Memory card interfaces keep pace with the internal bus evolution race: Part 2

Learning from the past is wise, as long as it’s not taken to excess. So, too, is adopting others’ ideas (in a non-infringing way). The post Memory card interfaces keep pace with the internal bus evolution race: Part 2 appeared first on EDN.

Memory card interfaces keep pace with the internal bus evolution race: Part 2

Learning from and adapting the lessons of the past is wise, as long as it’s not taken to overly constraining excess. So, too, is adopting others’ ideas (in a non-patent-infringing way, of course).

As you already know if you read last week’s blog post (and if not, please do so first before continuing with today’s…I’ll be right here, waiting for your return…), I initially planned on covering this topic in a single writeup. It ended up, however, being at least twice as long as I’d originally envisioned, so I basically chopped it in two. Part 1 covered the historical precedents that led to the ongoing memory card innovations of more modern times, which I’ll discuss this time.

Interface evolutions

I’m spending all this time on past-history factoids and trends because, as you’ll soon see, they conceptually continue(d) to repeat themselves multiple times over with the passage of time. To that point, one other historical example, involving performance, also bears mention. PCMCIA, introduced in 1990, tackled a mid-life enhancement five years later, from the 16-bit ISA bus-derived PC Card to the PCI bus-based and 32-bit, but still backwards-compatible, CardBus.

A more radical transformation, ExpressCard (originally called NEWCARD), followed roughly a decade after that. Based on the combination of PCI Express and USB 2.0, it was not directly backwards compatible with CardBus, far from with PC Card, thereby either forcing systems adopters to include slots for both standards in designs or forcing users to use clumsy adapters:

More generally, as my attempted blending of two Wikipedia entry excerpts notes:

Despite being much faster in speed/bandwidth, ExpressCard was not as popular as PC Card, due in part to the ubiquity of USB ports on modern computers. When the PC Card was introduced, the only other way to connect peripherals to a laptop computer was via RS-232 and parallel ports of limited performance, so it was widely adopted for many peripherals. More recently, virtually all equipment has Hi-Speed USB ports, and most types of peripherals which formerly used a PC Card connection are available for USB (and have the advantage of being compatible with desktop computers as well as portable devices) or are built-in, making the ExpressCard less necessary than the PC Card was in its day.

Wash, rinse, repeat

Let’s now fast-forward to more modern times. CFast, short for CompactFast, which I mentioned in both of my 2023 writeups (in the context of their use by my Blackmagic cameras), is based on CompactFlash (and is also managed by the CFA) but migrates from ATA to SATA. CFast 1.x dates from 2009 and is based on SATA 2.0; the backwards-compatible CFast 2.0 upgrades to SATA 3.0 but has seen limited-at-best industry uptake since being initially unveiled in 2012.

Why? Enter, for example, the alternative CFexpress, also managed by the CFA, which switches from SATA to the solid-state media-optimized NVM Express (i.e., NVMe) as its command set and to PCI Express (PCIe) as its hardware interface foundation (as I’d mentioned at the end of 2023), as well as coming in multiple dimensional options. The smaller Type A (at left in the following image) and larger Type B (right) card variants are today commonplace in the industry, with the even larger Type C conversely not yet in production to the best of my knowledge:

In this context, an overview of the earlier XQD standard also bears mention. XQD, once again now managed by the CFA (albeit initially announced solely by Sandisk, Sony and Nikon), dates from 2010. It’s dimensionally and connector-compatible with CFexpress Type B and is also based on PCIe, albeit only in a single-lane implementation (with PCIe 3.0 support added with XQD 2.0 in mid-2012). The XQD and CFExpress standards are therefore cross-compatible, although only to a degree, generally requiring firmware updates which not all camera, memory card reader and other system manufacturers have provided.

