Tuesday, June 20, 2017

Leica M10 vs Fujifilm X-T2 vs GFX 50S for High ISO Performance


The Leica M10 is very special. It's also very expensive. Of all the cameras I review, I get the most criticism from both fans and critics. The one thing we can't argue over is the final image quality. It is what it is. Although I rarely do back-to-back comparisons between various cameras, I was curious about the new sensor in the M10. I knew the new 24MP sensor in the Leica M10 has better higher ISO sensitivity, but how much better? It's definitely better than the older M240, but how about against the latest Fujifilm sensors, both the APS-C and the medium format sensor? I decided to shoot an impromptu portrait session with Marcello at Push Pull CafĂ© in Vancouver. Here is a quick indoor natural light comparison between the Leica M10, Fujifilm X-T2 and the GFX 50S at ISO 3200. Can you tell which sensor is which?
Can you tell which image belongs to which camera?
The first thing to mention is that both the Leica M10 and the Fujifilm X-T2 were shot with the latest Leica 35mm f/1.4 Summilux Aspherical. I used the Fujifilm M Mount Adapter on the X-T2, which made the lens a 52.5mm equivalent, while the Leica remained a 35mm. Because of this, I had to move the tripod closer to the subject to approximate the same focal length. The GF 63mm is a 50mm equivalent on the GFX 50S. Because none of the focal lengths and perspectives are equal (including the distance from me to the subject), the image will be different. Both the M10 and X-T2 have 24MP sensors, while the GFX has a 50MP sensor. The pixel density (relationship of how many given pixels per square area) is also different for each camera (X-T2 will have the most density, then the GFX, and finally the M10 will have the least, which is advantageous). Finally, both the GFX and X-T2 images are processed through the same in-camera Fujifilm X Processor Pro, while the Leica M10 images are processed through their own Maestro II processor.

Because of all these factors, there is no way we can compare these sensors on a level playing field. I did my very best to process all 3 images from their original RAW files in LightRoom with the least amount of post processing, trying my best to match the white balance of the back wall. Please do not focus too much on the colour or the rending of each image, as the lens, sensor and processor will all affect the look of the final image. My main goal was to look at high ISO performance, and the biggest surprise for me was how well the Leica M10 performed against the Fujifilm GFX 50S. Although the GFX ultimately has twice the pixels (allowing for a higher resolution image), the Leica M10 performed very well against the digital medium format camera when it came to high ISO noise performance. I also did outdoor tests at various ISOs, testing up to ISO 12,800, but I will save those for my full review.  Please click on each image for a full resolution JPEG.

Fujifilm X-T2 w/Leica 35mm Summilux Asph with M Adapter. ISO 3200 @ F/2.8. Full rez JPEG

Fujifilm GFX 50S w/GF 63mm f/4. ISO 3200 @ F/4. Full rez JPEG

Leica M10 w/35mm Summilux Asph. ISO 3200 @ F/4. Full rez JPEG
In conclusion, you can be the judge between the high ISO performance between the Leica M10 and GFX 50S. You can clearly see the X-T2's smaller APS-C sensor can not keep up with the larger sensor cameras at higher ISO, something that won't bother most cropped sensor shooters. You pay a premium for the luxury of high ISO performance of the Leica and the GFX, something that many of us are unwilling to pay for, myself included. At lower ISO, there is very little difference in visible noise, and the differences I do see have less to do with sensor size, and more to do with the lens I use and how the camera processed the image. What difference you do see between the sensors is micro contrast and dynamic range. As a general rule, larger the sensor the better, but how much better? This again is for another article. This was just a test comparing high(er) ISO performance.

My final thought between full frame (36mm x 24mm) and medium format (44mm x 33mm) is that there is a difference, but the gap isn't huge. The crop factor between the two formats is only 0.79X, versus the much larger gap between APS-C (24mm x 16mm) and full-frame of 1.5X. The biggest difference is the 24MP versus the 50MP sensors, something that Leica can address in the future, but at the cost of higher pixel density. For now, for the majority of photographers, the 24MP full frame sensor seems to be the sweet spot (although 36MP is still very doable), while the GFX's medium format sensor sits comfortably at 50MP (and capable of 100MP in the future). Unless there is a huge leap in sensor technology (e.g. organic sensors), the APS-C sensor should stay at a maximum of 24MP for now. Thanks for visiting and happy shooting.

