The high resolution of today’s cameras, coupled with their ability to capture high FPS photos and up to 8K video, places huge demands on the speed of memory cards. If you’ve ever had to wait dozens of painful seconds for the buffer to empty, you’ll know what I’m talking about.

Of course, memory card manufacturers know of this problem, which is why they compete on read and write speeds. In an ideal world, those proudly proclaimed numbers, glowing in gold lettering on the box, would be the perfect buying guide. But we live in the real world, where the information written on the box does not always mean very much.
Are manufacturers trying to lie to us? I would like to think not. However, their test conditions are far from the real world. That’s why I decided to find out the read and write speeds for myself.
For now, I have tested the 12 memory cards that I happen to own. I will expand this real-life speed ranking every time I use a new memory card in the future.
How I Tested
I performed the tests below using a Nikon Z9 or Nikon Z8 coupled with a 50mm f/1.8 lens. Both cameras have the same maximum frame rate, the same buffer capacity, and the same file sizes. The advantage of the Z8 is that it also allows you to test SD cards. I set lossless compressed RAW (average file size 54.8MB), 1/1000s, f/2.8, and continuous shooting at the highest rate of 20 FPS. I shot the test scene for 60 seconds. This was repeated for five different test scenes to minimize the measurement error and the effect of the scene on the file size.
To measure read speed, I used my MSI Creator Z16 laptop (Intel Core i9, Geforce RTX 3060, USB3.2 Gen2 ports). For the connection between the card and the laptop, I used Exascend USB 3.2 Gen 2 Dual-Slot Readers (CFExpress B and SD cards) and Lexar Professional XQD 2.0 USB 3.0. For testing I used the EaseUS DiskMark application. Again, I took 5 measurements and then averaged the results.
I did not test the cards’ ability to resist overheating, nor did I measure their speed as a function of temperature. This measurement would certainly be very interesting, but would require strictly standardized conditions, which are not easy to achieve. The original Nikon MC-CF660G CFexpress Type B or the Delkin Devices 325GB BLACK CFexpress Type B card are reported to give very good results in this respect. So if you plan to use the card primarily for video recording, you should consider this feature as well in addition to speed.
The Results
In the first few seconds of high-FPS shooting, I did not notice any significant differences between the different cards. This is due to the relatively deep buffer of the Nikon Z9 and Z8. However, once the buffer is full, the differences become apparent, as the slowest cards immediately grind to a halt. On the worst card I tested, the frame rate after the first few seconds dropped precipitously, and the average speed over 60 seconds was a mere 1.9 FPS.
With that, let’s get into the tests.
#1 Nikon MC-CF660G CFexpress Type B
Declared write speed: 1500 MB/s
Declared read speed: 1700 MB/s
Average measured write speed: 854.9 MB/s
Average measured frame rate: 15.6 fps
Suitable for 8K video: Yes
Regular price: $726
#2 Delkin Devices 325GB BLACK CFexpress Type B
Declared write speed: 1530 MB/s
Declared read speed: 1725 MB/s
Average measured write speed: 839.9 MB/s
Average measured frame rate: 15.3 fps
Average measured read speed: 1067.0 MB/s
Suitable for 8K video: Yes
Regular price: $425
#3 Delkin Devices 128GB POWER CFexpress Type B
Declared write speed: 1700 MB/s
Declared read speed: 1780 MB/s
Average measured write speed: 832.4 MB/s
Average measured frame rate: 15.2 fps
Average measured read speed: 1055.6 MB/s
Suitable for 8K video: Yes
Regular price: $110
#4 Delkin Devices 256GB BLACK CFexpress Type B
Declared write speed: 1400 MB/s
Declared read speed: 1645 MB/s
Average measured write speed: 831.0 MB/s
Average measured frame rate: 15.2 fps
Average measured read speed: 821.3 MB/s
Suitable for 8K video: Yes
Regular price: $_ _
Where to buy: No Longer Available
#5 Exascend Essential CFexpress B 1TB
Declared write speed: 1700 MB/s
Declared read speed: 1800 MB/s
Average measured write speed: 805.6 MB/s
Average measured frame rate: 14.8 fps
Average measured read speed: Not measured (to be added)
Suitable for 8K video: Yes
Regular price: $469
#6 SanDisk 128GB Extreme PRO CFexpress Type B
Declared write speed: 1200 MB/s
Declared read speed: 1700 MB/s
Average measured write speed: 427.6 MB/s
Average measured frame rate: 7.8 fps
Average measured read speed: 1064.2 MB/s
Suitable for 8K video: No
Regular price: $110
Declared write speed: 400 MB/s
Declared read speed: 440 MB/s
Average measured write speed: 355.1 MB/s
Average measured frame rate: 6.5 fps
Average measured read speed: 355.5 MB/s
Suitable for 8K video: No
Regular price: $90
Declared write speed: 400 MB/s
Declared read speed: 440 MB/s
Average measured write speed: 352.7 MB/s
Average measured frame rate: 6.4 fps
Average measured read speed: 402 MB/s
Suitable for 8K video: No
Regular price: $160
Where to buy: Amazon.
