PNY Premier-X 128GB: an honest card at the entry price

Is the PNY Premier-X any good?

PNY is the brand you find on the peg next to the SanDisk, a little cheaper, in a blister that looks like it was designed in 2020 because, according to the copyright line on the back, it was. The Premier-X is the entry line: 128GB, a 100 MB/s read on the front, C10, U3, V30 and A1, and on the back the sentence that made me buy it: Up to 100MB/s read speed, write speed is lower. No figure. That makes PNY the third vendor on my bench to state a read and decline to state a write, after SanDisk on the Ultra line and Lexar on the Blue, and across those cards the unstated write has turned out to be anywhere from 13% of the read claim to 96% of it. The footnote is true of all of them and tells a buyer nothing about which one they are holding.

So I bought one, the same day I bought an Amazon Basics 128GB, as a deliberate pair: two 128GB microSD cards, both C10 U3 V30, a few dollars apart, one from a brand that prints a lot on its box and one from a brand that prints almost nothing. I ran the PNY four times in one day, twice with a fill of every byte on my primary reader, once more on the same reader, and once on a second reader that can drive cards faster than the first. I read its identity registers on a Raspberry Pi. This review is what came back, and the answer to the question in the heading is yes, with one thing you should know about how it writes and one thing I did not expect to find about who built it.

What the card claims, and what the card did

Two rows need their reasoning attached. The read is met, not beaten, and could not have been beaten on this hardware. A standard UHS-I card runs a bus mode called SDR104 whose practical ceiling on a card reader is about 97 MB/s; every card on my bench that claims 100 measures 96.4 to 96.9 on it, and the SD Association’s own figure is the bus, not the flash. This card measured 96.8 and 96.9 on two readers, one of which can drive a faster non-standard mode and took the Amazon Basics to 206 the same morning, so there is no faster mode hiding in the PNY. The 100 is the honest number for the bus it speaks. A1 is not testable on any USB card reader, mine included: its floors assume a command queuing mode that USB readers do not perform. The card is neither credited nor penalised for it, and the numbers in the row are there because they are the closest thing this bench can measure, not because they decide anything.

A blister that prints more than cards costing twice as much

The Premier-X comes in a sealed retail blister with an SD adapter in a clear cradle. The front carries the marks: 128GB, 100 MB/s, C10, V30, U3, UHS-I, A1, 4K Ultra HD. The back is where PNY earns a paragraph. In three languages it prints the read-speed footnote, the capacity definition, the A1 line, the choking-hazard line and a photo-recovery software offer, then a warranty term (lifetime, limited), a model number (P-SDU128V31100PX-GE), the UPC, addresses for New Jersey, France and Taiwan, technical support numbers, and Made in Taiwan. Four proof icons sit along the bottom, temperature, shock, magnet and water, with no ranges attached to any of them.

I keep a running comparison of what cards print against what they cost, and this one is the low end of an inversion that keeps showing up: the least expensive branded 128GB card on my bench has the most complete back panel of the recent arrivals. That is worth exactly nothing if the card fails, and the corpus has cheap cards that did both ways: the Newegg TeamGroup was honest on capacity and failed its speed claim, and the three generic cards were all fakes. So the box gets noted and the bench decides. The card arrived formatted exFAT with 128 KB clusters, no volume label, and nothing on it.

Five minutes at 54.9 MB/s, in four-second heartbeats

This is the panel that decides a video class, and on this card it is the one to watch rather than read. The trace below is the scored run: five minutes of committed writing, one sample a second, with the speed-class floors marked. Most cards draw a line. This one draws a comb. It writes at 80.7 MB/s, a rate well above its own average, for a little under three seconds, then drops to about 3 MB/s for a second and a half, then does it again, every 4.2 seconds, 71 times in five minutes. The average across all of it is 54.9 MB/s, the burst in the first 30 seconds is 53.3, and the worst ten seconds average 47.4. Three more runs, one of them on a different reader, put the worst ten seconds at 47.2, 47.4 and 47.1. This card is nothing if not repeatable.