X-T2, GFX, M10 at approx 100% crop (for X-T2 and M10)

*Thanks for all your support for my YouTube channel. I have just started affiliated links to B&H Video and Amazon. If you decide to buy any of these featured items, or buy anything on B&H and Amazon, I will receive a commission on the purchase. It won't cost you any more to use these links, but it will help me to continue making content like this. Thank-you so much for your support of my channel:

Affiliate Link to B&H: https://bhpho.to/2vGUl6O
Affiliate Link to Amazon.com: http://amzn.to/2erHF0d

Affiliate Link to Amazon.ca: http://amzn.to/2wpvq8L

Fujifilm GFX 50S: https://bhpho.to/2vHcQrq
Fujifilm X-T2: https://bhpho.to/2wIj4vj
Fujifilm X-T2 with Grip kit: https://bhpho.to/2wno7yr
Fujifilm X-T2 Graphite Silver: https://bhpho.to/2wI75O3
Leica M10 (black): https://bhpho.to/2wn64bM
Leica M10 (silver): https://bhpho.to/2wIu8IB








18 comments:

  1. At first I agreed with your conclusions but on closer inspection I notice that the X-T2 image is sharper. I wonder if the Leica image has less noise at the expense of a little sharpness. Especially around the vermillion border of the subject's lips it is noticeable. The GFX image looks great in terms of both sharpness and noise.

    ReplyDelete
    Replies
    1. Because I wasn't able to take these pictures all at the same time, everything from my actual manual focusing, to the movement of the subject would be affected. I did take pictures of still life, and the final results were the same. GFX had the best noise at higher ISO, then the M10 was a close second (up to ISO 3200) and then the X-T2. When the X-T2 was at ISO 6400, the colour noise was very clear, even without 100% peaking.

      Delete
  2. would love to see the raw files for each of these images, any chance?

    ReplyDelete
    Replies
    1. possible, but you're using an Unknown profile.

      Delete
    2. Hi, profile should be visible now?

      Delete
  3. I wonder how processing the XT2 file in Iridient Transformer would have impacted this comparison.

    ReplyDelete
    Replies
    1. Processing can also make a difference, but then all 3 files would have to be processed similarly. I know you can't make everything equal, but as equal as possible helps level the playing field. Overall, it isn't surprising the M10's noise performance is closer to the GFX vs the X-T2 since the sensor sizes and sensor technology (both are Bayer colour filter array but without an anti aliasing filter)

      Delete
  4. Hi, if I may also add (background in electronic design) that sensor size is somewhat of a myth from a pure physics perspective. Pixels regardless whether small or large are not aware of their size or the size of their neighbours. A case in point, if you were to tape off an APSC area on a FF sensor such as the M10 - would the noise performance and dynamic range of those pixels be any worse all of a sudden? Or in real world terms, how does a 5d mark 1 compare with a modern A6500! What is more important at least from a technical perspective is the technology and efficiency of the sensor design and surrounding imaging pipeline - how efficiently photons can be converted to a clean electronic signal.
    To that end, the sensor within the GFX (at least as far as any information that I can gather) is a couple of pixel technology generations behind the technology in the XT2 (copper wiring, smaller process) and similarly with the M10. As a result I'm not too surprised that you are seeing results similar between the GFX with the M10 even accounting for the noise benefits to the GFX from downsampling to the same native 24mp resolution. What is surprising for me however, is the difference in results between the XT2 and M10 in your testing here. The Leica sensor is not piggy backing off of the latest and greatest Sony technologies (copper wiring, small process, etc...) and traditionally Leica has lagged in this area. I've personally compared an SL and XT2 and found the SL wanting from both a technical and also financial perspective. I'd expect the XT2 and M10 to be much closer in performance as a result of the more advanced pixel level technology in the Fuji/Sony sensor.
    The results here are interesting too https://www.dpreview.com/articles/4109350402/welcome-to-our-studio-test-scene and the two are very similar in noise at least up to 6400 with the XT2 perhaps edging it. I'd note too that in the DPreview studio scene the leica sensor +lens combo produces a noticeably sharper image to the Fuji on ACR where the fuji looks to have had very heavy handed NR applied. Download the two raws though, run both through iridient and the tables flip with the Fuji producing a superior image from both a resolution and also noise performance! Perhaps demonstrating that the choice of raw developer in the digital world - particularly as it applies to PR 6x6 CFA sensors such as the one in the XT2 is as much a controller on overall image quality output as was the choice of Rodinal vs D76!
    Anyway, the irony of making such a nerdy post is that I shoot film largely with a Leica M6 and TriX400 - I like grain and I feel that the reason that one chooses a Leica is for the shooting experience. Doubling back on my ranty post further, I wouldn't buy a digital leica M10 today when I can get a similar shooting experience buying a second hand film leica body for a lot less money , or even get close to the digital leica shooting experience with an XPro2, albeit on the digital side with better color profiles and a unique 'look' from the Xtrans that I prefer on the Fuji. So really it's all about subjective shooting experience.
    I really hope you didn't mind me chirping in like this. I love love love your channel, work over at Fujilove, youtube videos. You are one of my favourite street shooters going and I really value how you interact with your subjects on the street. A sincerely nice guy. Keep up your good work :)

