Memory cards: Specifications and (more) deceptions


Back in September 2022, I told you about two microSD cards that weren’t as advertised, capacity-wise:

A bit more recently (at the beginning of this year, specifically), I unsuccessfully tried to get inside one of them:

And now…well, it’s happened again, this time with an SD card. Back at the end of December, I bought a “Used-Very Good” condition 128 GByte SDXC card from Amazon’s Warehouse section for $27.45, versus $32.83 (its current brand-new price on Amazon as I write this):

This was the second “Warehouse”-sourced acquisition of the same card (vendor, capacity and performance specs) that I’d made in ~1 month, following in the footsteps of a “Used-Like New” purchase (ironically for $0.10 less) in late November. And at the same earlier time, I’d also bought two brand-new cards “Black Friday”-priced at $27.63 each. At first, all four cards seemingly worked fine. The only discrepancy I’d noticed was that the “Used-Very Good” card came absent its packaging, instead housed solely within a clear plastic “baggie”, but it had been advertised as repackaged, so I didn’t think anything of it.

Prompted by the advertised-vs-true capacity discussion I had with reader “Ducksoup_SD” as a follow-up to that earlier mentioned January 2023 writeup, I gave in to curiosity and did full-reformats (versus default “quick” formats) on all four PNY SD cards (plus four used Sony ones I’d subsequently bought from B&H Photo Video for another project, details of which I’ll save for another time) to ensure that they actually delivered the promised storage capacity in full.

The late-December-acquired Amazon Warehouse-sourced PNY passed format but seemed to complete slower than its peers, which was strange. Its label also fell off when I pulled it out of the computer’s SD card slot: more strangeness. So, further feeding the curiosity beast, after clumsily gluing the label back on, I ran Blackmagic’s Disk Speed test (here’s a direct link to the utility on Apple’s App Store) on all four of them. Here’s what I got on the two new ones and the “Used-Like New” one:

And here are the results on the “Used-Very Good” one, which previously had seemingly reformatted more slowly than the others:

Notice the disparity? All four cards delivered comparable read speeds, but the fourth card’s write performance was ~25% lower than the others. As soon as I flipped it and one of the other cards over, I had my answer:

Again, see the difference? Before continuing, I’ll also share a photo of both cards’ front sides:

The under-delivering “Used-Very Good” one is on the left in both pictures; you’ll need to take my word that it originally looked identical to its full-performance peer to the right. The label degradation you see was solely the result of my earlier-mentioned re-glue clumsiness.

To explain what I think is going on, here’s some preparatory background. While the underwhelming write performance may be a minor annoyance when you’re formatting a memory card, it’s a huge issue when you’re trying to sustainably camera-capture “raw” or other high bitrate-formatted 4K or higher resolution video (which is precisely what I’d bought these cards for), for example. The correctly outfitted card on the right is a UHS-II model; its second row of signal contacts, in combination with the earlier SD-to UHS transition to low-voltage differential signaling, enables it to deliver highest-possible interface transfer speeds (currently, at least; there’s also a UFS-III spec but I haven’t seen any cards based on it yet). The other one has a more conventional single row of apportioned contacts. Compare the two and you’ll likely come up with at least two correct conclusions:

  • UHS-II cards are intentionally designed to be backwards-compatible with UHS-I (and precursor) card readers, albeit running at lower transfer speeds in the process, and
  • The card slot in the system I used for my benchmarking, an early 2015 13” MacBook Pro, is obviously UHS-II cognizant, otherwise the correctly implemented cards wouldn’t have performed better than their slower sibling did.

So, what happened here? I suppose this could have been a screwup on PNY’s part from the get-go, sticking the wrong label on the card way back at the factory. But more likely, I suspect (particularly given that the label on the misbehaving one fell off on me), is that this is the latest in a long line of storage scams that have victimized many folks. Back in January, for example, I told you about a ripoff from mid-last year involving Walmart (inadvertently, I assume) selling supposedly 30 TByte portable SSDs for $39. Well, subsequent to my January writeup’s publication, another scam got lots of coverage: fake 16 TByte SSDs on Amazon for $100.

My guess? Someone printed up a bunch of fake PNY labels, stuck them on unknown-source SD cards (correct-capacity ones, at least) and returned them to Amazon, keeping the legit ones they’d previously purchased. I got one of the fakes. Who knows, frankly, how many times this particular card has circulated through Amazon Warehouse’s buy-return-resell (lather, rinse and repeat) cycle, and how many of these fake cards ended up unknowingly (and permanently) in scammed buyers’ hands. To wit, I almost didn’t bother returning the card, out of concern that Amazon might just turn around and resell it even though I’d documented its definitive flaws in my return-request submission. Instead, I thought about instead keeping it to add to the teardown pile; in retrospect, had I done so, I might have also been able to discern info about its origination via a perusal of its S.M.A.R.T. data using a utility such as CrystalDiskInfo.

More generally, the specs associated with the microSD and SD cards, and therefore the markings on the labels of them, are IMHO frankly a mess. In addition to the aforementioned bus interface evolution and options (default SD, high speed SD, and UHS-I, UHS-II and UHS-III, along with SD Express in the future) there are four different capacity range classifications: SD (up to 2 GBytes), SDHC (2-32 GBytes), SDXC (32 GBytes-2 TBytes) and SDUC (2-128 TBytes). And there are currently three different sets of media speed classifications, all of which overlap each other:

  • Original Speed Class (2, 4, 6 and 10)
  • UHS (presumably “Ultra High Speed”) Speed Class (U1, U2 and U3, which are different than the previously discussed UHS-I, UHS-II and UHS-III interface speed options), and
  • Video Speed Class (V30, V60 and V90)

See for yourself:

Further muddying the waters are various proprietary memory card implementations. Sandisk, for example, sells a single-row contacts family that looks like a UHS-1 form factor and therefore should max out at V30 transfer rate performance (in fairness, Sandisk does label them as such). But the company touts them as delivering up to 200 MByte/sec read and 140 MByte/sec write speeds. That’s because they optionally support a Sandisk-only DDR interface transfer mode which, to the best of my knowledge, is only comprehended by a few Sandisk-branded card readers; in industry-standard card slots they run at 104 MByte/sec max UHS-1 speeds.

And don’t get me started on all the other high-capacity and/or high-performance removable memory card form factors and spec options, industry standard and proprietary alike, that are now contending for consumers’ wallets, such as the Compact Flash Association’s CFast and CFexpress, the latter in both Type A and B variants…sigh. I could dive down into the next level of spec minutia, complete with more rants, but I think I’ll spare both you and my poor associate editor colleague the incremental wordcount and associated angst. Thoughts, supportive or not, on my situation, conclusion and overall industry observations? Sound off in the comments!

Brian Dipert is the Editor-in-Chief of the Edge AI and Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company’s online newsletter.

 Related Content


Scroll to Top