Fujifilm GFX100 II Engineers and The Truth About the “New” Sensor and “Up to Twice” Readout Speed
Fujifilm GFX100II
Over the past few weeks, many have contacted me asking why I don’t talk about the reports that claim:
- the Fujifilm GFX100 II sensor is not new
- the Fujifilm GFX100 II does not read out at twice the speed and faster readout speed is achieved only with software tricks, meaning dropping down at 12 bit when in 8 fps mode
The answer is simple:
When Fujifilm published their first Fujifilm GFX100II development story episodes, they said they would talk about the GFX100II sensor in future episodes.
This is why I decided to wait, hear Fujifilm’s position, and then write an article that is as comprehensive and complete as possible.
And I am happy I waited, because in the 5th X Lab episode dedicated to the GFX100II, we get some helpful insights in the inner workings of the sensor (full episode with summary below)..
But before we dive into it, let me express my admiration for Jim Kasson’s brilliant work in analyzing the inner workings of the GFX100II and to Petapixel for bringing his findings to the attention of the larger public.
I decided to wait for more details coming from Fujifilm before reporting about it, but maybe they would not have disclosed those in such detail, if Jim would not have done his great work, and Petapixel made a huge buzz out of it.
The Truth in a Nutshell
We will go more in depth into each point further down in the article.
But in short we can say this:
- there have been so many changes on the sensor stack, that we can indeed consider it a new sensor (and it is capable of achieving performances the previous sensor could not)
- the sensor can read out indeed up to twice as fast, but Fujifilm had to decide when to allow it read out faster and when not based on several criteria (for example power consumption)
But there is more to it. So let’s jump into it.
Fujifilm Isn’t Telling the Whole Truth About the GFX 100 II
Before we dig into the fact if Fujifilm lied to us or not, I would like to catch up on the headline of Petapixel: “Fujifilm Isn’t Telling the Whole Truth About the GFX 100 II”
Well, I agree, Fujifilm is not telling the whole truth.
But let me ask you: did Sony tell us why the electronic shutter of the new Sony A9III is reduced to 1/16,000 at larger than f/2 aperture? Nope, they didn’t. So they are not telling us the whole truth either!
And why does Nikon refuse to admit that the Nikon Z9 sensor is actually a Sony sensor? Why are they hiding such an obvious fact in front of their customers?
Truth is: camera companies, all of them, do not share everything, because they consider some parts of their cameras as confidential.
I hate it. I would like to get an answer to every single question I have. But I have to accept that if camera brands don’t want to, they also don’t have to give us answers.
Fujifilm was actually very transparent about this, as they even openly said at launch that they are not going to disclose all GFX100II details, as they consider some stuff on their cameras as confidential. In fact, in an interview with CineD (we reported here), they were asked about how the magic happens in F-Log2 D-range priority, and Fujifilm answered.
F-Log2 D-range priority: pick that in the menu for wider dynamic range. Camera reading is “a bit different” and this is actually confidential technology but it gives wider dynamic range
With that said, while I get it that Fujifilm does not want to disclose F-Log2 D-range priority secrets, telling us about the 12 bit at 8fps would not be that big of a secret, especially considering that it is common practice among all camera brands to lower bit precision at the highest frame rates.
Did Fujifilm Lie to Us?
Now back to the original topic.
While I applaud and admire (and share) Jim Kassons work here on FujiRumors, there is one limitation to it.
His claim that the sensor can’t really read out faster and it is actually the same sensor of the GFX100S, is based on the fact that at 8fps the GFX100II achieves faster readout speed by dropping down to 12 bit. And that’s a fact that I don’t question here.
My main issue is that just looking at one aspect of the sensor performance (8 fps) and concluding that the sensor is never able to read out twice as fast without software tricks might not be the best thing to do, especially considering that, as we will show below, Fujifilm stated that the faster readout speed has mainly advantages in terms of video rolling shutter, which sadly is something that has not been tested by reviewers so far.
