GFX100, a 15 fold increase of value compared to an XF IQ350 (per Bernard)

[BernardLanguillier]

I am NOT pointing fingers just at Fuji.

Right... you still had the choice to post this or not... and you choose to post it.

I assume in a generous move aimed at informing people that they may be about to buy a 10,000 US$ camera with major issue?

Cheers,
Bernard

[Steve Hendrix]

My understanding is that the faulty cameras were dropping all of the custom settings that you entered into the firmware each time you turned it off. No small matter for a camera with a 300 page manual filled with different settings that you customize. Moreover, there are other glitches being reported by first users. https://www.dpreview.com/forums/thread/4405899#forum-post-62836667. I am NOT pointing fingers just at Fuji. As you know, I have an X1D, which had more than its fair share of teething issues. I am sure Fuji will get it all sorted out in time, but I think most of us expect more when we drop $10k on a camera. I assume you would agree, Steve.

We've not had these reports yet and were only aware of the date/time setting issue, stands to reason this would apply to custom settings as well. Still, I am not aware that anyone has confirmed this is not an issue of the internal battery needing a more thorough charge. If that is the case, and cameras need to be exchanged - which I assume we would eventually need to take part in, that's a different story. Once I see that this is at the level of "all sorts of issues", then I'll post the reconsideration of my comment.


Steve Hendrix/CI

[eronald]

I expect B to present us a picture of a sailboat in action real soon! I mean, what better place to take a brand spanking new 15k jewel of envy to than over the middle of a deep trench in the ocean somewhere. You gotta be livin' on the edge, no?

Maybe his memory reset?

Edmund

[BernardLanguillier]

I expect B to present us a picture of a sailboat in action real soon! I mean, what better place to take a brand spanking new 15k jewel of envy to than over the middle of a deep trench in the ocean somewhere. You gotta be livin' on the edge, no?

Not very likely. Horses for courses, the Nikon Z7 is a much better fit for those situations, would it only be because it has much more useful - and still excellent - zoom lenses starting with the amazing 24-70 f2.8 S.

And I tend to apply the one year rule... don't take a new camera to a potentially risky outing until it has delivered value for at least a year. That's one of the reason why I took the D850 instead of the Z7 during a one week sailing trip last November... but every rule has exceptions obviously. 

Cheers,
Bernard

[chez]

Not very likely. Horses for courses, the Nikon Z7 is a much better fit for those situations, would it only be because it has much more useful - and still excellent - zoom lenses starting with the amazing 24-70 f2.8 S.

And I tend to apply the one year rule... don't take a new camera to a potentially risky outing until it has delivered value for at least a year. That's one of the reason why I took the D850 instead of the Z7 during a one week sailing trip last November... but every rule has exceptions obviously. 

Cheers,
Bernard

You mean depreciate in value don't you? 

[BernardLanguillier]

You mean depreciate in value don't you? 

I look at cameras as items that deliver value through usage.

Cheers,
Bernard

[eronald]

I look at cameras as items that deliver value through usage.

Cheers,bout the camera to
Bernard

Bernard, could we see an image with urban texture, or a person in it, or even a cat?

Edmund

[ErikKaffehr]

Hi Ronald,

I posted a question on DPReviews Science and Technology forums on costs for a new sensor. The best answer I got was:

Erik, if the pixel has already been proven via other products or test devices, then this is a pretty easy task. A design team of perhaps 10-12 could bang the design out in less than 6 months.  Then there is fab and characterization which could be 9 months.  Depending on what is found, there may need to be a revision.  Entering manufacturing might take another 6 months (production samples v. engineering samples).  Between personnel costs and mask costs, could easily be a few million $ to get to engineering samples, and a few more to get to production. So you are several million into NRE as an investment. Someone somewhere does this math more carefully, and looks at the market for price total production volume, and considers other strategic factors (e.g. does it enable a new Sony camera product) and someone else then decides if it is worth starting or not.  There are always way more product opportunities than there are design and fab capacities, so these decisions are not done in isolation.  For example, Sony might make more money making more stacked camera-phone sensors instead of using fab capacity for this new product.  It is a complicated business, and if only we had a perfect crystal ball for making decisions like this.