#9 Lexar XQD 32GB 2933x Professional
Declared write speed: 400 MB/s
Declared read speed: 440 MB/s
Average measured write speed: 351.1 MB/s
Average measured frame rate: 6.4 fps
Average measured read speed: 361.3 MB/s
Suitable for 8K video: No
Regular price: $_ _
Where to buy: No Longer Available
#10 Sony 64GB SF-G TOUGH Series UHS-II SDXC
Declared write speed: 299 MB/s
Declared read speed: 300 MB/s
Average measured write speed: 236.4 MB/s
Average measured frame rate: 4.3 fps
Average measured read speed: 94.3 MB/s
Suitable for 8K video: No
Regular price: $100
#11 SanDisk 64GB Extreme PRO UHS-II SDXC
Declared write speed: 260 MB/s
Declared read speed: 300 MB/s
Average measured write speed: 222.5 MB/s
Average measured frame rate: 4.1 fps
Average measured read speed: 94.5 MB/s
Suitable for 8K video: No
Regular price: $65
#12 SanDisk 64GB Extreme PRO UHS-I SDXC
Declared write speed: 90 MB/s
Declared read speed: 200 MB/s
Average measured write speed: 106.5 MB/s
Average measured frame rate: 1.9 fps
Average measured read speed: 95.6 MB/s
Suitable for 8K video: No
Regular price: $13.49
Analysis
What stood out to me was the dramatic difference between CFExpress Type B, XQD, and SD cards. All of the CFExpress B cards, with the disappointing exception of the SanDisk 128GB Extreme PRO, achieved roughly 15 FPS averaged over 60 seconds of high-speed shooting. The XQD cards hovered around 6.4 FPS. UHS-II SD cards were about 4.2 FPS, while the UHS-I card I tested achieved a mere 1.9 FPS.
A positive surprise was the Delkin Devices 128GB POWER CFexpress Type B card. It achieved a frame rate of 15.2 fps, just 0.4 fps behind the test winner, while costing much less ($110 rather than $726, granted, with a smaller memory capacity). However, Filip Bartak, a product specialist at Nikon Czech Republic, pointed out that the 256GB version of this card does not achieve these speeds. In an update to this article, I will support – or deny – this claim with my own data.
The disappointment of this test was the SanDisk 128GB Extreme PRO CFexpress Type B card. The average frame rate was only 7.8 fps, which is about half of what the aforementioned Delkin 128GB can do for exactly the same money.
Taking price into account, I was also impressed by the Exascend Essential CFexpress B 1TB card. This card has a huge capacity and a frame rate of 14.8 FPS, which is almost at the level of the winners of the test. All this for a price of $469, which is the second best price/capacity ratio after the SanDisk 64GB Extreme PRO UHS-I SDXC card.
Let me know what cards you would like to see me test next!
