5D4 performance with various SD cards (some test data)

Since getting the 5D4 I've been using the old 45MB/s SD card I had in my 5D3 (along with a fast CF card). The 5D3 has an "issue" that prevents it from writing to SD cards quickly, so at the time there was no point in buying a faster SD card. Having just obtained a 95MB/s SD card I decided to do a bit of testing of buffers and write rates. There's nothing new/groundbreaking here, but it might be of interest to some.

The cards I used were:

• 64GB Lexar Professional 1066x 160MB/s CF
  
• 32GB SanDisk Extreme 45MB/s SD
  
• 32GB SanDisk Ultra 80MB/s Micro SD (in an SD adaptor)
  
• 64GB SanDisk Extreme Pro 95MB/s SD
  

Note: The above figures are claimed read rates. The claimed write rate of the 95MB/s SD is 90MB/s. I don't have data to hand for the others.

The cards were formatted in-camera before each test, and the camera was running in high speed shooting with the OEM grip and two freshly charged batteries.

A non-IS lens was used, in manual focus, with 1/1000 shutter speed, wide open, ISO 100, and the lens cap fitted. The camera was set to record RAW and JPEG to both cards (if present).

The above settings resulted in consistent 26.1MB RAW files, and 590KB JPEGs (so 26.68MB of data per frame).

In general I found that writing all files to CF and SD simultaneously gave about the same performance as just writing to the SD. This is expected, as when writing the same files to both cards, the bottleneck will be the slowest device (and the CF card used here is faster than all the SD cards).

Using just the CF card resulted in 21 frames (@ 7fps) before the buffer was full. The camera then shoots at the rate it can write frames to the card. In this case, approx 4fps (so somewhere just over 100MB/s write rate). Shooting 22 frames (the last resulting in a tiny pause as it's just beyond "buffer full") clears to the card in about 6 seconds (a calculated rate around 98MB/s).

The 45MB/s SD card resulted in 16 frames before filling the buffer, and took 11 seconds to clear.

The 80MB/s Micro SD only achieved 14 frames, and took a huge 34 seconds to clear (more on this later).

The 95MB/s SD card gets 19 frames, and clears in 8 seconds (a resulting rate of around 64MB/s). Frames beyond "buffer full" were around 2.8fps; that's 2.8 * 26.68 ~= 75MB/s (probably close enough to 64MB/s, given the accuracy of the test).

Testing the SD cards in a USB3 card reader on a PC (using CrystalDiskMark) showed the 45MB/s SD card has read/write rates of approx 47MB/s and 43MB/s. The 95MB/s SD card was 93MB/s and 86MB/s. The 80MB/s Micro SD was 93MB/s and 30MB/s (either in the SD adaptor, or directly in a Micro SD slot on the card reader). This would explain the poor buffer clearing times of the Micro SD card.

I also did some calculations of how much data would likely be in the camera's buffer during each test - essentially the level of water in the bucket, as it fills quicker from the top (new frames) than it's draining from the bottom (writing to the card(s)). I'd estimate the buffer on the 5D4 to be around 300MB (albeit probably with a +/- 50MB level of accuracy).

In conclusion then:

• If you want the maximum number of frames, with the quickest clearing time to the card; shoot only with a high speed CF card
  
• An apparently fast Micro SD card can have a very poor write rate
  
• Recording to both cards doesn't appear to have much penalty (speeds are broadly what you'd get from just using the SD card)
  
• It is worth using a fast UHS-I SD card in the 5D4
  

I didn't test writing RAW to the CF and JPEG to the SD. If everything works as expected then that should result in a minor improvement over writing both file types to a CF card (assuming a fast CF and fast SD). However, I write all to both cards as a backup, and moving from the 45MB/s card to the 95MB/s card has given me an extra 3 frames at 7 fps, and much faster buffer clearing - for those times when I do need a fast shooting rate.

Real scenes (i.e. not the back of a lens cap) will produce bigger files; and especially high ISO shots, as the noise is harder to compress. I'd assume you could broadly scale the figures above; i.e. 21 frames around 26MB is about the same amount of data as 14 frames at 40MB. So, if your typical shooting conditions produce ~40MB RAW files, then the above CF card should get you about 14 frames at 7 fps on the 5D4. Continuing to shoot with a full buffer would get you roughly 100MB/s / 40MB = 2.5fps.

Well, that's 5 minutes of your life you're not getting back eh?