Canon 5d mark III | Record Separately Vs. Record to Multiple | Performance Comparison

UPDATE: 5/3/2013 if you’re using mRAW or sRAW in your workflow, check out my recent article to see if you’re actually saving yourself hard drive space.

UPDATED 4/3/2012 to add metric for mRAW to CF, sRAW to SD.

I am in love with my new Canon 5d3.  A huge sigh of relief for me is the dual card slots, so I can always have a backup.  Unfortunately there’s a big performance difference between the SD card slot and the CF card slot.

If Canon had chosen to support the UHS-1 standard, we’d get roughly equal performance.  As it is, they supported SDXC, but not UHS-1.  (For the record, the camera will work with a UHS-1 card, it just won’t support the improved write speeds.  I’m using one in my camera and it works fine.)

Assuming you’re not using super old CF cards, the write speed of the SD card represents the bottleneck, and is what will limit how fast the camera flushes the data from the buffer.  Given this, I suspected that writing RAW to the CF card, and mRAW to the SD card might help speed things along.  The SD card data is only there for an “oh shit” moment, and under these conditions mRaw’s 10-ish megapixels ought to be good enough to get by.

There’s a gotcha, however.  When writing the same file to one or both cards, the buffer gives us a solid 12 frames before hitting the bottleneck.  When writing different files though, we lose roughly half the buffer.

For example, configuring the camera to record full RAW to the CF card, and either mRAW or sRAW to the SD card results in a 6 frame buffer.  Configuring the camera to record full RAW to the CF card, and JPEG (regardless of size/quality) to the SD card results in a 7 frame buffer.  It stands to reason that a chunk of the buffer memory is being utilized to generate two different versions of the file.

So the question is, would you rather have a bigger buffer and slower flush speed, or a smaller buffer and faster flush speed?  Only you can answer that question for yourself, but here are some speed comparisons to help you decide.

This trial is not intended to test performance of any particular memory card.  It is intended to compare different write settings within the camera on a given set of memory cards.

To test, I used one 64GB 400x Lexar Professional memory card, and one 64GB LX Series Patriot Memory SDXC card (class 10 / UHS-1).  For each test I held down the shutter button until the buffer filled, then waited for the shutter to click ten more times.  You can see the difference simply by looking at the recording.  The longest tracks took the longest to complete.

Canon 5d mark III record--separately vs record-to-multiple performance chart comparing various settings.
Click for full size. Recorded using the open-source audio recording software, Audacity.

In order of performance, times approximate, measured to highest peak on the last recorded shutter click:

  1.  6.9s, Full RAW to CF card only (super fast – no surprise)
  2. 9.8s, mRAW to CF card, sRAW to SD card.  (Interesting how it takes a long time to warm up, then runs pretty quick.)
  3. 13.8s, Full RAW to CF card, Large/Fine JPEG to SD card.
  4. 14.4s, Full RAW to CF card, mRAW to SD card.
  5. 18.1s, Full RAW to SD card only
  6. 18.6s, Full RAW to both CF and SD cards

Again, remember that options 2 through 4 in green, while completing the test faster, also limit you to a 6 or 7 frame buffer.

Your workflow will determine which option is best for you.

Which will you use and why?  Tell us in the comments below!


Author: Gavin Farrington Photography

San Francisco & Los Angeles wedding photographer specializing in capturing natural, candid moments.

21 thoughts on “Canon 5d mark III | Record Separately Vs. Record to Multiple | Performance Comparison”

  1. Hi,

    my 5DmkIII is on it’s way, so I don’t have hands on experience with the new camera.
    For my work a backup is essentially. Shooting RAW only, on CF and SD card simultaneously, will be my solution.
    Thanks for your comparison, it’s the first I found on the net which is so detailed.
    I did some math with the diagram you provided (assuming the RAW file size is 27MB):

    1. Write speed to CF card is ~67MB/sec (with the tested Lexar 400x).
    17 frames at full speed.
    2. Write speed to SD card is ~20MB/sec (with the tested Patriot SDXC Class10)
    13 frames at full speed.

    Do you have any other SD Card to test?
    Might be interesting to see if another fast card gives more frames at full speed or if the cameras SD interface is limited to 20MB/sec.




    1. Hi Tom. A UHS-I USB3 card reader would be great to test this with, but I don’t have one. I can guarantee you this camera is SD interface limited though. Don’t take my word for it, it’s right from Canon’s own lips documentation. Page 32 of the manual states, “Although the camera does not comply with the UHS (Ultra-High Speed) speed class standard, UHS SDHC/SDXC cards can be used. The manual in PDF form can be downloaded here:

      You can also find the spec referenced here:

      I realize this will be quite a disappointment for many users, myself included. I can only speculate that the decision was made to maintain differentiation between the 5d mk III and the 1d-x.