Libor,
Thank you so much for publishing these test results and for taking the time. I’m curious on the testing method. Are the results inclusive of the buffer in both the Z9 and Z8? Or, were you able to somehow separate out just the shots written AFTER both bodies filled their buffers? I.e., are the results indicative of just the write speeds of the cards, or are they the average over the full 60 seconds, thus including all shots first filling the buffers? If it’s the latter, that might be falsely showing a closer gap between the different cards.
I also have both of these bodies and I was trying to think of how I could replicate this test to only report results AFTER the buffers are full and just measuring the write performance of the cards…but I’m failing to figure out how to separate out the initial shots required to fill the buffer. If you’ve found a way to do that, I’d be interested in know how you did that.
Kind regards,
Tony
Hi Tony, basically it’s probably not realistic to separate the buffer from the write to the card. It’s too dynamic a system. That’s also why I decided on a capture length of 60 seconds, so that the buffer size could be neglected. At the same time, even though the buffer introduces a system error into the test, this error and its size is constant for all cards, so essentially ignorable. If you wanted to replicate the test, take five consecutive 60-second bursts (on different test scenes). Set 14-bit RAW and 20fps. Calculate the average number of frames per second. Calculate the average write speed (total data/5*60sec). I will soon extend the test with other very interesting cards. I will be glad if you share your measurements in the comments. Thank you in advance.
Best,
Libor
Speed is certainly important, next is reliability of the card. Any idea wat the failure rates are of these? I’ve had Lexar fail but not my SanDisk. Though the SanDisk CF Express didn’t fare well in this test. It has been suitable for my sports shooting.
Reliability is indeed an essential property. Unfortunately, it is probably untestable. That’s why I would be very grateful for any stories about your card failing. But I think these failures are very rare. In the x years I’ve been shooting digital, I’ve only lost data once due to a defective card. Touch wood…
Good concept but I shoot JPG instead of RAW for sports and it would be neat to see how many images could be taken at 20FPS before the buffer filled.
Erase your card, set it 20 FPS and hold down the shutter release till you hear the buffer chocking up.. (most likely the card will be filled up JPEGS completely) otherwise see it here: www.youtube.com/watch…038;t=939s
Thanks for your test.
I would be very interested to see you test the Angelbird CF Express cards (several versions) from the austrian maker. The company reports the sustained read and write speeds of their cards.
Thank you Patrick, I will definitely be expanding the portfolio of cards I test. The Angelbird and ProGrade cards will definitely be among the ones I want to focus on.
It’s great to see your experiences and observations about these cards. In a way the speeds “declared” are theoretical. They are based on the fastest card readers they could find. And they are based on very short I/O reads or writes. In the way your camera has a frame buffer, these cards have a very fast I/O cache. In prolonged writes the cache gets saturated and the I/O (write) speed drops to the actual “media speed”.
The price per GB we pay depends on (a) the quality of the media behind the cache and (b) speeds this can reach. The quality lies in how many rewrites (“program’/erase cycles”, “P/E cycles”, in the jargon of an electrical engineer) the memory cells in the media section can sustain. Every program/erase is a slow “brown out” and one type card may fail after 300 times, another, like my Delkin (I asked them and they answered) after 30,000.
The media speed is often referenced in specifications as the maximum sustained write speed. (mutatis mutandis for reads).
One way to sell higher quality (more P/E cycles or more “TBW”) is to add more media in the card than is on the label. This is called “over-provisioning” in the IT industry. That “over” amount is available to your card (or SSD) ‘s controller/firmware that runs a process called “wear leveling”. So if you buy a 128GB card, there may be 128GB+64GB under the hood. What we have seen in the last few years is that my example 128 card is relabeled and now has 128+32=160GB media available to you. But the overprovisioning got reduced from 64 to 32 and now you pay more for one reason or another.
In this review, the 325GB and 660GB card must be suspected to have come to these media amounts on the label.