CFexpress 1.0, announced by the CFA in September 2016 as the successor to XQD, launched with support for PCIe 3.0, albeit this time in higher-bandwidth dual-lane form (for the size option now known as Type B and used by my high-end Canon and Panasonic cameras, among others). CFExpress 2.0, following in February 2019, added the single-lane PCIe Type A and quad-lane Type C options, along with upgrading the NVMe command set from 1.2 to 1.3. And the latest iteration, August 2023’s CFexpress 4.0, upgrades the supported PCIe interface to 4.0 (again, at up to four lanes with Type C), and the NVMe command set to 1.4. CFExpress 4.0-optimized systems are not yet in the market, to the best of my knowledge, but cards (such as this OWC Atlas Pro) are prevalent and backwards-compatible with existing cameras and such:

No, I don’t know what happened to CFexpress version 3.0, either. While buying a CFexpress 4.0 card now will leave potential performance “on the table” with CFexpress 2.0-only systems, it does provide obsolescence protection for subsequent camera-or-other upgrades you might make in the future. And conversely, if future-proofing isn’t a concern, you’ll be able to (as I’ve personally done) get some great deals on CFexpress 2.0 memory cards right now, despite overall semiconductor memory supply constraints, as manufactures strive to “fire sale” deplete their inventories of legacy product variants.

Don’t count out Donkey Kong

And what about the SD and related microSD card standards; are they in danger of falling by the wayside as these high-performance newcomers ramp into the market? Not if the SD Association has anything to say about it, specifically with next-generation “Express” offerings. See if you notice anything familiar trend-wise in the paragraphs that follow:

When the SD Association (SDA) first announced SD Express in June 2018, it set the bar high and opened a world of possibilities for manufacturers to integrate supercharged removable storage into their designs. SD Express is capable of delivering SSD performance levels of up to 4GB/sec. This makes it perfect for use in high-performance electronic devices and products. With the introduction of advanced security features in May 2022 found in the SD specification version 9, performance and versatility merge to create an innovative, and advanced powerhouse solution for SD memory cards.

SD Express leverages the PCI Express and NVMe interfaces and uses the well-known SD memory card form factor for compatibility with existing SD slot architectures. The SDA also introduced a microSD Express memory card format that is backward compatible with devices. SD Express is not just about SD memory cards getting faster, it is also about SD memory cards doing more.

After languishing for several years awaiting market demand that stubbornly refused to emerge, “Express” variants’ fortunes are finally looking up. Specifically, the microSD Express card is used in the Nintendo Switch 2 game console, notably (and singlehandedly) increasing the likelihood of a high-volume long-term future for the standard.

Blazing a trail

I’ll wrap up this writeup with coverage of a recently emergent sole-source memory card option (in spite of my earlier comment that I planned to avoid diving into past-history proprietary offerings) that I’d earlier caught mention of at The Verge and elsewhere. It’s Biwin’s Mini SSD:

Biwin is, if you hadn’t already guessed from the coin at left in this “stock” image, a China-based memory subsystem manufacturer (to the right of the 1-yuan coin is the rare U.S. $1 coin). Most of the products on the company’s website are industry standards-based: PCIe NVMe internal SSDs, for example, along with USB flash sticks and drives, DRAM DIMMs and SoDIMMs, SD/microSD and CFexpress memory cards (an image of which you saw earlier), and memory card readers. But with the Mini SSD, the company has apparently decided to try its hand at also going proprietary.

Interestingly, the Mini SSD is slightly larger (at 15x17x1.4 mm) than the microSD Express (15x11x1 mm) counterpart. And at least from a latest-generation ratified-spec standpoint, it’s seemingly no faster than microSD Express, either; both are based on dual-lane PCIe 4.0 and NVMe (once again: sound familiar?). The key differentiator that Biwin seems to be betting on is timing; as Ars Technica notes, currently available microSD Express cards “top out around 900MB per second, roughly the amount of bandwidth available from a single PCI Express 3.0 lane.”

Conversely, Biwin was demonstrating functional products at CES in January, claiming read speeds up to 3,700 MB/s and write speeds up to 3,400 MB/s (at least in combination with the company’s own card reader peripheral), and with capacities ranging from 512 GB to 2 TB. Biwin also touts Mini SSD’s IP68-rated dust- and water-proof chops. One note: while the company was referring to them as the “BL100” series late last summer, it’s now calling them “CL100”. Why?                                                             <div class= Read Original