Fujifilm’s GFX 100 – A medium format mirrorless camera | Hacker News Reader

Fujifilm’s GFX 100 – A medium format mirrorless camera

theverge.com

99 pointsTomte91 comments

I understand that's the actual, original title, but it is so semantically meaningless. GFX is a "Mirrorless Camera" (with a large sensor). So of course it "performs like a mirrorless".

I get what they're trying to say ("This is not your daddy's medium-frame"), but the notion that these are somehow qualities intrinsic in a "Mirrorless" designation irks me. (Certain) DLSRs were fast, small, portable, effective and efficient long before mirrorless.

As well, the notion that GFX isn't a "Mirrorless", or that "Mirrorless" implies a particular sensor size (such as m4/3rds or similar) is again erronenous.

Medium format was expensive, unwieldy and slow due to artificial market segmentation, limited competition, and essentially being at the far *-phile end of the price/feature curve.

Fuji is a new entry and an underdog competitor, which is what drives most of the attractive features, not just the mirrorless part.

All they mean is that medium format cameras are perceived to be slow, while 'mirrorless' cameras are not. Its no different than people thinking "AMD chips run hot" when that hasn't been true for years.

There is a lot of misinformation, incorrect assumptions, and sacred cows in the world. If every article was written to correct those assumptions, then (1) it could consume the entire article (2) nobody would actually even read them (3) no ad-revenue would mean such publications would eventually die out.

Which AMD chips ran hot? For me nothing will ever compare to the P4 for thermals (even though modern chips have higher power consumption).

The Athlon XPs did run hot. I don't have a handy link right this moment. Maybe someone can look it up..

The 15h (Bulldozer to Excavator) series had pretty lousy efficiency, with most desktop parts having a TDP of 95W or 125W. The FX-9590 achieved some level of notoriety for having a 220W TDP despite only having four real cores.

The real problem with bulldozer was by that point they were so far behind nobody in my circle bothered mentioning them at all.

I would agree in principle; in practice, my main desktop is still FX-8350. I keep looking for excuse to replace it... and not finding one :-/

Still available as new in mainstream shops [1], and at recent times strangely peaked specific price/performance chart. It boggles my mind to be honest :P

1: https://www.newegg.ca/p/pl?d=fx-8350&N=-1&isNodeId=1&Submit=...

That's fascinating though, because Mirrorless cameras until so very recently were "the slow ones", and very much perceived as such, primarily and quite legitimately due to:

* refresh rate of EVF (vs optical viewfinder);

* Contrast Detect vs Phase detect autofocusing;

* as well as some design/control/menu choices

Even to this day, highest-end mirrorless are just starting to be trully competitive in the professional sport / action arena.

In other words, no active photographer I follow would use "mirrorless" as a universal shortcut slang for "fast, speedy, responsive". On the contrary...

So I don't feel it's a case of "every article needs to work hard to dispell myths"; but rather, an article not actively, for no seemingly good reason, starting wrong myths from scratch :).

You are comparing historical mirrorless performance with DSLRs, but article is comparing them to medium format cameras, two different things. Medium format cameras were slower than both, older consumer/compact mirrorless cameras as well as DSLRs.

BTW.. Its fair to assume that these myths are common, than to assume that someone purposely chose to perpetuate them with bad intent.

Oh I don't at all feel it was bad intent and vicious ulterior motive :)

As I said, the title just irks me :>

Fuji also has a history with medium format rangefinders, lenses for which translate well to mirrorless digital designs since rangefinder cameras are... mirrorless.

That is a damn good argument. You should also include TLRs in the discussion.

I thought it was referring to the sensor stabilization. Medium format is generally just for studio, where you wouldn't need or want that. Seems like Fuji means for this to be portable... like any other "mirrorless".

> Medium format was expensive, unwieldy and slow

?! There are plenty of very pocketable, no fiddling required cameras that take 120/220 film from the 50s onwards. I have a few that I still regularly take on hikes and trips around town (although for shorter trips I can deal with the slightly bigger footprint of a TLR), and let me take pictures with a detail and tonal range most digital cameras can only dream of (particularly smartphones), all about the same size and weight as my Anker battery pack.

Yes there are also crazy bulky Mamiyas/Hasselblads etc but those were always targeted to studio use.

Come on, 500 series 'blads were far from bulky and were totally walk around cameras. If you are talking about RBs and RZs, I agree.

...and none of those were as bulky and unwieldy as Fuji's own GX680.

Apologies; I should've explicitly indicated "Digital Medium Format", as the most appropriate "like for like" comparison - shows my biases :->.