What you are looking at is a write cache. The card accepts data into a fast buffer of roughly 220 MB at 81 MB/s, and when the buffer is full it stops taking new data, almost completely, while it moves what it holds into the main flash. On this card that pause is short and hard: for about a second and a half it accepts one megabyte roughly every third of a second, which is where the run’s longest single block, 355 milliseconds, comes from, and then the buffer is clear and the next burst starts. Nothing in the five minutes came close to a real stall; the bench flags any block over half a second, and there were none on any of the four runs. The pattern held for the entire whole-card fill too, all 125 GB of it, with the speed across each minute of the fill sitting between 53.2 and 54.1 MB/s from the second minute to the last.

Two things follow for anyone deciding whether to buy it. The first is that the ten-second averages the speed classes are built on, and the worst of them, are honest about this card: 47.4 is what a recording device sees if it buffers even a few seconds of video, and every dash cam and camera does. The second is that the gap between the worst ten seconds and the average is not a slow patch somewhere in the run. It is the phase of the cycle: a ten-second window covers about two and a half heartbeats, and sometimes it catches three pauses instead of two. The best ten seconds of the same run average 60.1, and any full minute of it averages 53.5. I score the worst ten, as the rubric says, and it lands in the same band either way. For scale, the Botslab 128GB that came bundled with a dash cam floors at 53.9 on this bench, the SanDisk MAX Endurance at 51.6, and the Kioxia Exceria G2 that anchors the lower end of my control band at 62.6. This card is a step below all three, and a step above the FitcamX at 43.8 and the Newegg TeamGroup at 35.6.

What scattered writes do to it

Loop recording is not one long write. It is a long write interrupted constantly by small ones: event clips, directory updates, overwrites of the oldest file. Random 4K writes are the bench version of that abuse, three twenty-second passes of it, and the question is whether the third pass is as fast as the first. On the scored run the answer is yes: 457, 433 and 465 write IOPS, which is a 7% dip in the middle pass and a third pass that was the best of the three. The card came out of the churn leg and back to its full media rate inside the first ten seconds of continuous writing. Those are small numbers, and I will come back to that.

Now the small numbers. 452 write IOPS is 1.85 MB/s of random 4K writing, which puts this card in the bottom cluster of my bench alongside the Botslab (1.87) and the Newegg TeamGroup (1.89), where a SanDisk High Endurance manages about 4 and the Amazon Basics 8.5 on the same reader. At that level the fade measurement is mostly noise: the difference between a good pass and a bad one is a tenth of a megabyte a second, and across the four runs the first-to-third change read minus 2%, 5%, 13% and 7%. The rubric scores the primary reader’s scored run, which is the first of those, and the row is a 5; the other three runs would put it at 4.5, 4 and 4.5, and the overall score is 4.5 whichever run you pick, because the mean never moves far enough to change the rounding. I am telling you all four so the row does not read as more precise than the test is. What the four runs agree on is the part that matters for a dash cam: churn does not leave this card slower than it found it, and it is back at full speed within seconds.

Every byte, written and read back, twice

The bench wrote 117 files of unpredictable, dedup-proof data until the card was full, 125,062,610,944 bytes, then read every one of them back and compared. Zero mismatched bytes. Zero read errors. Then, four hours later, it did the whole thing again with a fresh random seed: 125,063,135,232 bytes that time, the difference being where the last file landed, and again zero mismatches and zero read errors. The write speed across the first, middle and last tenth of the card spread by 3%, all of it from a slow first minute as the card cleared its housekeeping from the five-minute leg before it, and the read spread by 0.3%. There is no slow region, no cache running out partway through, and the fill ran the same four-second heartbeat for all 125 GB.