    ReplyDelete
    Replies
    1. Fantastic post, Tom, thank you

      Mats

      Delete
    2. Hi Tom,

      Thanks for your comment. It is very difficult to make all things equal as I mentioned. Different sensor manufacturers, different technology, different processors, different pixel density, different pixel sizes and technology, different colour filter array, etc. However, pixel density makes a huge difference. As you said, if you mask off the M10 sensor from 35mm (36 x 24mm) to APS-C (24 x 16mm) will the dynamic range and ISO performance change? Absolutely not. It's the exact same sensor, same processor, and assuming you don't change the total pixel count, it'll be the same. However, if you change the pixel density, meaning when you mask off the M10 sensor to an ASP-C size, but you also double the amount of pixels in a much smaller area, both having the same amount of pixels, then yes the image quality will significantly change.

      As for comparing the SL/M240/Leica Q to the X-T2, I had the same results. I even said many times on my blog and YouTube channel that the X-T2 sensor outperforms the Leica M240 sensor at higher ISOs. However, the new M10 processor is all new. That was mentioned in the very first paragraph of this article. It is not a new Sony sensor, but made by another manufacturer that makes this sensor exclusively for Leica. Since Leica does not upgrade sensors as often as other manufacturers, I was curious about the high ISO performance. I was surprised they didn't up the megapixel count, but I can now see why. They wanted to keep the pixel density low to improve on high ISO performance.

      This is why I said I was curious of how it would perform from the last generation sensor. It is much improved. I have tested all the Leica camera sensors in the last 5 years, and I've done the same with Fujifilm. I have lots of sample images and I am very confident this current test is an accurate depiction of high ISO performance comparison between the Leica M10, Fujifilm X-T2 and GFX 50S. I have more sample images, but this was just my first test. More coming up.

      Finally, thank-you for your support of my website, as well as my articles on FujiLove. As you probably know, I too am a huge film fan. Switching film is like swapping sensors, something you can't really do in the digital world (the Ricoh GXR got pretty close actually). As for your comment, I encourage dialogue. I find I don't get a lot of interaction via my website, so the more you comment, disagree, agree, share, the better!!

      Thanks again Dublin Tom!!

      BHT

      Delete
    3. Masking off the full frame sensor will NOT change the pixel density, there will just be LESS pixels being activated. It probably would not affect the IQ perception on a 11" monitor, but would definitely limit the cropability and printing size (kinda like the digital zoom on the Leica Q). which leads to another question - which is more important, the way an image looks on someone's monitor, or the way it looks printed large-ish? i'd say the gold standard is probably the printed result. But printing is a lost (and expensive) art and i'd say 99.9999% of images are ultimately viewed on various sized monitors, anyway.

      For my money I can't really tell those images apart, but all things being equal i'd rather shoot with that M10. As far as film - the magic is real. But the expense of film (don't give me that 'you should develop your own' business) and the hit-miss ratio really make it prohibitively expensive and not practical and ultimately negate any cost benefit derived from buying a $2000 M7 vs a $6500 M10.

      By the time i shoot a roll of cinestill and have it developed and scanned i've spent $24 for maybe 5 share-able images. So that's roughly $5 per 'keeper'. At that rate, it wouldn't take long to justify the exra $4500 for the M10 and it's bottomless supply of practice shots and multiple shots at varying exposures and framing angles to make SURE I got it right.

      Delete
  5. I find comparisons between cameras based on numerical ISO setting to be tricky, to say the least. IMO, it would be more helpful to compare cameras based on same/equivalent amounts of light given to the sensor. Thus, I would suggest using the same DoF and amount of motion blur, adjust image brightness to match during processing, and then compare IQ. after all, only "old film guys" (no offense intended) walk up to a scene and set the ISO first. Nowadays, typically, one would define the artistic parameters (DoF and motion blur and framing) first, then worry about signal amplification and image brightening.

    ReplyDelete
    Replies
    1. Hi Nixda,

      Yes it's very very tricky. The most objective way to do a test like this is using static objects in a controlled lighting situation. However, this test still gives a good overall idea of how these sensors perform at ISO 3200. The results will not be way different by equalizing things even more. I did my very best to be transparent with how I did this test, as well as pointing out areas where there will be differences. As for DoF, I did my best to match the X-T2's f/2.8 with Leica's f/4, since the cropped APS-C sensor has a 1.5x advantage in DoF at the same aperture. Again, you're right, comparisons are difficult. That's why I started this article by saying that I rarely post articles like this. I didn't want anyone to take this test as an absolute and definitive comparison, just something to look at and give them a general idea of the performance of these sensors.