(I would edit for clarity but I believe time has expired for that)

Thx for pointing it out!

Here's a sample picture at full resolution (100M pixels):

https://1.img-dpreview.com/files/p/sample_galleries/60445538...

The amount of details is impressive.

I agree as is the dynamic range in the shadow areas of the image. You can see a few people in heavily shadowed areas as well as detail on white buildings that otherwise would be blown out.

If you look at the fine details, everything is soft. Not one single line in the whole photo looks crisp and sharp.

This is not an issue with the camera, mind you — just a fact of life when taking photos at this sort of range — but it does highlight why you don't get nearly as much of a benefit from that 100MP sensor versus their 50MP medium format cameras as you'd think you do.

This is undoubtedly a brilliant studio camera for applications like fashion or product photography, but completely and utterly pointless as a landscape camera, where physics severely limits the benefits you're getting for the extra price.

> If you look at the fine details, everything is soft. Not one single line in the whole photo looks crisp and sharp.

This is generally true with color digital cameras that use color filter arrays, which is most digital cameras. I have yet to see a camera with a color filter array which looks "sharp" at 1:1 pixels.

To me the detail looks quite acceptable, but I'm not directly comparing it against other photos so I might be wrong.

> I have yet to see a camera with a color filter array which looks "sharp" at 1:1 pixels.

Check out Foveon sensors in Sigma cameras. They are really impressive.

agreed, every pixel from the 3-layer foveon images is great, they have a very unique look

they dont have a color filter array. They have three distinct color sensors, very different tech from usual cameras.

> To me the detail looks quite acceptable, but I'm not directly comparing it against other photos so I might be wrong.

Oh, absolutely — the detail is pretty good. My point is simply that "pretty good" is the best you can achieve in this situation, because the limiting factor for the amount of detail is distance and atmospheric conditions, not the sensor — but that, in turn, shows that the step up from 50 to 100MP is not that useful, rather than showing off the advantages of the 100MP sensor.

I’m not sure what is expected, because it’s only a 40% difference in resolution.

That's not inconsiderable.

I use both a FF and a 1.5x crop APS-C camera, and part of the reason I got the APS-C body at all is because I get 24MP from that instead of roughly 11 for the equivalent crop from my FF body when I go birding.

Interesting problem to have. I assume you already have the longest possible focal length lens for your camera system?

Pretty much, yeah. I'm using the a7iii and the a6400 with the 100-400 GM. Only way to reach any longer on all-sony is by adding a teleconverter to the mix. At any rate, it's fairly common advice to use e.g. a 7D mk2 for wildlife even if you're mostly a 5D user.

Any reason you don't go for the 1.4X or 2X teleconverter? Would the loss in shutter speed make it a wash for birds, where I assume that actually matters a lot for birds in flight?

I still have the option to grab the TC at some point and potentially use both the a6400 and the TC in tandem (though I doubt the lens is quite sharp enough to justify that — odds are the 100-400 + 1.4x already has a circle of confusion bigger than the pixel sites on the APS-C sensor)

The reasons I went with the APS-C body instead of the TC is that I actively wanted a second body for other reasons. Namely: I wanted a smaller, lighter body that I'm happy to just leave in my backpack all the time, and I wanted the versatility of having two lenses available at all times for some situations.

Also, not the most amazing photo ever, but definitely a good example of the sort of sharpness you can get out of this — https://i.imgur.com/p9H42aG.jpg

> I have yet to see a camera with a color filter array which looks "sharp" at 1:1 pixels

This is why I was absolutely blown away when I got my first high-DPI screen: a MBP in 2013.

At the time I was a pro photographer. Because the pixels were half the size as before, a true 1:1 viewing looked so much sharper. I found I could tune my sharpening (almost always required) for print based on screen viewing alone, and no longer had to run as many test prints.

In addition, in the grid/filmstrip views of photos I could set the thumbnails to just slightly bigger than the old 35mm slides, and found I could quickly edit down my photos often just based on that alone.

Excuse the rant... for some reason it triggered a moment of nostalgia about how amazed I was with that screen.

And it's a JPEG. It would be interesting to see the RAW image.

The image is soft, probably out of focus. Large megapixel sensors tend to accentuate small focus errors, which is one of their downsides (another are the huge raw file sizes). With 100MP, you have to make extra sure your focus is spot on, and perhaps use a tripod.

I agree with the other replies that mention the image softness with 100MP sensors, and how hard it can be to achieve image crispness.

That said, I think they are missing the point. This is a 100MP medium-format camera for $US10k. This is markedly cheaper than comparable cameras from other medium format manufacturers. The big question is how the GFX 100 stacks up to the competition in other areas.

