E-M1 II Shooting Rates with SD cards

Discussion in 'Olympus Cameras' started by bassman, Jan 3, 2017.

  1. bassman

    bassman Mu-43 Top Veteran

    800
    Apr 22, 2013
    New Jersey
    The Bassman
    After a recent conversation on another forum about usage of the two card slots in the E-M1ii, I decided to try a test using the cards in hand.

    The setup I have been using since I got the camera is:
    Slot 1: Sandisk Extreme PRO 300MB/s UHS-II 64G for raw files
    Slot 2: Sandisk Extreme PRO 96MB/s UHS-I 32G for LF JPEGs

    The concept is to have the larger raw files on the faster card in Slot 1, and JPEGs in the slower card for both backup in the event the first card fails, and to be able to use the JPEGs while traveling for editing and sharing on my iPad.

    I ran five tests:

    Test 1: Slot 1 UHS-II Raw and Slot 2 UHS-I LF JPEG
    Test 2: Slot 1 UHS-II Raw and Slot 2 empty
    Test 3: Slot 1 UHS-I Raw and Slot 2 empty
    Test 4: Slot 1 UHS-II Raw and Slot 2 UHS-I Raw
    Test 5: Slot 1 UHS-II Raw+LF and Slot 2 UHS-I Raw

    For all tests:
    - I formatted the cards in camera before each test
    - The camera was set to High Speed @ 15 FPS
    - The camera was set for manual exposure and focusing: 1/50s, f/2.8
    - The camera was on a tripod
    - I held the shutter for 30 seconds, and timed how long the camera took to finish writing to the cards with a stopwatch.
    - Timings for the frame rate were taken from the timestamps on the images to the nearest second; there may be some rounding, but it appears to be not significant.

    Here are some of the findings.
    - The camera never exceeded 12 FPS. There are a few seconds in all of the samples where either 11 or 13 images were recorded, but this is just rounding from the timestamps.
    - The best performance was in Test 2, using only the fast card: 155 total images in 30 seconds (5.2/sec) and 45 seconds until the camera finished writing (3.4/sec). About 60 shots in 5 sec before the buffer filled and the camera slowed down, running at 3.7/sec for the remainder of the 30 seconds.
    - Tests 3 and 4 were very close: 110/109 total shots in 30 seconds, 54/55 seconds to complete writing, about 50 shots in 5 sec until the camera slowed down to about 2.2/sec. This implies that the camera was limited by the slower card in either slot. I'd like to test two fast cards, but I only have one so far. It also implies that the camera writes simultaneously to both slots.
    - Tests 1 and 5 had the worst performance. Test 1 got less than 50 shots in less than 5 seconds before slowing down to 1.3/sec, completing 85 shots in the 30 seconds and taking 70 seconds to finish writing. Test 5 got about 50 shots before slowing down to 0.9/sec, completing 76 shots and taking 81 seconds to finish writing.

    Here are a few conclusions of mine.
    1. There's a bit of marketing hyperbole going on with the 15 FPS second. The camera never achieved that with any configuration of cards, hitting 12 FPS in each case.
    2. If you need pure speed, keep the second slot empty and use a fast card in Slot 1 (subject to my testing with a second fast card).
    3. Writing raw to both cards is faster than writing raw to the fast card and LF to the second. Since the write speeds to the two slots are identical, the timing difference must be the JPEG generation process - either it's quite slow, or it's done serially and not simultaneously with writes to Slot 1.
    4. Going to full flexibility - two copies of the raw file on two cards, plus a LF JPEG for SOOC sharing - really slows you down. In 30 seconds of shooting, you get 76 images rather than 155, and it takes 81 seconds compared to 45 to complete writing. This is not surprising, of course.
    5. You get the maximum speed of 12/sec for at least 4 seconds in each test, achieving it for 5 seconds only with the single fast card in Slot 1.

