I’ve been looking for a card reader for my CFexpress Type B cards that I use in my camera. There are not very many devices on the market, but they have a large price span: around 10x.

Since I was wondering whether the price correlates with speed, I decided to perform some simple tests using my existing hardware:

• a UGREEN CFexpress Type B card reader
• a PGYTECH CreateMate CFE-B/SD card reader
• a Nextorage B1 Pro CFexpress Type B card with 165 GB capacity
• the built-in NVMe SSD of my Thinkpad laptop, a Samsung SSD 990 PRO with 2 TB storage capacity.

The two external card readers are attached via USB-C/USB 3.2 Gen 2 to my Thinkpad Thunderbolt 3 dock. I am generally happy with both of them, I think they have good quality.

By the way, why should I even look for a new card reader in the first place? Well, the CreateMate card reader is a bit bulky and heavy and I once forgot to take it with me on vacation. So I was looking for something smaller and lighter.

Sequential read and write tests

I did not want to overcomplicate things, just get an impression if there are any differences between card readers or not. Therefore I decided to perform sequential read and write tests using dd as described at Baeldung.com.

These are the commands that I used in Fish shell; Bash has slightly different syntax.

#!/usr/bin/fish
echo "*** Sequential write test ***"
for i in (seq 1 5); dd if=/dev/zero of=/run/media/daniel/NIKON\ Z6_3/tempfile bs=1M count=1024 conv=fdatasync; end
echo
echo "*** Sequential read test ***"
for in in (seq 1 5); sudo sh -c "/usr/bin/echo 3 > /proc/sys/vm/drop_caches" && dd if=/run/media/daniel/NIKON\ Z6_3/tempfile of=/dev/null bs=1M count=1024; end
rm /run/media/daniel/NIKON\ Z6_3/tempfile

Results

The graph below shows the means and standard deviations of 5 tests for each device. The “direct” label denotes tests with the card reader attached directly to a USB-C/USB 3.2 Gen 2 port of my Thinkpad P14s laptop, rather than via the Thunderbolt 3 dock.

When attached to the Thunderbolt dock, the PGYTECH CreateMate reads from the CFexpress card much faster than the UGREEN reader. It does also write a bit faster than the UGREEN reader.

Attaching the card readers directly to the laptop results in increased speeds overall, and the differences between the UGREEN and the PGYTECH readers seems negligible.

However, regardless of how the readers are connected to the laptop, both are left in the dust by the internal SSD!

Is the observed difference in read speeds between the UGREEN and the PGYTECH card readers “clinically relevant”? Does it justify spending a lot more money for the PGYTECH reader? Given the very large difference of both of these card readers to the internal SSD, I don’t really think so.

What to expect from a “USB-C” or rather, USB 3.2 Gen 2 connection

It should be noted that the USB 3.2 Gen 2 standard has a nominal transfer speed of 10 GBit/s or 1.25 GB/s. As stated in the Wikipedia article on USB 3.2, actual transfer rates are round 0.8-1.0 GB/s. Still higher than what I measured.

My Thinkpad’s USB-C ports a really Thunderbolt 4/USB 4 ports that should be capable of transferring up to 40 Gbit/s or 5 GB/s. This would be in the internal SSD’s ballpark. However, the card readers have “only” USB 3.2 Gen 2. PGYTECH claims 1000 MB/s for the CreateMate and UGREEN 1250 MB/s for their device. The card itself claims read speeds of up to 1950 MB/s and write speeds of up to 1900 MB/s, which roughly corresponds to 1.86 and 1.90 GB/s. Obviously, the 0.3 to 0.6 GB/s that I measured are nowhere near any of these numbers.

Using Darktable’s local copies feature

Evidently, nothing is faster than the internal NVMe SSD.

This reminds me that Darktable has a feature to copy image files to the built-in SSD. This not only allows for faster read and write speeds, but also makes it possible to work on the images without having the card reader attached to the laptop: Darktable’s user manual on local copies.

Local copies (or rather, the XMP sidecar files) are synchronized whenever the external storage is attached when Darktable is starting up. I don’t know if Darktable will automatically remove synchronized local copies in case disk space gets low.

Conclusion

In practice, I am going to create local copies of my images and select and work on the local copies. Yes, the initial process of copying the files will take a few seconds less with the PGYTECH reader than with the UGREEN reader, but that is negligible given the amount of time that I spend on selecting the images and working on the RAW file development in total.

Therefore, if I decide to purchase a new reader, I will focus on the size and build, and the UGREEN reader is a strong contender when it comes to small size.

The above was edited as usual in Darktable. I wanted to let the lights in the windows glow a bit and have them contrast with the otherwise rather dark scene. I did not have a mist filter on my camera, so I attempted to simulate one with Darktable’s Diffuse or Sharpen module. In my opinion, the effect of mist filters can quickly destroy an image, it’s like emptying an entire bottle of ketchup all at once onto your fries. I was wary of that and attempted to make it subtle. Don’t know if this was the way to go.

See my Diffuse or Sharpen mist filter preset over at github.com/bovender/darktable-presets-and-styles.

The screenshot above shows the mask overlay in Darktable. The screenshow below shows the settings of the ‘g’ channel: I want to exclude dark image areas from the filter effect.

The next screenshot shows the ‘hz’ channel, which I adjusted to include warmish colors only.

Finally, I applied a large blurring radius so that I can achieve a glowing effect.

I’m not sure yet whether to exclude the moon from the mist filter effect or not.

Original image developed in Darktable. Raw image file available upon reasonable request. All rights reserved.

Here’s one of several video tutorials on the Diffuse or Sharpen module by Boris Hajdukovic. He’s a master of image editing with Darktable. Highly recommended.

Images developed in Darktable. Raw image file available upon reasonable request. All rights reserved.

Here are the first two of them, both taken on “Schwanheimer Wiese” a meadow in the forests of Frankfurt-Schwanheim.

The “Struwwelpeter” is a tree named after a figure from a famous children’s book:

Camera: Pentax K-3 APS-C, 18-55 mm kit lens (crop factor 1.5), image developed in Darktable. Raw image file available upon reasonable request. All rights reserved.