I don't quite understand what I see: it doesn't look like lens blur. It's more like the bad JPEGs from my Fuji S5600. But why would dpreview not do a proper raw conversion?

It's a pre-production camera. Typically reviewers don't get to see anything but JPEGs until it's released.

Meh, there's terrible smearing of details in the leaves, grass, and stone due to clumsy noise reduction. It will be interesting to see what a decent raw converter can do with files from the camera.

Not a fair evaluation - in a standard resolution picture, say 24MP, with the same FOV, that grass patch would be a green blob.

Yeah but you could buy a 24MP camera and a whole kit of high quality glass for the same price as this body.

And that's enough for most people. MF cameras have a specific set of uses and some trade-offs.

What you're seeing is the physical limits of resolving detail through thousands of feet of warm air. There's probably no need for any noise reduction in these conditions.

True. The closer shots sadly also have a plastic look from noise reduction. I'm sure that a decent raw converter will get much better results out of the files.

People here talking about image quality should be aware of this being a jpeg. I'd say most problems discussed are jpeg issues. You can't discuss image quality over a lossy codec.

How so? JPEG compression is orthogonal to resolution. At the macroblock (or per pixel) level, a 100MP image is not degraded more than a 0.3MP image. So if softness can be seen in a 24MP jpg, it should also be discernible in a 100MP jpg.

Correct, but with a higher resolution image there's more incentive to use higher compression (lower quality) to bring the file size down.

Indeed. Even 1/8 crop images of that photo look great.

As for some of the negative comments, pixel peepers are a bane to the camera industry. Fact is, there's always a limit to the amount of detail that can go through the lens itself. Our eyes only see so much. This is why MF doesn't matter for most people. Even FF is too much for the average person. Our eyes can't resolve more than 4k at most distances, so that allows a lot of cropping even at 16mp. MF is for blown up studio or landscape images. Zooming in on a 100mp image taken at a long distance in the first place searching for flaws is entirely missing the point.

Tldr: this camera and the image are fantastic, pixel peeping is silly.

That's a bit of an odd title. This is a mirrorless camera.

I always found it interesting that medium format cameras generally don't have the fastest lenses. The fastest autofocus primes I see are f/2.8 and f/4, which are roughly equivalent in speed to an f/1.6 or f/2.3 respectively. I'm sure the quality is better, but I'd love to know why manufacturers don't seem to be pushing the aperture as hard with medium format as they do with full frame.

Of course, I'd love to play around with one of these... Even if it would be massive overkill for anything I'm doing.

> why manufacturers don't seem to be pushing the aperture as hard with medium format as they do with full frame.

It's usually a lot more expensive to manufacture larger aperture lenses, for one, and the more the aperture increases the more your focus is going to be razor sharp (1.8 on medium format will mean your focus fully open will probably be only a few mms deep in short distance) so the applications are very limited. And for further away subjects then you won't see much difference anyway. Digital sensors have variable ISOs and not too much noise even at several thousands of ISOs now so you could take pictures in the evening even with a narrower aperture.

Oh, no doubt it would be expensive, but medium format digital cameras already tend to be prohibitively expensive for anyone who doesn't have an urgent professional use for one.

If you can do without AF then most of the cost is just the back. You can totally shoot digital MF economically if you're open to not using the latest greatest back.

I paid about 1.5K total for my back, body and lens, to provide one anecdote of how cheap it can be.

There is a Fuji G 110mm f/2 prime which is autofocus. Its image quality is sublime.

There are number of reasons as you can imagine. Those lenses would be enormous (think 10lb lens + camera). It is very difficult to manufacture large lenses (with high quality), requiring a complete change in the tooling. It would also be labor intensive and thus only a very limited quantity can be made. At f/1.8, the lens price would probably be you just directly paying the technicians salary plus overhead.

So you want like the medium format equivalent of a Noctilux, but with autofocus?

I'm surprised something LIKE that doesn't exist, but I'm thinking more of f/1.2 equivalent glass.

>but I'd love to know why manufacturers don't seem to be pushing the aperture as hard with medium format as they do with full frame.

For one because the DoF would be so small as to make focusing almost impossible wide open...

This is true. I have an 80mm f2.8 from a 6x9cm camera that I mounted onto a recessed 4x5 lens board. Wide open the depth of field is so insanely shallow that it's tough to focus even under a dark cloth.

I have a 110mm f/2.8, it took me several weeks of experience with the lens and camera system before I could confidently get accurate focus. But I think that the shallow DOF is a point in its favor, because if you can get good focus at f/2.8 you're going to get a very sharp image at f/11.