      Delete
    2. I think it would be helpful to print the exposure values associated with each capture, or at least the exposure times along with the f-numbers used. There are significant differences between how these cameras derive exposure parameters from the metering. The X-Trans cameras, for example, bias exposure by -0.7EV from the metered values in order to provide 0.7EV extra highlight room. Thus, base ISO is 200 on these cameras, and it may in fact be more accurate to compare ISO 200 on the X-Trans cameras with ISO100 on the others.
      Another thing to keep in mind is that the camera might be set to a certain ISO value, but when the brightness of the image changes during processing, so does the ISO value associated with the output JPEG file. Say, you need to brighten the X-Trans image but 1EV in order to match its middle grey to that of the Leica image, the ISO is now 1EV higher than was set in the camera. I rarely find reviewers or bloggers take that into account when they present such comparisons.
      Therefore, I would again suggest to simply use equivalent exposures, match image brightness (through middle grey), then compare IQ, and basically not worry about the exact ISO number associated with the resulting images. These numbers aren't standardized to begin with, and when the midpoint is calibrated differently between cameras, they'll be off from the start.

      Delete
    3. Hi Nixda,

      Thanks again for your reply. Everything you mentioned has validity, but do you see then why when I started this comparison, I said it wouldn't be easy to make all things equal?

      Your first point of comparing the M10's ISO 100 versus the X-T2's ISO 200, would equate to testing the Leica M10's ISO 1600 to the X-T2's 3200, which is actually a disadvantage to the X-T2. I did think about matching the actual exposure (make sure both shutter speed and aperture were the same) but then that would give the advantage to the X-T2 since it would have a greater depth of field which would translate to a sharper image.

      In the end I decided to keep the ISO and aperture as similar as possible (the 0.79x difference between the M10 and GFX was less than a 1/3 stop difference so I kept the GFX at the same F/4, which put it at a slight disadvantage, although I feel the GF 63mm is actually sharper across the frame at that aperture than the Leica's 35mm Summilux). I also made sure the X-T2 and M10 used the same lens, but that also puts a disadvantage to the Fujifilm somewhat. However, if I used the XF 23mm f/2 lens up against the Leica 35mm Summilux Asph, I knew people would say it was a huge unfair advantage for Leica (although I disagree).

      As I mentioned, I have more comparison tests I took with all three cameras, again trying my best to equalize as best as I could. However, the lens, processor, sensor technology, etc. all affect the final image. I can get an overall idea after using all three cameras for a few weeks in various lighting situations and at various ISOs.

      Remember I had a pre-production copy of the X-Pro2 and X-T2, so I've probably shot with these cameras more than any other reviewer. I also had a pre-production copy of the GFX, so I really have a good idea of how these files will look at any given ISO. I did only have the M10 for just over 2 weeks, but I put a shot a lot of images and could easily compare against the other two cameras. The Leica M10's new sensor is amazing. It's much better than the previous 2 sensors (I've tested every Leica M body of the past 5 years). It outperforms the X-T2's sensor at higher ISO for noise, and is very close to the GFX's sensor from ISO 1600-6400. I don't think my results are out of line with other reviewers, and once DPReview finishes their long term review of the M10, I think we will see how their more controlled studio results will also reflect similar findings. Saying all this, I think we all know that high ISO noise is not the be all end all test. Colour saturation, sharpness, micro contrast, dynamic range, these things also matter. In fact, I can put up with some noise if I don't lose sharpness, something that the previous X-Trans sensor was criticized for (waxing skin and smudging of fine details like grass).

      Delete

  6. It's going to be finish of mine day, except before finish I am reading this enormous article to improve my know-how. craigslist colorado springs

    ReplyDelete
  7. Sorry for the late comment but in my experience, ISO3200 etc with a Leica M10 is not the same as ISO3200 etc with Fuji, and in my casual tests, is not even the same as the Leica Q. If the aperture & ISO are kept constant between the Q and M10 then there seems to be a big difference in the shutter speed need to produce the same exposure. A couple of stops in favour of the Q in my experience. I'm wondering if, like Fuji, Leica has gamed the M10's ISO to make it look good.

    ReplyDelete
  8. You are creating amazing content these Days.

    ReplyDelete

Note: Only a member of this blog may post a comment.