Speederlash 128GB: a better card than its box says

What is the Speederlash 128GB?

Speederlash is a name I found only on marketplace listings. No brand website turned up, the pack names no company, only an address in Pinghu Street, Longgang District, Shenzhen and Made in China, and after-sales support is a Gmail address. The card comes in a retail blister with an SD adapter, and neither the pack nor the card states a speed; the Amazon listing does. I bought this one sealed on Amazon on 6 October 2026, and it is the first Speederlash card on my bench.

It is worth testing for the same reason it is easy to dismiss. Plenty of no-name cards are fakes, and this one ticks the boxes that make me suspicious before I open it: a brand with no footprint, a pack with odd English, and fine print that contradicts itself. A fake would show up in the capacity test, and an inflated speed grade in the sustained write. The card is in CKBench now, the running index of every card on my bench, next to other budget 128 GB cards I have bought this autumn, among them the PNY Premier-X, the Silicon Power Superior Pro and the LinkMore XV23 Lite.

What the Speederlash 128GB claims, and what it did

The card prints class marks and nothing else, so the class marks are what it is judged on, and it holds every one of them with room. The read figure exists only on the Amazon listing, which says up to 95 MB/s. My rubric scores a speed only when the card, its pack or the maker’s own published spec states it, and a retailer listing does not count; the card would meet it anyway, at 96.8 MB/s on the ProGrade and 96.9 on the QuickFlow, the practical ceiling of the SDR104 mode, where most standard UHS-I cards on my bench land. A1 is not testable on a USB reader either: at queue depth one the card cleared both A1 floors on the QuickFlow and fell just short of the write floor on the ProGrade, which is indicative and nothing more.

A box that gets its own card wrong

The front of the blister carries the brand, a holographic 128GB sticker, three lines of feature copy (Data transformation security, presumably meaning transfer, Strong compatibility and reliability and 4K Ultra HD video quality) and the card itself under the plastic beside its adapter. There is no speed figure anywhere on it. The back adds icons for video, photos and audio, a durability line, the SD trademark notice, an address in Shenzhen with no company name, an operating range of 0 to 70 °C and one set of class marks. Amazon’s fulfilment sticker covers the end of the address line. There was no returns sticker on any panel.

Those class marks are where the box and the card part ways. The pack prints V30 beside a U1, the lowest UHS speed grade, with a 10 MB/s floor. The card under the plastic prints V30 beside a U3, whose floor is 30 MB/s, plus Class 10 and A1, neither of which appears on the pack. V30 with U1 is an odd pair in any case: V30 promises the same 30 MB/s sustained write that U3 does, so a card that earns V30 earns U3 as well. The card’s own registers, which I read before its first run, declare U3, and so does the Amazon listing. The pack is the odd one out, and it errs in the card’s favour: the card holds U3, as the sustained write below shows. The capacity footnote has a slip of its own. It defines 1GB as 1,000,000 bytes, short by a factor of a thousand; a gigabyte on a memory card label is 1,000,000,000 bytes, and that is the definition the card is measured against below.

The card arrived empty, formatted exFAT with 128 KiB clusters and no volume label. One detail of that factory format is unusual on my bench: its partition starts 32 KiB into the card, where every other card whose factory format I have recorded starts its partition on a 1, 4 or 16 MiB boundary. It suggests a different factory tool from the other cards’. I ran the card as it came, so the results below include whatever that costs.

Five minutes at 47 MB/s, three and a half seconds at a time

The timed write starts after an untimed warm-up that runs until the speed stops changing, so the five minutes that decide the video class start on a settled card. On this card the warm-up settled at the earliest point the tool can declare, 60 seconds, at 46.1 MB/s. The trace below is the scored run, one sample a second, with the speed-class floors marked. The first thirty seconds averaged 46.3 MB/s and the last thirty 47.1, the whole five minutes 47.0, and the worst ten seconds were 44.25 MB/s, at 1:27. On the second reader, after the whole card had been written once, the same five minutes read 46.9 steady with worst ten seconds of 43.94.

44.25 is 14.25 MB/s above the V30 floor, so V30 and U3 hold with room to spare. The trace is not a smooth line, though, and the jagged band is the card, not noise. It writes in a rhythm about three and a half seconds long: roughly 146 MiB goes in at about 80 MB/s, then the card writes at about 6 MB/s for a second and a half, and it repeats, 86 times in the five minutes and all the way through a write of the whole card. A one-second sample cannot draw a cycle that short cleanly, so the trace jumps between seconds as low as 5 MB/s and as high as 83; a fit to the samples gives a period of 3.46 seconds in the timed run and 3.54 through the fill, with the same 155 MiB of data in every cycle either way.