That, and the sheer amount of glass you'd need would make for an excellent workout session every time you went shooting.

Say, a 250mm f/1.4 with an image circle large enough to be useful on 6x9 would be quite the beast (and impossible to focus on my Texas Leica.)

You've got computers to focus for you, for the reasonable price of adding a motor strong enough to move all that glass ;)

> I'm sure the quality is better, but I'd love to know why manufacturers don't seem to be pushing the aperture as hard with medium format as they do with full frame.

Wide open, the lens is less sharp and you are going to have large portions of the image out of perfect focus. At that point there is no point using MF, you might as well use a smaller camera.

My personal experience... I mostly use a 110mm f/2.8 prime, and when I shoot wide open it's beautiful, but at that aperture I'm definitely not getting near the sharpness that the system provides.

But all of this applies to full frame with an f/1.4 85mm lens as well, yet even f/1.2 isn't that rare in the full frame world.

That is incorrect, it does not apply to an f/1.2 full-frame lens, because you cannot switch to a smaller format. At least not without spending an exorbitant amount of money.

I said it applied to an f/1.4 lens. In that case, you'd have to switch to an APSC f/0.95 lens. Those exist, but only in manual focus... and they're only around $500

If you're shooting MF at f/2.8 on a 110mm lens with ~0.5x crop factor, you'll get similar results to shooting 55mm at f/1.4. So, not much advantage to making MF lenses with larger aperture.

If you're shooting f/1.4 on a full-frame camera, and swap to APS-C at f/0.95, you're not getting similar results. An f/0.95 is a somewhat extreme lens design and will have noticeable aberrations wide open. At least, the f/0.95 lenses that I've seen.

For the same reason, you rarely see LF lenses that open wider than, say, f/5.6 or f/4.

> The fastest autofocus primes I see are f/2.8 and f/4, which are roughly equivalent in speed to an f/1.6 or f/2.3 respectively. <

Can you elaborate on that? What do you mean by "equivalent in speed?"

Equivalent to ƒ/1.6 and ƒ/2.3 on a "full frame" sensor.

Similar depth of field and total light gathered.

I'm aware that sensor size doesn't change depth of field, however if you compensate the focal length to produce the same field of view on both sensors, you then need to compensate the aperture to arrive at the same depth of field.

So a 50mm f1.4 on a full frame sensor would be roughly equivalent to an 85mm f/2.4 on one of these digital medium format sensors.

The best way to look at total light gathered is to talk about the aperture diameter, not f-ratio. The f-ratio is what it is; it doesn't make sense to multiply it or divide it to 'compensate' for different aperture diameters. You should just use the aperture diameter at that point.

And an f/4 for any size format will be gathering the same amount of light per sensor area. It's just that the sensor area is larger on larger formats, so the total amount of light (but not light per pixel!) is more. This is why telescopes have such wide apertures; they're limited by total amount of light.

The amount of light per area at the sensor is directly proportional to the area of the aperture and inversely proportional to the focal length. That's why we use the f-number, which is a ratio of focal length to aperture diameter.

Please note that the f-number doesn't indicate total light hitting the sensor. It indicates total light per unit area at the sensor. Thus, for the same f-number, a larger sensor will gather more light.

All of this info can be found in the wikipedia page: https://en.wikipedia.org/wiki/F-number

I don't really get the obsession with fast lenses, shooting wide open suits a fairly limited range of subjects, and generally lens performance is terrible wide open. I have some Zeiss Planar f/1.4 lenses for 35mm, and I basically never shoot wider than f/5.6. Also never really shoot wider than f/8 on medium format Zeiss CF lenses. But I don't shoot stuff that needs fast lenses I suppose.

In broader photographic terms (outside of full-frame / mirrorless), f/4 is fast.

I suspect that people shooting medium format aren't as interested in having awesome bokeh and a super fast lens with a lower number, and instead care about lens performance in general.

This sensor has a 55 mm diagonal vs the 43 mm diagonal you see on a full-frame sensor. That's 1.7X the total sensor surface area.

I guess there's some people who will "need" this, but the vast majority of photographers (even professionals) can make do with smaller and cheaper cameras/lenses. I have an M4/3 camera on the other end of the scale that fits in my jacket pocket when I put a pancake lens on it.

Note that while this indeed qualifies as medium format, it's pretty far from the classic 6x7 film format. As a fan of medium format (something about the compression and distorsion of a portrait lens, but with the viewport taking in more environment), I really hope we're going to see a more "true" medium format in the future.