- Product Page
[in the video section of the product page] High speed rolling shutter: Follow high speed subjects with confidence – GFX100 II’s increased readout rate suppresses rolling shutter for true-to-life results
In addition to that, Fujifilm says in today’s video that they deliberatly chose not to enable fastest readout speed in certain situations (for the reasons we will share below).
With that said, did Fujifilm really lie to us when they said the Fujifilm GFX100II sensor reads out “twice as fast“.
Well, if Fujfiilm would have really said that the sensor reads out “twice as fast“, then yes, they would have lied to us.
The problem is that they didn’t say this in their official launch material.
What they really said is that it reads out “up to twice” as fast, and this changes everything.
- Press Release
The GFX100 II features the newly-developed 102MP high-speed image sensor “GFX 102MP CMOS II HS” and the high-speed image processing engine “X-Processor 5” to deliver up to double the signal readout speed compared to the current models [GFX100 and GFX100S] […] Users can film a moving subject naturally due to the increased sensor readout speed and suppression of the rolling shutter effect.
There is huge difference between saying “up to double” or “double”. And today we learn that there are situations in which the sensor reads out faster, because Fujifilm wants it to, and other shooting conditions in which Fujifilm decided to keep the readout slower, because not always a faster readout is needed or beneficial (see below).
And while video rolling shutter has not been tested, we know thanks to Jim’s testing that there is no increased readout speed in stills shooting.
Why?
My guess is that Fujifilm decided not to increase readout speed in stills, because in stills people are going to use the mechanical shutter anyway where rolling shutter is not an issue, allowing Fujifilm to keep the benefits of a lower sensor readout.
This makes a whole lot of sense especially considering that, with full sensor readout, the mechanical shutter is faster anyway (8fps MS vs 5.3fps ES). On contrary, on cameras like the X-H2S, where the electronic shutter unlocks higher frame rates, a faster sensor readout is crucial to keep rolling shutter in check.
New Sensor or Not?
Also in the video today (summed up for you below) Fujifilm insists: this is a new sensor. Not the silicon wafer itself is new (that’s something they never claimed, as it is still 102MP BSI CMOS), but the readout circuits, AD converters, mircorlenses etc are improved.
In fact, during the X summit here, Fujifilm said the sensor is new because of the “newly designed readout circuit” and “new micorlenses”
Now, one could argue that the readout circuit and micorlenses are not the sensor.
But to me reducing the performance of the sensor to its silicon wafer alone would be like reducing the speed of a vehicle to the power of its motor without considering the impact of its aerodynamics, weight etc on its actual top speed performance. I mean, if a car and a truck share the same horsepower, it does not mean they will also reach the same top speed.
In fact, just as speed performance of a car is a combination of several factors, so the sensor performance is the combination of it’s pixel structure, AD converters, readout circuit, microlenses, color filter array (X-Trans for examples also influences readout speed and image quality) and more.
So I guess the changes made on the whole sensor stack are enough that we can accept this indeed as a “new” sensor. And it comes with improvements, such as less rolling shutter in video.
Some Questions are Still Open
So no, Fujifilm did not lie to us: this is a new sensor, if with “new” we mean the whole sensor stack (silicon wafer, readout circuits, microlenses, etc).
And they probably did not lie to us even about the “up to twice” readout speed, as there might be situations, especially in video, where certain specs (4K/60p) and reduced rolling shutter can only be achieved thanks to the newly designed readout circuits that allow for faster readout.
And yet, the fact remains that Fujifilm’s formulation (“up to twice as fast sensor readout”) might be accurate, but remains vague.
So I believe Fujifilm should give a more detailed explanation about the “up to twice the readout speed” over the GFX100S, so that people know when exactly and under which conditions they can expect a readout improvement over the GFX100S.
Does this improvement apply only to video (less rolling shutter)? Or does the GFX100S drop to 12 bit at 5fps whereas the GFX100II drops to 12 bit only at 8fps? Are there other benefits resulting from the improved readout speed, such as higher AF calculation frequency? Is the EVF blackout shorter on GFX100II over GFX100S when used at the same settings?