The slow part is not a stall. The longest any single 1 MiB write took was 240 milliseconds, and the bench flags anything over half a second. A camera buffers a few seconds of video and never sees the rhythm; what it sees is the average, about 47 MB/s, and the worst ten seconds, which this card measures honestly. They depend a little on where a window falls in the cycle, 40.2 MB/s for the worst nine seconds and 46.4 for the worst fourteen, and every one of them clears V30. Among the budget cards on my bench that puts it beside the PNY Premier-X (47.1) and the FitcamX (43.8), above the Silicon Power Superior Pro (37.8) and the Newegg TeamGroup (35.5), and well below the LinkMore XV23 Lite (77.3).

The Gigastone’s registers, one flag apart

Every SD card carries a set of registers, written at the factory, that a USB reader cannot show you and a native SD host can. One names the card: manufacturer ID, OEM ID, product name, revision, serial and build date. The others describe it: how big it is, which SD specification it follows, which speed classes it declares. I read this card’s on a Raspberry Pi before its first run, and apart from the serial, the build date and a single flag they are those of the Gigastone High Endurance Pro 128GB, a card from an established brand that I benched in September. The flag sits in the SD specification register and says whether erased memory reads back as zeros or as ones; there the Speederlash matches the Newegg TeamGroup 128GB to the byte instead.

Manufacturer ID 0xFE is not on any public list, so it names no company; my manufacturer ID register has five cards under it now, two of them free cards bundled with dash cams. The SD Status block, the 64 bytes that declare the speed classes and a few performance fields, is identical on all five and on the PNY Premier-X, which reports a different manufacturer ID, and it carries a value of 4 in a performance field that no card off this platform reports. Whatever is inside these cards, it is one platform, and on my bench alone it is sold or bundled under six names, which the identity explainer lays out register by register.

That lets me compare how the platform’s cards behave, and they are not all the same card. Every one of them takes data at about 80 MB/s; what differs is what happens when that first stage fills.

The Speederlash carries the Gigastone’s identity and writes like the other end of the family: a short cycle that drops to about 6 MB/s, not the Gigastone’s nine-second saw-tooth with a 52 MB/s lower step. It sits between the TeamGroup and the PNY on the length of its cycle, the size of its fast stage, its average and its worst ten seconds. The registers name the platform; they do not tell you which build of it you are holding, or which flash is behind it. The LinkMore XV23 Lite taught the same lesson on another platform, and it is the reason I test every card rather than trusting a register match.

125.1 GB, every byte verified

The capacity test wrote 117 files of unpredictable, dedup-proof data until the card was full, 125,065,101,312 bytes, then read every one of them back and compared. Zero mismatched bytes. Zero read errors. The fill took 45 minutes 37 seconds and the read-back 22 minutes 19 seconds. Across the first, middle and last tenth of the card the write ran at 46.2, 45.9 and 45.9 MB/s, a spread of 0.76%, and the read spread by 0.94%: flat from one end of the card to the other, where the three no-brand fakes on my bench spread by 10 to 23%.

The three-and-a-half-second rhythm ran through the whole fill, at the same 155 MiB a cycle from the first gigabyte to the last, and the worst ten seconds anywhere in it were 40.37 MB/s, still V30. The chart below puts this fill beside the two other cards on my bench that declare exactly the same capacity, all three filled on the same reader on their first run.

The card declares 125,069,950,976 bytes, 97.71% of the 128 GB on the label, which is ordinary overhead for a genuine card and the same figure as the Gigastone and the PNY. Windows will show it as about 116 GB, because it counts in binary, and that is the figure the Amazon listing quotes as usable space. That is arithmetic, not a shortfall; the methodology page has the conversion. The capacity row of the rubric scores the declared share of the label, and 97.71% is a 4.

No slower after a whole-card write

Some cards write much more slowly once they have been filled, until something tells them which space is free again; why a used microSD card tests slower covers the effect. A card reader cannot send that signal, so a card filled and emptied over USB stays in that used state, and it is the state a dash cam card lives in. The Gigastone, which reports this card’s identity, showed it on my bench. So the Speederlash’s first run ended by filling and emptying the whole card, and its second run started 29 minutes later on the other reader.