BTW,  speaking of Sony design capacities, looks like they need more capacity: article

That was from Eric Fossum, the inventor of the CMOS sensor.

I would also guess that NRE costs for the 44x33 sensors are paid off and that production may be moved to an older 'fab' while the 100 MP sensor is probably made in a fab supportingnew design features like BSI and copper.

In general, new products are introduced at higher prices in lower volumes. After a while prices go down and availability gets better.

Best regards
Erik

It is my belief - founded on my 40 year old education in IC design, that chips are priced by surface. (The root cause of this is that defects follow a Poisson statistic).

Hence the cost of a 50Mp or a 100MP 44x33 chip do not differ vey significantly.

Which makes me think that the cost to Fuji of a 44x33 chip is well under $1K, and makes the pricing of the GFX100 completely incomprehensible, a more reasonable expectation of pricing would be somewhere north of $5K where we find the other "pro" Nikon and Canon bodies ... and the Hasselblad X1DII.

I think Fuji may have been indulging in a bit of puffery with the overlarge and shiny booster body and noisy launch PR, and we can expect some quick steps downwards as soon as the first wave of impulse buying subsides.

The positioning of the GFX100 in the Fuji line, as a straight jump up from APS-C is very smart. However it becomes less clever when the price differential between APS-C flagship and the crop MF flagship is 4x rather than 2x.

I intend no disrespect to Fuji, nor to the GFX100 which seems an excellent photographic device, but the discussion of pricing has always been a prerogative of the consumer.

Edmund

[BernardLanguillier]

Bernard, could we see an image with urban texture, or a person in it, or even a cat?

My daughter has a bad cat allergy and I prefer smooth cities...

This being said I'll probably be in Singapore soon, I may bring the compact camera with me.

Cheers,
Bernard

[Dan Wells]

If we're looking at "several million into NRE as an investment" on a sensor that has (so far) appeared in one expensive camera, there's quite a bit of that investment stuck to each sensor. It's a guess, but I suspect Sony is attaching at least $1000 extra per sensor to cover that investment, the opportunity cost of making the sensor in a newer fab (the 50 MP is probably made on a line that has fewer other possibilities) and extra profit on a new product. It could easily be twice that...

Another guess: Fuji is charging the end user about twice what the sensor costs them for the sensor difference. Much of this is margin, but it also covers things like a unique filter layer.

This puts the total sensor (and filter stack) cost difference at retail somewhere between $2000 and $4000 from a GFX 50 to a GFX 100. This doesn't count the cost of the (limited-production) IBIS unit or anything else aside from "what's your best guess about what the sensor's responsible for".

This is based on a production of ~1000 cameras month (or possibly less). Those are D3x-ish numbers, and I can't think of a better comparable (for which the numbers are available - Nikon numbers are easier to get than anything else). The scraps of information that have come out on medium format suggest that the much less expensive GFX 50S and R probably sell somewhat better than the D3x ever did (especially in combination), but that nothing else has probably reached D3x numbers (the ONLY information I have on the Pentax is Fuji's statement that they "doubled the size of the medium format market" even before the GFX 50R). If the Pentax had been selling 1000+ per month plus various Phase and Hasselblad models, Fuji would have to have been selling quite a few cameras to make that true...

[eronald]

Re. Sony, I think that what Sony Semiconductor does these days is basically churn out sensors by customising their technologies, and they do this quickly and easily, at a price which ensures them increasing marketshare. The 100MP sensor used in the GFX100 has already been out for some time in remote sensing/mapping  applications, and I would expect that the design costs are amortised, and Sony won't get too greedy because they will still hope to resell the same sensor to that chinese drone company which keeps a luxury camera as a side-brand, and also see a compact GF100R as another future release.  Fuji's investment in oversized IBIS and doubly oversized shutter is a different story.

