ProGrade PGM0.5 review: your card reader is a measuring instrument

Nobody reviews the thing doing the measuring

I have published a lot of memory card reviews. Every single number in every one of them passed through one small metal box on my desk, and until this week I had never tested it. That is a strange thing to admit on a site whose whole pitch is measurement, so I fixed it.

A card reader is not a cable. It is an active device that negotiates a bus speed, moves every byte, and decides what your card is allowed to be. Plug a fast card into a slow reader and you do not get a slightly slower number. You get a different answer, and nothing on screen tells you which one you are looking at.

This is a review of the ProGrade Digital PGM0.5, a dual slot SDXC and microSDXC reader. It is also, unavoidably, a review of my own bench, because that is what this reader is.

How much faster is a UHS-II reader, really?

Here is the cleanest test I can run. One card, a ProGrade SDXC UHS-II V60 128GB. Two readers. Same afternoon, same machine, freshly formatted before each run, nothing else changed.

Through the PGM0.5 the card read 296.9 MB/s. Through my Transcend TS-RDF5, a competent UHS-I reader, the same card read 95.0 MB/s. That is 3.1 times, and it is entirely the reader.

The write side matters too, and it is the part people forget. Writes went from 108.4 MB/s down to 88.5. Smaller in percentage terms, but it is the direction that decides whether your camera drops frames, and the slow reader caps it.

The slow reader does not just move fewer bytes, it waits longer

Bandwidth is the number everyone quotes. It is not the only thing that changes.

I measure small random reads at queue depth 1, which is a single 4 KB request at a time with nothing else in flight. That is a latency test wearing a throughput costume. Through the PGM0.5 the card managed 2087 of those per second. Through the Transcend, 1014.

Turn it around and it is easier to feel: each individual read took 0.48 milliseconds through the fast reader and 0.99 through the slow one. The interface did not just narrow the pipe, it doubled the wait on every single request. If you have ever browsed a card full of RAW files and watched thumbnails crawl in, that is the number you were feeling, and it is not your card's fault.

Can you really offload two cards at once?

The PGM0.5 has a full size SD slot and a microSD slot, and the first thing I found is genuinely useful: it presents them as two independent drives. Both drive letters appear the moment you plug the reader in and both stay there with both slots empty. A cheaper muxed reader shows you one drive and swaps whatever is inserted underneath it.

So two cards, two drive letters, copy from both at the same time. That is what I assumed. It is wrong.

I put a card in each slot and read from both simultaneously, sampling the actual device counters once a second. Across two separate experiments totalling 330 seconds, the number of seconds where both slots were moving data was zero. Not degraded. Not sharing. Zero.

What actually happens is that one slot seizes the reader and holds it. In the longer test the microSD slot ran alone for 59 seconds, then the full size slot took over and held it for the remaining 181 seconds while the microSD side sat at a dead stop. The starved side kept asking the whole time. It simply did not get served.

The control makes it conclusive. I moved the second card into a separate reader on the same machine and ran the identical test. Both streams ran at full speed simultaneously, for every second of the test. So this is not Windows being cautious. It is one command pipe behind two slots.

The slots are too close together

A small thing that no spec sheet will tell you: with a full size SD card seated, getting the microSD card back out is genuinely difficult. The cards sit close enough that there is not much room to pinch, and the microSD is flush enough that a fingernail slides off it.

I ended up using tweezers. That is not a disaster, and it is only a problem when both slots are occupied, which given everything above is a situation you have less reason to be in than you thought. But if you have big hands or short nails, budget for a little swearing.

The rest of the physical design is fine. The plug is molded straight into the housing rather than hanging off a cable, which removes an entire category of failure and makes it a genuine throw-it-in-a-bag item. It is also the reason one troubleshooting step later in this article was unavailable to me.

The morning my own instrument started lying

I sat down to bench two new UHS-II cards. The first one produced a sustained write trace that looked like a heart monitor: full speed for about twelve seconds, then a complete stop for three or four, over and over, for the whole five minute test. Twenty two percent of the test time, the card accepted nothing at all.

That is a serious result. It dragged the card's worst ten second window down to 53.7 MB/s, below the 60 MB/s floor its printed V60 badge promises. On paper I had just caught a professional card failing its own rating.

Then the second card, from a different manufacturer, did exactly the same thing. Two premium cards failing identically is not a coincidence, it is a clue, so I started eliminating. The cards were not at fault, because a third card failed too. The UHS-II bus was not at fault, because a UHS-I card failed as well. The slot was not at fault, because both slots failed. The USB port was not at fault, because three different ports failed. The hub was not at fault, because plugging straight into the motherboard failed. My own benchmark software was not at fault, because I reproduced the whole thing with plain file copies and Windows' own performance counters.

At that point I had eliminated everything except the reader, and I very nearly published that. I had written the verdict. What saved me was the oldest question in technical support, which somebody asked me before I hit send: had I tried turning it off and on again?

I had power cycled the reader repeatedly. I had never rebooted the computer. I rebooted, ran the identical test on the identical hardware, and the stalls were gone. Completely. The fault was wedged state somewhere in the host's USB stack, and it survived unplugging the device, changing ports, and bypassing the hub, which is exactly why it looked so convincingly like broken hardware.

Why this belongs in a reader review

The animation above is the same card, in the same reader, running the same test twice. One of those runs says it is a mediocre V30 card that cannot hold its rating. The other says it comfortably clears V90, two full classes above what is printed on it. Only one of them is true.

