 bytes written across 59 files, the same number read back and compared, zero mismatched bytes and zero read errors. That is a pass, and after three counterfeit 128GB cards this month that returned 66, 48 and 65 GB of real space, it is never a formality, supplied unit or not.

The 97.4% is not a shortfall to hold against it. Every card reserves flash for spares and metadata, no usable-capacity figure is printed on this package for it to miss, and genuine cards on this bench land between 96.7% and 100.1% of their label. Its two siblings declare 97.4% and 99.1%. Your computer will show it as about 58 GB, because Windows counts in binary and the label counts in decimal; that is arithmetic, not missing space. Throughput across the address range was flat too: 58.7, 59.1 and 58.6 MB/s writing across the first, middle and last tenth of the card, a spread of 0.8%. Tiered fake flash shows its seams there, and this card has none.

How hot does it get, and does it slow down?

Five minutes of uncooled sustained writing, with an infrared frame captured the moment the timed write stops, because a card sheds heat in seconds. The question is always two questions: how hot, and did the card give up any speed getting there.

The hottest of the three, and the least margin

It did not throttle. The write trace is flat at 60.2 MB/s for the full five minutes with no decay at the end, the second reader's trace is flat at 60.3, and the 62.4 GB fill held its speed too. On the test that matters most, heat cost this card nothing.

But 174 °F (78.9 °C) is the highest peak of the three VIOFO High Endurance capacities, 14 °F above the 128GB and 21 °F above the 32GB at the same 76 °F ambient, and it ties two premium 512GB cards for the second-highest peak on this bench, behind only an SD Express card running in a PCIe reader. It reached that while writing 60 MB/s, which is ordinary work. Two things make that worth a section rather than a footnote. First, the reading is a floor, not a ceiling. The frame below sees the part of the card outside the reader's slot, and the controller sits inside it; when I have ejected cards at the end of a run and imaged them face-on, the in-slot reading has under-reported the surface by 10 to 14 °F, which would put this card's true surface in the mid 80s °C, at the roughly 85 °C these parts are typically rated for. VIOFO prints no operating range of its own to check that against.

Second, and this is the part that matters for this specific product: my ambient was 76 °F in a climate-controlled room, and a windshield in summer is not 76 °F. This is a card sold to sit behind glass in a parked car. It cost no speed on my bench and I am not calling it a defect. It is the one dimension where this card is measurably behind its siblings, and it is behind them in exactly the environment it was built for.

Two readers, one answer

Every card on this bench now gets a second run on a second card reader with a different bridge chip, because a reader can change an answer and the only way to know is to ask twice. This one ran the full suite with the capacity fill on the ProGrade PGM0.5, then everything but the fill again on the SanDisk SDDR-B751. The two agree on everything that decides a verdict.

The one row that moves is random 4K, and it moves the way it always does between these two readers: the SanDisk's bridge turns a single small command around about 130 microseconds faster, which is a third of a 4K operation and under 1% of a 1 MiB one. That is a property of the reader, not the card, it is the reason I report the A-class as indicative, and it is why the A1 figures above are attributed to a reader rather than quoted bare. Its first-ever run was its canonical run, and unlike its 32GB sibling, nothing about it needed excluding.

Where it sits between the 32GB and the 128GB

Three capacities of one product line have now been through the same bench, which no other brand here can say. VIOFO's own write claims run 30, 40 and 80 MB/s up the ladder. The bench measured 34, 59 and 56 through the worst ten seconds, so the middle card writes fastest, the top card is the only one that misses its own number, and the claims track the capacities rather than the silicon.

The identity registers settle which sibling this card belongs to. The pair review found the 32GB built by Longsys and the 128GB by a manufacturer whose ID is not on any public list, and the 64GB reports that second maker's ID, with a configuration register that matches the 128GB's byte for byte. So the line is Longsys at 32GB and the other platform from 64GB up, and this card and the 128GB are the same controller family at two capacities. That is exactly why the 64GB out-writing the 128GB is interesting: same silicon, half the claim, a faster result. The three-card table also settles the endurance question for the line as a whole. None of them prints a rating.

How the 4 breaks down

The bottom line

Judged on what it does, this is the best card in its line. It beats the write figure on its box by half, its worst ten seconds are twice its printed class and a whisker under the class above, it does not fade under churn, its capacity is entirely real, and two readers agree on all of it. If one arrives with your camera, or you buy one, there is nothing here that should worry you about the card itself.

Two things should shape whether you buy this one rather than its bigger sibling. The first is capacity, which is the only thing 64 GB is short of: a 4K dash cam overwrites it in under three hours, and the 128GB of the same line costs you nothing in speed. The second is heat, the one measured mark against it, and it lands squarely in the use case the card was built for. Neither is a reason to avoid it. Both are reasons to know what you are buying.

Disclosure: VIOFO supplied this card alongside a VIOFO A119M Pro dash cam they sent for review. I did not buy it, and it was not inside the camera's retail box; the card is sold separately. VIOFO had no editorial input, saw no result before publication, and had no opportunity to influence any number on this page. 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.