    I'd be happy to have others make similar tests, perhaps with different cards. I'm going to order a second fast card, and I'll repeat the tests after that.
     
    • Informative Informative x 3
    • Useful Useful x 1
  2. Growltiger

    Growltiger Mu-43 Top Veteran

    854
    Mar 26, 2014
    UK
    I would be very interested in the results from these two configurations:
    1. Slot 1 UHS II Raw+LF JPG. Slot 2 enpty.
    2. Slot 2 UHS II LF JPG. Slot 2 empty.
    I predict that config 2 will deliver 15 fps.
     
  3. bassman

    bassman Mu-43 Top Veteran

    800
    Apr 22, 2013
    New Jersey
    The Bassman
    I just happened to receive a second UHS-II card, slightly slower at 280 MB/s but otherwise the same as my first fast card. A quick run of Test 5 with the new card in slot 2 showed no significant improvement. That config appears to be limited by the need to write two files to slot 1 - obviously one at a time - and create the LF JPEG in addition to the writes.

    I'm not home now but I'll repeat the above tests plus the two additional ones later.

    Let me add to the shooting conditions, as they might conceivably have an impact (although I don't think so):

    - S-IS Auto
    - using LV rather than the EVF
     
  4. barry13

    barry13 Super Moderator; Photon Wrangler Subscribing Member

    Mar 7, 2014
    Southern California
    Barry
    Hi, which lens did you use? Some lenses have slower aperture mechanisms and that can limit frame rate (even wide open in some cases).
    Thanks
     
    • Agree Agree x 1
  5. AussiePhil

    AussiePhil Mu-43 All-Pro

    Jun 1, 2014
    Canberra, ACT, Aust
    Phil
    Couple of things I'd like to see/know
    Which lens?
    And can the shutter speed be increased to at least 1/200.

    Thanks for this, largely mirrors my impressions except that I'm sure that slot 2 doesn't need to be empty.... I'll see if I can replicate this on the weekend
     
  6. bassman

    bassman Mu-43 Top Veteran

    800
    Apr 22, 2013
    New Jersey
    The Bassman
    Should have specified the lens: Panasonic 12-35/2.8. If the camera can't drive a hi-end lens like that at the quoted speed when shooting Raw on a single fast card, I claim bogus advertising. Although I'm sure Oly specifies one of their own lenses in the claim. I'll try the Olympus 45/1.8 later.

    For the first 5 tests I set the SS to give a proper exposure in the room (its dark and rainy today). 1/50s would give up to 20 fps ignoring any intershot time. ISO was 3200. For the test with two fast cards, SS was about 1/250, as I was outside.
     
    Last edited: Jan 3, 2017
  7. Olympus micro 4/3 lenses are required for Pro Capture, so it's possible they may work better.
     
  8. bassman

    bassman Mu-43 Top Veteran

    800
    Apr 22, 2013
    New Jersey
    The Bassman
    BTW, I'd be happy if someone achieved better results and posted the info on what I'm doing wrong. There's no ego here (well, not much) and I'm just trying to help everyone get the most out of this (expensive) machine.
     
    • Like Like x 1
  9. Ramsey

    Ramsey Mu-43 Top Veteran

    738
    Jan 9, 2013
    Zagreb, Croatia
    I think everyone is grateful for the tests and i sincerely doubt there was any criticism aimed at you. Make no mistake, YOU're the man.

    I have to admit i read the end to your initial post wrong, sth to the extent "if anyone wants, i'd be happy to make similar tests". Skipping certain stuff while reading may or may not have caused that.

    Lens used, AF settings and SS are the first things people are curious about in these kind of tests. Again, thx for the test.
     