125.1 GB is 97.7% of the 128 GB on the label. PNY’s own footnote says user-available storage may be less than listed, and it is, by 2.9 GB, which is ordinary provisioning overhead and sits inside the 96.7% to 100.1% range every genuine card on this bench occupies. It is at the lower end of that range: SanDisk declares 99.9 to 100.1, the Longsys-made cards 98.2 to 99.1, and this card’s exact declared byte count, 125,069,950,976, matches one other card on my bench to the byte, the Gigastone High Endurance Pro 128GB, which the identity section explains. Windows will show you about 116 GB, because it counts in binary; that is not a shortfall, it is arithmetic, and the methodology page has the conversion. The capacity row of the rubric scores the declared share of the label, and 97.7% is a 4.

129 °F, and no rating to measure it against

PNY prints a temperature-proof icon and no range, so there is no published maximum to hold this card to; the rubric assumes 85 °C (185 °F) for a card that publishes nothing, which is the figure most of the industry uses. Against that, the card is comfortable. At the end of five minutes of uncooled writing in a 76 °F room the thermal imager read 129 °F (53.9 °C) on the card body at the reader’s slot mouth on the scored run, and 125 °F (51.7 °C) on the two later runs, one on each reader. That is 31 °C of headroom, and it puts the Premier-X among the coolest cards I have measured: the Amazon Basics bought the same day peaked at 171 °F, the SanDisk MAX Endurance at 140, the Botslab at 135.

It did not throttle. The last thirty seconds of the scored run averaged 55.7 MB/s against 53.3 for the first thirty, and the cache cycle at the end of the run is the same shape as at the start. The reading is a floor rather than a ceiling, because the imager sees the part of the card outside the slot and the controller is inside it, but a card that is writing at 55 MB/s and sits at 129 °F on the surface is not working hard. In a dash cam on a summer windshield the temperature it reaches will have more to do with the car than with anything in this test; what this test says is that the card brings a lot of room to that situation.

Two readers, one answer

Every card I review runs on my primary reader, a ProGrade PGM0.5, and any card that might be limited by it runs again on a SanDisk QuickFlow SDDR-B751, which drives the fast non-standard UHS-I mode at a higher clock and took the Amazon Basics from 156 MB/s sustained to 188 the same morning. The PNY went on it last. Here is the same card, the same day, on both.

This is what a card that is not reader-limited looks like. The PNY speaks standard UHS-I and nothing faster, it writes at half the rate either reader can carry, and so the reader you own changes nothing about it except the random-read figure, which moves with the reader on every card I have tested. If you have read my reader tiers post and wondered whether it applies to a card like this, the answer is that a plain UHS-I reader will get everything this card has. The reader guide still matters for the faster cards in the same slot.

Who actually makes the PNY Premier-X

Every SD card carries a set of identity and capability registers, written at the factory, that a USB reader cannot show you but a native SD host can. I read this card’s on a Raspberry Pi. The identity register is the interesting one, and not for the reason I expected. Its manufacturer ID is 0x74, the OEM field reads J`, the product name field reads simply SD, and the revision is 6.1. The manufacturer ID is one PNY does not appear on any public list under; the lists that exist, all of them compiled by hobbyists reading cards because the SD Association keeps the assignments confidential, attribute 0x74 to Transcend, another Taiwanese company. PNY’s blister says the card was made in Taiwan and gives an office in Zhubei. Whether that means Transcend assembles the Premier-X for PNY, or PNY’s Taiwan operation writes an ID a list has filed under a different name, I cannot tell from here, and I am not going to guess. The register is now in my manufacturer ID register as what it is: the first 0x74 on the bench. What it shares with four cards that report 0xFE, byte for byte, is the mirror-image case in the identity explainer.