Edmund

If we're looking at "several million into NRE as an investment" on a sensor that has (so far) appeared in one expensive camera, there's quite a bit of that investment stuck to each sensor. It's a guess, but I suspect Sony is attaching at least $1000 extra per sensor to cover that investment, the opportunity cost of making the sensor in a newer fab (the 50 MP is probably made on a line that has fewer other possibilities) and extra profit on a new product. It could easily be twice that...

Another guess: Fuji is charging the end user about twice what the sensor costs them for the sensor difference. Much of this is margin, but it also covers things like a unique filter layer.

This puts the total sensor (and filter stack) cost difference at retail somewhere between $2000 and $4000 from a GFX 50 to a GFX 100. This doesn't count the cost of the (limited-production) IBIS unit or anything else aside from "what's your best guess about what the sensor's responsible for".

This is based on a production of ~1000 cameras month (or possibly less). Those are D3x-ish numbers, and I can't think of a better comparable (for which the numbers are available - Nikon numbers are easier to get than anything else). The scraps of information that have come out on medium format suggest that the much less expensive GFX 50S and R probably sell somewhat better than the D3x ever did (especially in combination), but that nothing else has probably reached D3x numbers (the ONLY information I have on the Pentax is Fuji's statement that they "doubled the size of the medium format market" even before the GFX 50R). If the Pentax had been selling 1000+ per month plus various Phase and Hasselblad models, Fuji would have to have been selling quite a few cameras to make that true...

[Dan Wells]

I wonder what the numbers of sensors used in remote sensing/mapping applications are? I have no idea - is what we call photography a minor sideshow in the use of high-resolution sensors, or is the aerial market a small percentage of what gets used in cameras, or is it roughly equal?

[ErikKaffehr]

Hi Dan,

A few years ago, it was often mentioned that the MFD market may be 10000 cameras a year or less.

A few months ago, Lens Rentals published it's "Most Popular Photography and Videography Products of 2018", that places Fujifilm on 4.67% 2017 and 4.25% 2018. So, it seems that the GF system may not have a large impact on Fujifilm rentals.

Would Fujifilm sell 20000 GFX systems a year and and would say 5000 of that be GFX100, we may assume, just as an example:

NRE costs, say 20M$US (Eric Fossum indicated like 10M$US on a sensor based on an existing pixel design, but this is a new design)
Say that pay off is calculated for three years. So, that would mean 2e7 / 15000 -> 1300 $US, just covering NRE costs. To that comes production costs.

Pricing is actually about what customers are willing to pay, more than about costs. Any manufacturer selling below fully covered manufacturing costs will go kerplunk over time of course.

Best regards
Erik

If we're looking at "several million into NRE as an investment" on a sensor that has (so far) appeared in one expensive camera, there's quite a bit of that investment stuck to each sensor. It's a guess, but I suspect Sony is attaching at least $1000 extra per sensor to cover that investment, the opportunity cost of making the sensor in a newer fab (the 50 MP is probably made on a line that has fewer other possibilities) and extra profit on a new product. It could easily be twice that...

Another guess: Fuji is charging the end user about twice what the sensor costs them for the sensor difference. Much of this is margin, but it also covers things like a unique filter layer.

This puts the total sensor (and filter stack) cost difference at retail somewhere between $2000 and $4000 from a GFX 50 to a GFX 100. This doesn't count the cost of the (limited-production) IBIS unit or anything else aside from "what's your best guess about what the sensor's responsible for".

This is based on a production of ~1000 cameras month (or possibly less). Those are D3x-ish numbers, and I can't think of a better comparable (for which the numbers are available - Nikon numbers are easier to get than anything else). The scraps of information that have come out on medium format suggest that the much less expensive GFX 50S and R probably sell somewhat better than the D3x ever did (especially in combination), but that nothing else has probably reached D3x numbers (the ONLY information I have on the Pentax is Fuji's statement that they "doubled the size of the medium format market" even before the GFX 50R). If the Pentax had been selling 1000+ per month plus various Phase and Hasselblad models, Fuji would have to have been selling quite a few cameras to make that true...