Nothing on screen distinguished them. There was no error, no warning, no dropped device. The card mounted, the copy completed, every byte verified correctly on read back. The fault was pure latency, and if I had been offloading photos instead of running a benchmark I would have noticed nothing except that it felt a bit slow that morning.

That is the actual argument for caring which reader you own and how you verify it. Not because readers usually fail, but because when the measurement chain misbehaves it does not announce itself. It quietly changes the answer and lets you draw the conclusion.

Three things saved those two cards from a published accusation: I keep the raw per second samples instead of just the summary numbers, I own a second characterized reader to check against, and I changed one variable at a time. If you are benchmarking anything, those three habits are worth more than any single instrument.

Does it get hot under sustained load?

The housing is metal, which is the right choice, and there is no temperature sensor to read so this one comes down to touch. After roughly eight minutes of continuous maximum rate reading, with the card streaming at 280 MB/s the entire time, the reader body was warm. Not hot, not uncomfortable, warm.

More usefully, throughput did not sag across that window. A device that is thermally throttling shows you a slow decline. This one held its rate from the first second to the last, which is the answer that actually matters.

What is a UHS-II reader actually worth?

On the same afternoon I ran two very different cards through both readers. One is the UHS-II ProGrade V60 already in this article. The other is a Samsung PRO Ultimate 512GB, a UHS-I card, four times the capacity, and at the time of testing all three cards I used cost roughly the same.

On the UHS-I reader, the two cards are the same card. The UHS-II card read 95.0 MB/s and wrote 88.5. The UHS-I card read 96.3 and wrote 89.2. The cheaper, larger, older-standard card is fractionally ahead, and the difference is inside run to run noise. Every bit of the expensive card's advantage is sitting on the other side of the reader.

Put a UHS-II reader in front of them and they separate immediately: 296.9 against 169.8. But the number that actually settles the argument is not read speed at all, it is sustained write, because that is what decides the video speed class your camera can record at.

In a UHS-II host the ProGrade card holds V90. Its worst ten second window was 102.8 MB/s, comfortably over the 90 floor. The Samsung, given the UHS-I bus a camera would give it, manages 88.7 and holds V60. It misses V90 by 1.3 MB/s.

That is a whole video speed class, not a percentage. It is the strongest argument for UHS-II hardware I have measured, and it is the reason to own a reader that can actually reach it.

Should you buy the ProGrade PGM0.5?

Buy it if your cards are UHS-II. That is the whole case, and on a UHS-II card it is emphatic. Three times the read speed is not a tweak, it is the difference between waiting for an offload and not waiting for one. If you shoot volume on V60 or V90 cards, this pays for itself in minutes saved.

Buy it if you own a card with a proprietary fast mode. Some UHS-I cards go faster than the standard permits when the reader speaks the same extension, and this reader does. A Samsung PRO Ultimate 512GB read 169.8 MB/s through the PGM0.5 and 96.3 MB/s through my UHS-I backup. That is the same card, and the extension alone is worth 76% more read speed and 54% more write speed. My own analyzer placed the 169.8 within 5% of the DDR200 ceiling, so it named the extension before I did. A basic reader would never show you any of it.

Do not buy it for ordinary UHS-I cards. A standard card tops out around 96 MB/s no matter what you plug it into, and my cheap backup reader hits that number too. If everything you own is a regular V30 microSD, this reader will change nothing and you should spend the money on a better card instead.

Do not buy it expecting parallel offloads. See above. If that is the feature you are shopping for, buy two readers.

What I could not test

Being straight about the edges of the data, because a review that only tells you what it proved is half a review.

Two readers is a small sample. Every comparison here is against one specific UHS-I reader. It is a well characterized one that has agreed with the PGM0.5 to within 0.7% on cards where both are on equal footing, which is why I trust it as a control, but it is still one comparison point rather than a field.

I could not test the microSD slot at UHS-II speeds. ProGrade states both slots are UHS-II. I have no UHS-II microSD card, so that claim is the manufacturer's, not mine. Everything I measured on the microSD side ran at UHS-I speeds because that is all the cards could do.

I could not separate the reader from the card at the top end. The fastest read I recorded came within 5% of the theoretical ceiling for this interface, and my own tooling flagged it: at that point the reader may be the limit rather than the card. Answering that needs a faster card than I own.

The wedged host state was never explained. A reboot cleared it and it has not returned. I do not know what caused it, and anyone telling you they would have diagnosed it faster has not spent an afternoon eliminating nine other suspects first.

Worth it, with one honest caveat

I would buy this reader again, and the numbers are the reason: 3.1 times the read speed and half the latency on a card that can use it, from a small metal box with no cable to lose. If you own UHS-II cards and you are still reading them through whatever came in a bundle, this is one of the few upgrades where the improvement is immediate and obvious.

Two caveats stay attached. The dual slot is not the parallel offload feature it looks like, and you should plan around one card at a time. And a reader remains a device you cannot audit while you rely on it, which I now know from the inside rather than as a theory. That is not really a mark against this product. Every reader has that property. This is just the only one I know of where somebody has published what happens when it goes wrong.

If you want to see what this reader measures when it is behaving, every card on the CKBench bench was tested through it, and the scoring method explains what the numbers mean. The counterfeit SanDisk is the one that best shows why measurement integrity is not an academic concern.