The other three registers are where the surprise is. The capability registers, the ones that describe the card’s capacity, its bus support and the classes it declares to a host, are written by the controller firmware rather than by whoever prints the label, and on this card they are byte-for-byte identical to a family of cards I have tested before under four other names: the Newegg TeamGroup 128GB, the Botslab 128GB, the FitcamX 64GB and the Gigastone High Endurance Pro 128GB, all of which report manufacturer ID 0xFE. The SD Status register matches all four of them across all 64 bytes, including two fields that no other card on my bench sets to those values. The bus register matches the Botslab, FitcamX and Gigastone exactly. The capacity register matches the Gigastone’s to the last byte, which no two cards on that platform had managed before: the Gigastone review notes that its four cards declare four different capacities, and this fifth card declares the Gigastone’s. The product name, SD, is the TeamGroup’s and the Gigastone’s string. Even the serial number fits the family: every major-brand card I have dumped carries a nine or ten digit serial, and these cards carry four and five digit ones. And then the card writes exactly the way that family writes.

All five cards take data into the same size of cache at the same 80-odd MB/s. What separates them is what happens when the cache folds into the main flash: the TeamGroup drops to 5 MB/s and stays there for most of its cycle, the Botslab and the Gigastone drop to a second, slower rate of about 52, and the PNY stops almost dead but only for a second and a half, which is why its average lands between them. One more difference is worth naming because the Gigastone review makes so much of it: that card’s floor gave way after its first whole-card fill, from 59.6 to 31.0, until a native erase restored it. The PNY’s did not. Its third run, after two whole-card fills and 299 GB written, read the same 47.4 as its first. Whether that is firmware, the number of flash dies, or the grade of the flash, the bench cannot tell, and I am not claiming it can. What the registers do establish is narrower and, I think, more useful: the identity register names the company that assembled the card, not the platform inside it, and this is the first card on my bench where those are visibly two different things.

PNY Premier-X or Amazon Basics?

I bought these two cards together to answer one question: what does the step up from the entry-price branded card to the house-brand card buy you? Both are 128GB microSDXC, both print C10, U3 and V30, and they were tested on the same day on the same two readers under the same protocol. They score the same, 4.5, and they get there from opposite directions.

The step up buys a great deal of speed: three and a half times the sustained write and twice the read on a reader that can drive the Amazon card, three times the write and 1.7 times the read on a plain one, and four to six times the random-write rate. It also buys a card that runs 42 °F hotter against a lower published limit, and a box that tells you less. If your device writes V30 video and you never offload a full card in a hurry, the PNY gives up nothing you would notice and runs cool doing it. If you offload often, shoot above V30, or want the fastest UHS-I card you can put in a phone or a drone, the Amazon Basics is the better card and it is not close. The PNY loses its points on speed and capacity; the Amazon Basics loses its on heat. Which of those you would rather live with is the whole decision.

How the 4.5 breaks down

One line in that block deserves a plain-language footnote. This is the first review scored entirely on the fixed tables that the methodology page published this month, with no judgment call in any row, and it shows what those tables do: a modest card that keeps its promises scores three 5s and two 4s, and the two 4s are exactly the two things a buyer would want to know about it.

The bottom line

The PNY Premier-X 128GB is the kind of card that makes a review like this one feel almost unnecessary, which is the highest compliment the bench can pay a blister. It prints six claims and keeps all six, it declines to print the one number it would have had to be modest about, it verifies its capacity twice over, and it runs cooler under load than anything else I bought this month. It is also slow in a specific, measurable, repeatable way that no marketing sheet would ever show you: a four-second heartbeat of burst and pause, averaging out to an upper-V30 card, on a platform I have now met under five names.

If your device asks for V30, and most do, this card has it with room to spare and a name on it with a lifetime warranty behind it. If you want the fastest card the slot can take, or you fill and empty cards for a living, the same money and a little more buys a great deal more speed from the card I tested alongside it. Neither answer is a mark against the PNY. It is what its box says it is, and on this bench that turns out to be rarer than it should be.

Sample size: one unit, purchased retail from Amazon on 15 September 2026, sealed. Four runs on two readers on 17 September 2026 under CKBench protocol v1.21.

Disclosure: I bought this card at retail from Amazon on 15 September 2026. PNY 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.