[Dan Wells]

Hi Erik-
     Sounds about right, although Fuji probably needs to sell somewhat more than 20,000 cameras/year to support all that lens development (and three bodies). How much more? I don't know... 1/4 GFX 100 is kind of what I'm thinking, too...

     Does this count as an existing or a new pixel design? It is a new design, but shared among three sensors - the 33x44mm, the presumably very low-volume (unless there's some use that isn't photographic in our sense) IQ4 150 sensor and (importantly, because it could be a lot of volume over time), the X-T3 sensor, which will probably migrate into Sonys, Nikons and Pentaxes as well. There's a datasheet for a fourth sensor with this pixel, which we haven't seen in a camera yet - 24x36mm, ~65 MP. How much is medium format paying for the development of the pixel?

Dan

[eronald]

Hi Erik-
     Sounds about right, although Fuji probably needs to sell somewhat more than 20,000 cameras/year to support all that lens development (and three bodies). How much more? I don't know... 1/4 GFX 100 is kind of what I'm thinking, too...

     Does this count as an existing or a new pixel design? It is a new design, but shared among three sensors - the 33x44mm, the presumably very low-volume (unless there's some use that isn't photographic in our sense) IQ4 150 sensor and (importantly, because it could be a lot of volume over time), the X-T3 sensor, which will probably migrate into Sonys, Nikons and Pentaxes as well. There's a datasheet for a fourth sensor with this pixel, which we haven't seen in a camera yet - 24x36mm, ~65 MP. How much is medium format paying for the development of the pixel?

Dan

Not only is it not a new pixel design, but it's part of a whole family of chips with the same interfaces. In my view, it's a trivial scaling down of the 150MP chip. If I were feeling unpleasant, I would say that Sony were probably able to finalize the design for the 100MP in one day after doing the 150, by changing just a few parameters in a CAD macro file.

Edmund

[Christopher]

And still no 36mm FF camera using that design.

[mcbroomf]

Not only is it not a new pixel design, but it's part of a whole family of chips with the same interfaces. In my view, it's a trivial scaling down of the 150MP chip. If I were feeling unpleasant, I would say that Sony were probably able to finalize the design for the 100MP in one day after doing the 150, by changing just a few parameters in a CAD macro file.

Edmund

Start thinking about making a donation then Edmund .. you're not even close    each new chip needs to be designed from the ground up.  Of course as previously discussed libraries can be used, and those familiar with them will get the work done fast, but I'd expect it to take weeks at least.

[eronald]

Start thinking about making a donation then Edmund .. you're not even close    each new chip needs to be designed from the ground up.  Of course as previously discussed libraries can be used, and those familiar with them will get the work done fast, but I'd expect it to take weeks at least.

I disagree with the "from the ground up" assertion if you already have a large array and want to scale down, and you're in a company set up to cycle through custom sensor designs quickly, but weeks seems correct. I see no reason why such a design would have a high setup cost and indeed a photographer here once recounted that Dalsa made a custom "digital polaroid" 8x10 sensor for him for about $100K.

Edmund

[mcbroomf]

I'm just telling you design just doesn't happen like that.  A new sensor with a new cell would take a lot longer than one using a previously used and characterized cell, but much of that time would be saved in avoiding repeated steppings after debug, and likely a shorter QA test requirement.

Anyway, I think this is done to death ... 

[eronald]

I'm sorry - there may have been a misunderstanding here - by "scale down" I meant reduce the number of pixels in x or y or dimensions or both. Obviously, if you scale down the pixel for a new process, you need new characterisation, but in my opinion just leaving out some pixels from an existing and fabricated array - eg. the 150MP Sony chip won't create major new issues. If you think it does, go ahead and educate me.

Edmund

I'm just telling you design just doesn't happen like that.  A new sensor with a new cell would take a lot longer than one using a previously used and characterized cell, but much of that time would be saved in avoiding repeated steppings after debug, and likely a shorter QA test requirement.

Anyway, I think this is done to death ...