Assuming equal resolution (i.e. not equal pixel size), the nice thing about moving to medium format sensors from full frame is that it relaxes the tolerances required to create high quality lenses as far as optical quality is concerned. But OTOH, you have to deal with a larger mass of glass. On the sensor side, the yield on the wafer is probably going to be very low when you have to create such large sensors without a single hot pixel. Then there's the total light gathered, which is always going to favour larger sensors, as long as you can get your exit pupil to be large enough. Its going to be interesting to see how this plays out over the next decade.

What test would you run to see if your camera's sensor had a hot pixel which was replaced in processing by neighboring pixels?

Well, for one, neighboring pixels are already being used during demosaicing.

But I understand your question. For an end-user to detect such overt manipulation, you'd probably need to take successive frames with the lens cap on, and observe how the RAW sensor values change over time and find patterns in the noise levels. A pixel thats duplicated will stand out. That's all I have :^) I'll have to think a bunch to understand it at a deeper level...

Fuji was a major player in the rangefinder (mirrorless) medium format film cameras market (G645, G690W, etc), so this is a good return to form for them!

Well they also have lots of other mirrorless digital cameras, the newsworthy part here is that this one is a medium format camera (of which there aren't too many with or without mirrors, it's a very niche segment).

I always found it hard getting lenses for my XT-10, but maybe I am just bad at looking. The kit lens it comes with is not great, and it seems the only other options are £300 prime lenses.

Hmm; I don't want to say "It's you..." ;), but the Fujifilm X mount is one of the most developed of the "new" mounts: http://www.sansmirror.com/lenses/lens-database/lenses-from-c...

In particular, the zooms: https://www.fujifilm.com/products/digital_cameras/x/index.ht...

Now, it IS expensive, on average. But by all accounts their lenses are mostly worth the price. And 300 for a prime is largely normal.

In all mounts, lens price is variable but high when you get to even remotely non-junk stuff. In my world of Nikkors for example, the Nikon 85mm 1.8g is widely considered a "bargain", at average price of $500 (compared to the $1,500 1.4G version).

I own the $100 "nifty fifty" and the $2000 70-200mm 2.8, and that's all prosumer country. Functional exotics can be $5000-$8000 for working pros (not Leica luxury items;).

Do you know https://alikgriffin.com/a-complete-list-of-fujifilm-x-mount-...

Given the insane advantage today's lens designers have in terms of materials, coatings, lens simulators, etc it is becoming harder and harder to design a bad lens. Maybe you just have a manufacturing defect in yours?

> becoming harder and harder to design a bad lens.

On the other hand it's not so much of a problem anymore to use a "bad lens" when cameras do signal processing to remove distortion and chromatic aberrations. On film this would have been impractical to correct, on digital medium it's almost done real time.

That's not really accurate. You can't 'remove' optical flaws, but they can be masked - to an extent. Correcting for those flaws via software isn't 'free', you can only do it at the expense of image quality.

A high quality £300 prime lens is not a bad price. The 35mm f/2 essentially lives on my XT-2. For its size, weight, and price, it’s a terrific lens.

I agree and use the 23mm f/2 more than my other primes. It is small and keeps the weight distributed nicely on the body. Although the 56 mm f/1.2 is a king for depth of field and bokeh. I sold my 16 mm and kept the 23mm so wider shots.

I have a Fuji XT20 with the 18-55 kit lens. I also own a number of Sony and Canon cameras.

If I rate the Canon 70-200 F2.8 at A+, and the Canon 50mm 1.8 an A, then I rate the Fuji 18-55 kit lens at perhaps an A-. And its rather cheap for a zoom lens if I remember correctly.

Yup, I have an XT2. I had to buy a grey market zoom because the kit pricing for the graphite body is insane or not available (can't remember which), but the kit lens was something like $350.

First of all, if it wasn't obvious to anyone, the Fujifilm X series is totally different from the GFX line discussed in this post.

> The kit lens it comes with is not great

If you're talking about the 18-55mm, it's considered quite good (and good value) for a kit lens.

Fujifilm has probably the best lineup of lenses for APS-C sized sensor. They're not cheap, but they're well-made and cheaper/smaller/lighter than equivalent lenses for full-frame cameras, which is why I love the X-mount, especially for travel.

Micro Four Thirds is an even smaller sensor, but it's standardized across many manufacturers so there are many options.

You can buy a lense adapter, and go looking for some cheap old lenses. I just bought two old Zenit lenses: Helios-44 (58mm) and Jupyter 37A (135mm) + single adaptor, all for ~100 eur, and I'm pretty happy with them!

I'm not sure if these exact lenses available where you live, but for sure there should be some options of nice old lenses you can buy for little money.

UPD: I have x-t20

Reply on news.ycombinator.com