The Steam Deck is spec’ed with a UHS-1/SDHC slot, which means that you can’t use SDUC-class cards and you won’t get much benefit from using comparable cards with a UHS-II/UHS-III bus mark compared to one with a UHS-I mark, even if the other marks otherwise suggest better performance. You can basically ignore the A/V markings because they’re not granular enough to help with comparing cards at this particular performance level (you should instead compare “Random Read”/“Random Write” performance benchmark scores).

Note that there remains a considerable amount of variance among similarly marked cards. For example, the Sandisk Extreme Pro (Bus: UHS-I, Speed: 3) can benchmark write speeds which are almost twice as fast as the Sandisk Extreme (Bus: UHS-I, Speed: 3).

tl;dr: The ideal card will have the following markings:

• Capacity Standard: SDXC (SDUC is not compatible)
• UHS Bus Speed: I (higher is fine, but not helpful)
• Speed Class: 3 (though you should really be comparing benchmark scores instead!)

I can’t say the class or anything like that, but… A lot of the real performance depends on factors that are well outside of any general consumer specified marketing. There used to be a YT conference video by Andrew Bunny Haung about the security of SD cards I would link to. It was only unlisted for years, but has been set to private now. SD cards run an entire internal microcontroller, usually an 8051 clone. These can be hacked, and he showed this. The issue will never get addressed as a security bug either. The microcontroller onboard is managing the actual page blocks in flash memory. A lot of low cost and low spec/smaller storage sizes are the same dies used for top specification units, but have a range of problems that do not fit within a simplified marketing spec. Like if part of the storage area on the die has issues with write speed, depending in the severity of the issue, the slow block may be masked off as if it does not exist, or the entire card may just get sorted and sold at a lower speed rating. This is why bench marks are junk data.

Your best bet is to buy cards from brands that actually own their own memory chip fab. If you also buy their high end offerings and largest capacity, (I know, sounds like I work for these guys as a shill, but I don’t), you are much more likely to get their absolute best performance possible. These will contain the best dies from the fab yield and will not have the types of problems that may exist in smaller sizes and lower classes. They usually do not make several dies specifically for all the various sizes and classes you see for sale. There are only a few dies made, and the defects determine the end rated product.

I bought an “U3 V30 A2” SD card and it works great. I might be wrong, but I think I’ve heard that it’s useless to go anything faster.

Also, the most important thing: buy from a reputable source. There are A LOT of fake SD cards out there.

Samsung Evo 512 is really good for the Steam Deck. Fast & reliable.

I remember reading a reddit post that led me to picking this card. I’ve been using it for a while now. No complaints at all.

https://www.amazon.com/dp/B0BDZ4GRB5

The SDUC standard exists, but no, there aren’t any commercial 2TB microSD cards yet, though they’re coming.

Now I want a 128 TB V90 A2 UHS-III Class 3 SDUC card. Which limb do I need to donate?

SanDisk Extreme Pro is one of the best, in my opinion, they excel in IOPS for small random read and write which is especially important when using it in a computer like Steam Deck or Raspberry Pi. A lot of other SD cards are optimized for cameras where it is writing a lot of sequential data, but are slower in small reads and writes.

Application performance class, launching games is a rendom workload so better random performance is much more important then sequential.

Out of those, application performance class is the one you want. Even better is a real-world random read benchmark.

• The capacity standard isn’t super helpful. Everything from 64GB to 2TB is SDXC, which is supported.
• The Steam Deck only uses UHS-I. It’ll work with UHS-II and UHS-III cards, but they won’t have any meaningful benefits.
• Pretty much any decent microSD Card in 2023 is class 10. If it’s anything else, that’s a red flag.
• Higher UHS speed class and video speed class are probably better, but they’re measuring write performance. For playing games, random read performance is far more important.

The SD Association has this page explaining the speeds, and from what I can see, it’s just to guarantee speeds with certain bus types. Most devices currently use a UHS-I bus, and the older speed classes cap out at pretty pathetic numbers, so they’re kind of useless. For the steam deck, any of the higher-end cards are going to be about the same, and the IOPS are going to be more important in terms of loading times anyway.

Not yet, the specifications are finished but the first SDUC hasn’t launched yet.

1 bit of information in flash memory is stored using one cell. Each cells is very similar to a transistor in a processor. We already know that through years of development we can fit 100s of millions of transistors in a very small space - these days they’re only a few thousand atoms big. If you look at any chip on a circuit board mostly what you’re seeing is the plastic packaging - the actual functional area of the chip (called a die) is much smaller. Even a massive processor like an i7 is only a few millimeters square.

Flash can actually pack in even smaller than a CPU. The structure of cells is much more regular, and they can be packed in 3 dimensions as well rather than just laid out on one flat surface. And that’s how you can fit a trillion memory cells into a micro SD card.

I have no clue, that’s just science gone right. Having so many transistors packed so dense comes with various problems, temperature being the biggest affecting it in lifetime and speed.

Be happy that not every SSD has just one microSD with a header inside.

We don’t use them for everything because they are much, much more expensive than other storage mediums.

2TB cards don’t exist quite yet.

Because there are orthogonal performance properties and making them all get “better” with a higher “performance” number would be very expensive. Someone taking photos wants high capacity. Someone taking 8K video wants high write bandwidth. Someone playing games or as phone app offload storage wants high IOPS.

It’s why yo can buy a 1TB low-IOPS card for $40, or you can buy a 256GB high-IOPS card for $40. But if you want 1TB with high-iops, you’re going to pay about $200.

Different use cases (photo/film/smartphone/console) have different requirements. Just one number would not be enough.