    • Like Like x 1
  10. bassman

    bassman Mu-43 Top Veteran

    800
    Apr 22, 2013
    New Jersey
    The Bassman
    Screen Shot 2017-01-03 at 5.32.34 PM.

    Ok, re-ran Test 1 as Test 6. Used the 300MB/s UHS-II card in slot 1, the 280MB/s UHS-II card in slot 2. Slot 1 raw, Slot 2 LF. To avoid any cross-manufacturer issues, I used the Olympus 17/1.8 this time, and shot at 1/250s, f/2.5, ISO 3200. Results were (drum roll ...) the beast got up to 15 FPS for less than 2 seconds - the first and second full seconds in the timestamps. Shots until is slowed down was about 55, ahead of Test 1 with the slow card, which got almost 50. Total shots were the same as Test 1 in 30 seconds at 84, and time to flush the buffer was also about the same at 72 seconds.

    The left image is the SOOC JPEG, and the right is the raw image with the Huelight Standard profile.

    I finally ran a new test, #7, with LF JPEG only assigned to the 300MB/s card in slot 1, and slot 2 empty. It ran faster the Test 3 with raw only, clocking about 15 FPS up to almost 60 images at about 4 seconds, then trucking along at about 4 FPS for a total of 154 images in 30 seconds. And then it spent about 15 seconds flushing the buffer, achieving 3.4 FPS over the 30 seconds.

    I reran the raw/raw test as #8 with the two fast cards. This also achieved a peak FPS of 15, but slowed to about 2.1 FPS after the buffer filled. 112 total images in 30 seconds.

    This this mystified me, since the dual raw test achieved a higher peak FPS than the single raw Test 3, so I reran the test. This time I achieved approximately the same results as the LF only test: peak of 15 FPS, about 4 FPS once the buffer filled, and 152 total images.

    There you have it.


    View attachment 500554
     
    Last edited: Jan 3, 2017
    • Informative Informative x 4
  11. retiredfromlife

    retiredfromlife Mu-43 Top Veteran

    716
    May 15, 2016
    Syndey, Australia
    Thanks for the very comprehensive testing, results still not too bad though.
    I thought I read somewhere that the fastest speeds were from the 12-40 pro only.

    Good to see some real world testing, especially with other brand lenses as that is what will happen in the real word use.
     
  12. bassman

    bassman Mu-43 Top Veteran

    800
    Apr 22, 2013
    New Jersey
    The Bassman
    So the answer - at least with the mechanical shutter - is that you can indeed achieve 15 FPS, for for no more than 4 seconds. Then the buffer fills, and the camera slows to either 2 or 4 FPS and seems to be able to run at that rate indefinitely.

    I also double checked whether shooting at 1/50s or 1/250s made any difference - it didn't.

    Finally, this testing all consumed about 60% of the battery. Just the tests I reported were about 1,000 shots; I easily shot 3x that number because of all my mistakes and reruns. This just reconfirms what we already knew - the battery is more limited by On time and especially LCD/EVF on time, and not by actual shots.
     
    • Informative Informative x 1
    • Appreciate Appreciate x 1
    • Useful Useful x 1
  13. Cederic

    Cederic Mu-43 Veteran

    258
    Nov 14, 2012
    Nottingham
    So you've just told me that the camera in transit to me can only do 30 full sensor photographs in two seconds using its mechanical shutter?

    That's terrible, that's only 12 more than any camera I've ever previously owned! Maybe I should get a refund :(

    Interesting results though, and thank you for sharing. I've invested in two cheap UHS-II cards (£55 each for 128GB) that only do 150MB/s so I'll run a test myself and see how it performs.
     
  14. Growltiger

    Growltiger Mu-43 Top Veteran

    854
    Mar 26, 2014
    UK
    Are you saying that if you simply fire off JPGs and have a fast UHS II card it won't take more than 30 photos at 15fps before slowing up? I find that hard to believe.
     
    • Agree Agree x 1
  15. bassman

    bassman Mu-43 Top Veteran

    800
    Apr 22, 2013
    New Jersey
    The Bassman
    So did I. It's more like 55 shots. The high speed burst rate is (I believe) constrained by the buffer "capacity", which surprisingly is about the same regardless of shooting raw or LF. Total write time in each case was about 45 seconds (I'm timing that with a stop watch, so it could be a second or two off in either direction).