The Speederlash came back within a third of a megabyte a second of its first run, on a different reader, which is the kind of agreement two readers give on a card that has not changed. The Gigastone lost almost half its worst ten seconds after its first fill, on the same reader, and needed a native SD erase to recover; the Speederlash, like the PNY, has nothing to recover from.

Random writes barely move it

Loop recording is a long write interrupted constantly by small ones, and random 4K writes are the bench version of that: three twenty-second passes, and the question is whether the third is as fast as the first. On this card it nearly was. The scored run did 479, 458 and 465 write IOPS, 1.96, 1.88 and 1.90 MB/s, a fade of 3.06%, inside the 5% the bench itself treats as noise, and the card was back at 90% of its sustained rate within 12 seconds of continuous writing afterwards.

The second reader’s run did 518, 524 and 517 write IOPS and read 22% faster at random than the ProGrade, 1,864 IOPS against 1,523. Random figures move with the reader on this bench, which is part of why the A classes cannot be tested on a USB reader, and the rubric scores the primary reader’s run, where this is a 5.

132 °F at its hottest, under its own 70 °C limit

This is one of the few cards on my bench that prints an operating range, 0 to 70 °C, and the rubric holds it to that maximum rather than the 85 °C it assumes for a card that prints none. At the end of five minutes of uncooled writing in a 76 °F room the card read 131 °F on the primary reader, and the thermal frame taken in the last seconds of that write reads 131.0 °F at its hottest point. On the second reader it read 132 °F. The rubric takes the highest reading on file: 132 °F is 55.6 °C, 14.4 °C under the printed 70, which scores a 3.5.

It did not throttle: the last thirty seconds of the timed write ran at 47.1 MB/s against 46.3 for the first thirty. 132 °F is a cool reading by the standards of my bench, mostly because the card writes slowly, and heat follows work. The row’s 3.5 is the price of printing a range at all: had the pack printed nothing, the rubric would have assumed 85 °C and the same 132 °F would have scored a 5. I would rather a card printed its real limit. 14.4 °C of margin in a 76 °F room is comfortable for a dash cam; a windshield in summer is hotter than a reader with its fan off, so it is a reason not to bury this card in a camera that already runs warm, not a reason to avoid it.

Two readers, one answer

Every card I review runs on my primary reader, a ProGrade PGM0.5, and again on a SanDisk QuickFlow SDDR-B751, which can drive a faster bus mode on cards that support one. This card ran once on each, the ProGrade run first and with the whole-card fill.

The two readers agree to within 1% on every write figure, so the card, not the reader, sets its speed, and the QuickFlow found no faster bus mode to use: the card reads at the SDR104 ceiling on both. Random reads run higher on the QuickFlow on every card I have run on both readers, so that part is normal.

How the 4 breaks down

Both readers give the same rows: the lowest class claim is a 5 at 146.5% or 147.5%, and the worst ten seconds are a 3 at 43.94 or 44.25.

The bottom line

The Speederlash 128GB is a better card than its box says. The pack prints U1 and gets its own capacity footnote wrong; the card prints U3, declares U3 in its registers and holds U3, and V30, through a write of the whole card and after it. It verifies every byte of an ordinary 97.71% of its label, runs cool, and stays inside the 70 °C its pack prints. It is not fast: about 47 MB/s, written in a three-and-a-half-second rhythm a camera never notices, with reads at the standard UHS-I ceiling. For a dash cam, a security camera or an action cam recording at V30 rates it does the job; for anything that wants V60, or quick offloads on a premium reader, look higher.

It is also another reminder that the identity a card reports belongs to a platform, not a product. This card reports the Gigastone High Endurance Pro’s identity, capacity and declared classes to the byte and writes nothing like it, and it shares the PNY Premier-X’s capacity register while sitting between the PNY and the Newegg TeamGroup on the length of its write cycle, its average and its worst ten seconds. Only the bench says which card you are holding. And a no-name brand with a Gmail address is not the same thing as a fake: this one is honest where it matters, in the capacity test and the class marks, and only its box is careless.

Sample size: one unit, purchased sealed at retail from Amazon on 6 October 2026. Two runs on two readers on 21 October 2026 under CKBench protocol v1.27.

Disclosure: I bought this card at retail from Amazon on 6 October 2026. Speederlash had no involvement in or knowledge of this review, 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.