As to offloading media from my CFexpress Type B cards, I use a double slot Sonnet card reader for that on Thunderbolt (40 Gbps peak) – where the cards can run at maximum speed – it doesn’t get faster than that.
Thank you for the addition. It would certainly be interesting to know how many overwrites each card can withstand. And also how it handles high temperatures. But these are parameters that are quite difficult to test. Fortunately, in all the time I’ve been shooting with digital cameras, I’ve only had two problems with cards. The first time was with a SanDisk Compact Flash 32GB card, which gave me an error message when it reached a certain fill level. I did not lose any data at that time. The second case happened last year when my Delkin 256GB POWER CFexpress Type B card failed. I lost all the photos on the card from a morning shoot. Fortunately, I was able to return to the location and mitigate the loss. Since then, my faith in the immortality of memory cards has been shaken, which is actually a good thing. I’d love to hear about your card failures.
Delkin states on their website that the Type B G4 cards have two different sustained write speeds based on capacity:
128-512GB – 805MB/s
650-2TB – 1490MB/s
My 1TB card gets 1240MB/s write and 1496MB/s reads with Blackmagic speed test.
Interesting, one would expect cards with less capacity to be faster.
Sabrent CFX Rocket Pro 512GB has been very good with my Z8. I wonder what your tests would show. Bought at amazon’s swedish site for 3280 swedish crowns (275 euro as of today).
I probably won’t be able to get my hands on the Sabrent CFX Rocket Pro cards at the moment, but if you run the test as I’ve described it, you should get the same results as if I had run it. I’d love to hear from you if you let me know how the test turned out.
The faster CFX results are so close – have you inadvertently measured the camera’s maximum write speed? Another thought occurs, of course, which is that they differ only on the outside and are rolling off the same production line…
In the limit, it’s measuring the minimum of the sustained write speed of both the camera and the card. The initial buffer will fill within the first 5-10 seconds, and everything after that is the camera + card sustained write combo.
It is quite possible that they only differ on the outside. Nikon is not a flash manufacturer, nor are they a flash controller manufacturer, so almost certainly they’re white-labeling somebody else’s card, perhaps with some customization and requirements on minimum write speed (not listed), heat tolerance (for sustained 8K video), and longevity.
The #2 Delkin Black advertises a minimum write speed of slightly more than the Nikon’s advertised peak write speed, but given the performance it seems likely that Nikon is low-balling their performance. Given the price differential between the two cards, at least there’s *some* difference between them.
Given the discrepancy between the minimum write speed e.g. Delkin Black lists and the achieved write speed, I could believe that it is limited by the camera’s bandwidth, and that’s a bit above 800 MB/s.
It is very likely (certain) that the Nikon cards are rebranded cards from another manufacturer. In any case, Nikon has made sure that these cards meet certain criteria and guarantees them. According to the Nikon specialist I spoke to on the subject, Nikon cards are very resistant to overheating. There are a number of parameters that are not written in gold letters on the box, but they play a very important role. At the same time, they are difficult to test seriously.
Great information on the testing and yes the Delkin Power CFX version is surprising! I was surprised that you didn’t have any Prograde cards in the test (Gold or Cobalt). Interesting to see how they score. I use the Prograde Gold and Delkin Black with no issues in Airshow or BIF photography. Minimal video at this point, but sounds like I am set with the two TOP brands.
Thank you Mike, I’m going to have the ProGrade Cards available soon, so I’m going to expand this review by including them.
Grief!
I could buy a very good used camera for the price of these things!
Just shows why one should be very wary of upgrading.
It’s not just the camera – or even the lenses – it’s the laptop and the cards as well.
The only thing that doesn’t need an upgrade is the monitor.
Isn’t it … ?
I’m okay for now, Robert: my array of thirty two HD monitors on the wall gives me a 63 megapixel display with which to examine everyone’s images at the pixel level 🤣
Even the monitor is not immortal, Robert. The cave paintings in Altamira Cave will fade millions of years after tomorrow’s Eizo :-)