Now, I get it that Fujifilm did not dig deep into this stuff at launch day, since product launches are not here to bore people with lengthy technical explanations, but more to trigger the emotions of potential customers.
I mean, just to make a fresh comparison, also Sony did not disclose in their Sony A9III launch material that the camera has no dual gain ISO, how high the dynamic range is, why it starts at base ISO250 only, why the electronic shutter is reduced to 1/16,000 at larger than f/2 aperture.
But if Fujifilm wants to get rid of the more or less veiled allusions that they cheated, time has come to disclose more than what they did until now. And while todays video goes quite a bit more in depth into this matter, at least in its google translated version it is still not the full technical explanation we want.
The Reasons to Upgrade from GFX100S to GFX100 II
But besides all tech talk and drama, what I can tell you from my own GFX100II hands on experience is that the GFX100II feels more than twice as fast than the GFX100S in terms of autofocus, simply because the algorithm is way superior, the processor much faster and the deep buffer allows for long bursts I can only dream of on my GFX100S.
At the end of the day, why would one upgrade to GFX100II over GFX100 or GFX100S? These are in my opinion the main reasons:
- vastly improved autofocus algorithm (better eye AF, more reliable tracking, tracking from further distance)
- richer autofocus features (subject detection)
- slight increase in dynamic range at base ISO (30% better, which means 1/3 of a stop better)
- better corner sharper performance thanks to new microlenses designed for better performance on the edges of the frame
- better IBIS performance
- vastly improved video specs
- snappier camera thanks to X Processor 5 and CFexpress Card
- battery grip option
- superb build quality
- Frame.io integration
- best EVF on the market
Last but not least, at the bottom of this article you will find the video as well as a summary I made out of not optimal google translation.
Get Yours
- Fujifilm GFX100 II – New
BHphoto / Amazon / Adorama - GF 55mm f/1.7 R WR – New
BHphoto / Amazon / Adorama - GF 30mm f/5.6 Tilt Shift – New
BHphoto / Amazon / Adorama - GF 110mm f/5.6 Tilt Shift – New
BHphoto / Amazon / Adorama
GFX Savings
- Fujifilm GFX100S – save $800
BHphoto / AmazonUS / Adorama / Moment / Focuscamera - GFX 50S II body or w/GF35-70 – save $800
BHphoto / AmazonUS / Adorama / Moment - GF 32-64mmF4 – save $450
BHphoto / Adorama / AmazonUS - GF 45-100mmF4 – save $450
BHphoto, AmazonUS, Adorama - GF 100-200mmF5.6 – save $400
BHphoto / AmazonUS / Adorama - GF 80mmF1.7 – save $450
BHphoto / AmazonUS / Adorama / Moment - GF 110mm F2 – save $550
BHphoto / AmazonUS / Adorama
- newly developed 102 MP sensor, the GFX102II CMOS II HS (HS for High Speed)
- speed was not enough with the previous sensor
- older 102 MP sensor allowed for 4K/30P max
- the goal was to achieve up to twice the readout speed
- that’s difficult to achieve with such a large sensor
- there are 102 million wires and it takes some time for the electrical signal to stabilize
- if you simply make the signal faster, then the signal will become distorted and there will be a lot of noise
- Both analogue and digital circuits become faster and there is a part were electrical signals and converted into digital signals by AD converters
- the speed of the digital signals has also been increased
- how fast does it tick? It’s called the “counter”, and counters are getting faster
- but if you make everything faster, it consumes more power, so Fujifilm has to balance that too
- Fujifilm controls the power supply in a very detailed way: when they want to increase speed, the do that by using electricity all at once and take a break when they need to
- This is how they achieved twice the speed
- Fujifilm is also paying attention to the disadvantages of making it faster
- 8fps bursts with mechanical shutter, but X series (X-H2S) is better for sports, even if now the AF algorithm is the same
- considerable eye autofocus improvements
- new zone custom feature, where now you can choose freely the Zone AF box size and shape (previously only three fixed sizes for zone AF tracking box available)