    Here's the frames achieved in each of the first 8 seconds when shooting LF onto a single fast card in slot 1. The first second in the LF test appears slow because I'm only looking at whole-second resolution from the image data, and I obviously couldn't time the start of pressing the shutter button to 1/15th of a second. In the Raw test, I started right at the beginning of the first second and the camera slowed before the end of the fourth second. Both tests are with the same card, same lens, same everything (afaict) aside from LF vs. Raw. I held the shutter down for 30 seconds. The data is collected by looking at the timestamp and frame# recorded by the camera for each image.

    LF LF Raw Raw
    Second Frame# # frames Frame# # frames
    1 1-11 11 1-15 15
    2 12-27 16 16-30 15
    3 28-42 15 31-45 15
    4 43-57 15 46-56 11
    5 58-62 5 57-60 4
    6 63-66 4 61-64 4
    7 67-69 3 65-67 3
    8 70-73 4 68-71 4
    9-30 74-153 80 72-152 80

    As I mentioned above, if someone could explain either a flaw in my methodology or my camera setup which explains these results, I'd be very happy.
     
  16. Growltiger

    Growltiger Mu-43 Top Veteran

    854
    Mar 26, 2014
    UK
    I can see nothing wrong with what you have done.
    But the buffer size not getting larger when shooting just JPG doesn't make any sense.
     
  17. Speedliner

    Speedliner Mu-43 Hall of Famer Subscribing Member

    Mar 2, 2015
    Southern NJ, USA
    Rob
    so what does this mean for 60fps RAW?
     
  18. hoodlum

    hoodlum Mu-43 Veteran

    255
    Jul 16, 2012
    Toronto Canada
    I just did Test#1 and #2 with the Lexar 2000x UHS-II in slot 1 and a Sandisk extreme Pro UHS-I in slot 2. I got the exact same results for number of images shost within 30 seconds and the time to clear.

    IR was able to get much better results. I wonder what they are doing different. I am no where near 9.5fps when the buffer is full.

    Olympus E-M1 II Review: Now Shooting! - Performance

    Continuous H mode
    RAW

    0.07 second
    (15.4 fps);
    102 frames total;
    7 seconds to clear*
    Time per shot, averaged over buffer length of 102 frames, then slows to an average of 9.5 fps when buffer is full, which is still fast!
     
  19. barry13

    barry13 Super Moderator; Photon Wrangler Subscribing Member

    Mar 7, 2014
    Southern California
    Barry
    Hi, I believe the buffer holds raw data from the sensor, which is converted to ORF and/or JPEG during the write pipeline.
     
  20. hoodlum

    hoodlum Mu-43 Veteran

    255
    Jul 16, 2012
    Toronto Canada
    OK. I found the problem. I needed to increase my shutter speed but more importantly lower the ISO. It seems like high ISO has a large impact on how the images are processed even when shooting only RAW. I shot straight at a light indoors so that I could shoot at base ISO and at 1/1000s with my 300mm f4 while still having a normally lit image like outdoors. With Test#2 I shots approximately 320 images in 30 seconds (tried it twice with same results). It took under 7 seconds to clear the buffer. This matches up with the results from IR.

    Edit: I also quickly tried test#1 and it shot 110 images in 30 seconds and took just over 20 seconds to clear the buffer. Based on this results I have decided not to write JPEGs to slot 2 as I am using the M1ii mainly in burst mode for birds and the JPEG limits the burst capabilities. I will leave the 2nd slot as a fail over in case the first card becomes full or enable JPEG adhoc if required to send through WiFi.

    I hope this helps.
     
    Last edited: Jan 4, 2017
    • Informative Informative x 6
    • Like Like x 1
    • Appreciate Appreciate x 1