    2. The first and foremost would be a true relaese date! I’m in the market for an upgrade, want the mark ii but since it was relaesed two years ago I think I would be foolish not to hold out for the iii. A built in vertical grip like the 1d’s would be nice. NEVER EVER EVER AN ARTICULATING LCD!!! increased FPS would also be a great feature. 7fps anyone?


      1. Are you having trouble finding a mk III, Borjong? I was on the preorder list with B&H within minutes of their preorder page going live, so I was one of the lucky ones who missed the supply problems. I’m planning to get the battery grip next month. I have one on my mk II and I miss it a lot.


    1. This is not typically something that can be enabled with a firmware update, Jan. If it were, it would indicate that they’re putting the more expensive controller chips inside the camera, but electing to disable them. It’s not a common practice, but I can’t say stranger things have never happened.


  2. Hi
    Slightly off topic but given the speed issues, if I wanted to use a Eye Fi card would I be better using the Eye Fi card straight in its SD card form, or getting a CF/SD card adaptor?


    1. Hi Mike. I’m not sure I follow. Are you asking if the Eye Fi would perform better if it were not being written to at the same time as the CF card? There’s no reason to suspect that it would.

      I follow news about EyeFi products, but I’ve never actually bought one for myself to use. Aren’t they primarily designed for use with JPG files? So I’m assuming you would write RAW to the CF, and JPG to the SD?


      1. Hi Gavin
        Meant that given the slower write speed of the SD vs CF slots, if using Eye Fi for JPEG files would file writing and transfer to a network be slower using the Eye Fi card in the SD slot, and quicker by placing the Eye Fi card in a CF/SD card adaptor and using it in the CF slot?


      2. Hi Gavin, My experience of CF adapters for SD cards (though never used for Eye-fi) is that that they are extraordinarily slow no matter how fast the SD card going into it. Not fast enough to maintain a video write after buffer expires on 7D and the camera will not be aware of of the eye-fi card and may shut off power to it prematurely using energy saving settings if in CF.


      3. Hi Joe. Yeah, there are two different bottlenecks. One is the card itself, and the other is the interface to the card. The interfaces haven’t needed to be very fast in the past, because the cards have not been very fast. Also since primarily consumer cameras have used SD, there hasn’t been much pressure to really boost the performance. Class 10 sounds fast in the SD world, but it’s pathetically slow compared to even a mid-range CF card. Enter UHS-I, which finally creates a SD interface which is truly fast. It is a part of the SDXC spec, but manufacturers can chose to support SDXC without supporting UHS-I. This was Canon’s choice on the SD slot. I’m not much for conspiracy theories, but it’s a pretty safe bet that Canon was worried about stealing market share from the dual-CF-card-slot 1D-X.


  3. According to the 1DX Owner’s Manual page 120, despite supporting dual UDMA-7 CF cards it slows down dramatically if writing to both.

    E.g, 1DX max raw burst size to a single UDMA-7 CF card is 38 frames. If writing separately raw + large jpg to two CF cards, it drops to 17 frames — despite both cards being UDMA-7.

    To me this indicates the 5D3’s lack of support for UHS-1 SD cards is less an issue than some think.


    1. I’d love to try it out myself, Joe. I suspect that both the 1D-x and the 5d III need to work twice as hard to produce two different types of files, and thus use up twice as much memory in the process, thereby reducing the maximum capacity of the burst buffer. Regardless, the time to FLUSH the data will still remain dependent on the bandwidth to the cards, so having two high speed CF cards will still return you to a ready-state much much faster on the 1D-X.


  4. Wish I had read this a few days ago. Found out the hard way when I lost a day’s worth of video footage when the camera only wrote to the slower SD card and not to the CF at all. And no, I didn’t have it set wrong. It can’t sustain the ALL-I write speeds to the CF card when it has an SD card in the other slot.


    1. That’s disappointing, Alison. I don’t do video work of any kind, but I’ve heard the 5d3 will not write video to both cards, regardless of the configuration. Can you confirm? Is it true that this is only an option for stills?


    1. Hi Edgar,

      I always shoot full RAW to both cards. Having the camera halve the buffer for two different types of files is counterproductive. Also, should the worst happen and a card fails, having JPEGs to fall back on would not count as a “backup” in my opinion. Lastly, 64GB cards are just so damn cheap now.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.