Partly because of my recent big update to the Z9 book, partly because I have a workshop coming up with lots of new Z9 users, partly because the the number of people encountering “issues” keeps rising as cameras get more sophisticated, I’m getting a lot of questions about write speeds of CFexpress Type B cards.
Let me just state it up front: the label on the front of the card mostly lies to you, so ignore it. Basically, the label often only identifies maximum read speed of a card. So a card marked 1700MBs should be regarded as one that has a maximum burst read speed of 1700MBs. If you try to ingest a whole card worth of images, you’re not going to hit that number, though you may come close with some cards.
But notice I wrote “read speed.” The write speed will be lower, almost always. And even on cards that do have the maximum write speed on the label, you still aren’t getting the information you probably need.
I’m going to use the latest announcement from Wise as an example. They just announced Mark II versions of their 512GB and 1TB cards (to be available in March sometime). Here are the numbers they put in their press release:
- 512GB original: 1700MBs max read, 1550MBs max write, 400MBs minimum sustained write
- 512GB Mark II: 1700MBs max read, 1550MBs max write, 800MBs minimum sustained write
- 1TB original: 1700MBs max read, 1550MBs max write, 850MBs minimum sustained write
- 1TB Mark II: 1700MBs max read, 1550MBs max write, 1300MBs minimum sustained write
Note that this will get confusing in the retail store, as all four cards are only identified with the maximum read/write speed, which didn’t change. You need to look for the faint M-II next to the card size.
Let’s first deal with why Wise says the Mark II cards are faster at sustained work: they use NAND chips made with a smaller process size. The benefits are faster access and lower heat.
But you’re still wondering about maximum speeds versus minimum sustained speeds. The Z9 worst case is when it is recording 12-bit 8K raw video. It’s a little unclear what the minimum sustained bar for that is—more on that in a bit—but it’s higher than 700MBs in continuous write speed. Likewise, when you take stills at maximum frame rate, you’ll hit the initial buffer at about 80 Lossless Compressed raw images, but what happens after that depends upon the minimum sustained write speed of the card. With “fast” sustained speed cards, you’ll get perhaps 2 to 4 fps of reliable continued use after the buffer fills. In other words, four seconds of 20 fps to buffer full, 2 to 4 fps of reliable continued use with a full buffer. I do see a hiccup every now and then in the full buffer situation with some fast cards, but still, the camera doesn't shut down, and you get a near continuous stream. Moreover, take your finger off the shutter release for a couple of seconds and you’re back to full speed for a full buffer run.
With cards that don’t have fast minimum sustained rates—I’ve seen cards with as low as 125MBs sustained—what tends to happen is you still get the 80 frame raw buffer, and then you get a slower and much more sporadic sequence after that. Click, wait, click, click, wait, wait, click, etc. You’re not getting a regular motion recording after the buffer fills.
Video is more pernicious. Particularly the high bit demands the Z9 can put on a card. Perhaps you’ve started recording video and had the camera terminate the video early. Almost certainly, the card isn’t keeping up. Video at the higher levels requires a fast sustained rate card; as I noted before, probably 700MBs or faster.
Why am I equivocating on a sustained speed bar? Because here’s another dirty secret that no one’s telling you: the maximum/minimum write speeds of a card drop as it fills. All the manufacturer-released performance specs on cards I’ve seen have to do with “initial performance with an empty card.” Larger capacity cards tend to do better maintaining their performance as they start to fill because they have more chips to spread the load over. How that load is spread in the card varies some from maker to maker, depending upon the memory controller they’re using. But pretty much every card I’ve looked at will have worse performance as it starts to fill, it’s just a difference of where that performance degrades.
What’s the takeaway? If absolute buffer speed and/or raw video performance is of any interest to you: (1) buy larger cards than you need and don’t come close to filling them; and (2) make sure that card is capable of >700MBs minimum sustained write speed.
A note to card manufacturers: you’re doing yourself and your customer a huge disservice. Yes, putting big numbers on the label and promoting on price will net you more short term sales and maybe even market share for awhile. But you’re annoying your most demanding customer, the one that recommends cards to everyone else (e.g. Thom Hogan, other Web personalities, other workshop instructors, etc.). Eventually that will come back to haunt you.
We need better testing standards (e.g. length minimum sustained speed can be attained, heat propagation, etc.). We need better and clearer labeling. We’re starting to see sites pop up that are doing more extensive testing, and they’re anointing winners and losers. As a card maker you can try to make your reputation and profit by exaggerated claims—some cards only have maximum read speed on their label—but it won’t stick. Fewer cameras are being sold these days, and to more sophisticated users. You only have yourselves to blame if you annoy them by not providing useful information.
I continue to recommend only a few brands (and lines within a brand) that perform best on a Z9. At the moment, I can vouch for Angelbird Pro (SE/SX/XT), Delkin Black, and ProGrade Cobalt, as all have worked under sustained loads at byThom without hiccups. I personally won’t buy (or recommend) a card without published minimum sustained speeds (note that many hide that stat deep in their Web pages).
