BM 8K 60FPS camera is coming...

[L. Langer]

I don't think 8K will be much of an issue with the next gen hardware coming out in the next two years, especially with stacked memory and other advances in memory and solid state storage being made. We're at the every 10-15 year generational leap in tech, where huge performance increases come with new hardware designs. The only area in which we are lacking in terms of advancement is archival storage options and that has always been behind the curve for one reason or another. If we get advances in superior lossless A/V compression tech like REDRAY, it helps balance things out on the storage side of things, but that area of industry R&D has stalled or at the very least, has been put on the back-burner.

 

[Elsie N]

I'm curious how far along they are with their R & D. On the surface it looks like a reaction to RED coming out with the 8k VV sensor. I'm thinking RED may have been working on theirs for some time and only announced it when Jarred had a working model.

Still being a young company (BMD) sometimes saying something and actually doing it may not be as easy as the engineers tell them it is. Ask RED about that in years past.

 

[Marc Wielage]

I don't think 8K will be much of an issue with the next gen hardware coming out in the next two years, especially with stacked memory and other advances in memory and solid state storage being made. We're at the every 10-15 year generational leap in tech, where huge performance increases come with new hardware designs.

I haven't seen nearly as gigantic a leap in drive speed. Even with fairly fast drives under (say) Thunderbolt 2, it still takes 4 hours to copy 1TB. I would bet if you had fast fiber channel drives, you might be able to cut this in half. But if you're shooting 15TB of material a day, that's still 30 hours just in backups. And how long to make dailies and editable copies? Certainly, SSD RAIDs might push the data through faster, but just how much money do you want to spend, given that a 2TB SSD RAID is $2500? If you fill up 7 of these a day, how many are you going to need for a 30 or 40-day production? What if you go six months, like ​Captain America?

Maybe a bigger question is, how fast is your computer going to be and how well can any software handle it? Michael Cioni, who is as much of an expert on post as anybody around here, is doing a demo at Cinegear showing how Pablo can do real-time 8K and playback on an 8K display. But even the Pablo was struggling a bit at times with handling 4GB a second. Still amazingly impressive and impossible on most comparable systems. But how practical will this be for the real world?

There are still $200 million movies coming out this year that were not made with an end-to-end 4K workflow. They can barely handle 2K for a massive VFX pipeline. 4K will happen within another year or two. But 8K? You're dreaming. It absolutely could be done, given enough time and money, but I think it's unrealistic to expect that it'll be practical for 97% of users here.

 

[Mike Nagel]

The data rates for 8K are fairly intense:

8K RAW (8192 x 4320), 24fps, 10-bit: 1.012 Gigabytes per second – 60.75 Gigabytes per minute – 3.645 Terabytes per hour
8K RAW (8192 x 4320), 24fps, 12-bit: 1.215 Gigabytes per second – 72.9 Gigabytes per minute – 4.374 Terabytes per hour
8K RAW (8192 x 4320), 24fps, 16-bit: 1.620 Gigabytes per second – 97.2 Gigabytes per minute – 5.832 Terabytes per hour

I know very few feature productions that could cope with shooting (say) 15TB of material every day, assuming 4-5 hours of material. Multiply that times 40 days of shooting, you've got 600TB. Plus backups. Plus dailies. Plus more if you do A&B camera. And it will create an enormous speed and storage bottleneck for most editors, post houses, VFX houses, and colorists. And to my knowledge, there's currently no easy way to actually watch 8K video.

I think 8K could be useful for very, very high-budgeted Hollywood projects with significant resources, but not for many others. You can draw many apt comparisons to the companies that shot 65mm film in the 1980s and 1990s -- not a lot of indies, shorts, or documentary people doing that, as I recall.

while these numbers are close to accurate (it's 1061.68 MB per sec for 8K 10-bit 24fps), they don't include compression, so (fortunately) these numbers do not apply to R3D files, which will need way less space...

8K 16-bit 24 fps w/ 6:1 compression needs "just" 283.12 MB per sec...

but for all productions following a strict DPX/EXR workkflow, yeah they will need a couple of new hard drives... ;-))))

 

[M Most]

I haven't seen nearly as gigantic a leap in drive speed. Even with fairly fast drives under (say) Thunderbolt 2, it still takes 4 hours to copy 1TB. I would bet if you had fast fiber channel drives, you might be able to cut this in half. But if you're shooting 15TB of material a day, that's still 30 hours just in backups. And how long to make dailies and editable copies? Certainly, SSD RAIDs might push the data through faster, but just how much money do you want to spend, given that a 2TB SSD RAID is $2500? If you fill up 7 of these a day, how many are you going to need for a 30 or 40-day production? What if you go six months, like ​Captain America?
.

No large production I've ever been involved with archives to anything other than LTO tape, for a number of still valid reasons. And nobody uses single platter drives to move that volume of data (usually either a small SAS tower with multiple drives and much better throughput, or multiple SSD's). So the number of drives they need is only relevant to how quickly multiple LTO's can be turned around, and even then, drives usually aren't involved. More often than not, the camera cards are handed over to the dailies operator, and kept hot until the LTO's are done and verified. Then they're recycled. The only drive based backup might be a production backup and/or a large storage unit (often a SAN) being used by the dailies crew. And since you mention Captain America, I would point out that I understand they're shooting almost exclusively Arri 65 for at least a month. And they're using 3 of them. So if you add it up, they're probably producing about 10TB per day, possibly more. The way that you keep up with that is by having multiple everything (multiple LTO stations, multiple dailies systems, etc.) and splitting up the material. It can (and is) being done. But it's not for the inexperienced or the feint of heart.

And BTW, regardless of the budget, and regardless of the camera being used, on nearly every project we do the first thing we hear is that we have to get the LTO archives done "quicker" (yeah, right.......) because they're running out of cards. We usually hear that at the end of day 1.......

 

[Wayne Morellini]

I don't think 8K will be much of an issue with the next gen hardware coming out in the next two years, especially with stacked memory and other advances in memory and solid state storage being made. We're at the every 10-15 year generational leap in tech, where huge performance increases come with new hardware designs. The only area in which we are lacking in terms of advancement is archival storage options and that has always been behind the curve for one reason or another. If we get advances in superior lossless A/V compression tech like REDRAY, it helps balance things out on the storage side of things, but that area of industry R&D has stalled or at the very least, has been put on the back-burner.

Have they ever shown redray 9mb/s to be anything but visually lossless? I'm curious, Red always talk the talk about doing one thing and coming clean on the insides but O rber a few occasions when that dies not hsppem and they do something else and don't talk about details.

L, I dont know about 10-15 years, except as an average but you are right stacked memory and the resistive ram I meant earlier promise a great advance. There are advances in archival but I don't remember which at the moment. All I can remember recently Sony and Panasonic has released an archival UV disk, and they are aiming at 6 terabytes eventually, not much.

I'm curious how far along they are with their R & D. On the surface it looks like a reaction to RED coming out with the 8k VV sensor. I'm thinking RED may have been working on theirs for some time and only announced it when Jarred had a working model.

Still being a young company (BMD) sometimes saying something and actually doing it may not be as easy as the engineers tell them it is. Ask RED about that in years past.

From the talk of them pouring in millions into development of the sensor for them, I guess this is a long term strategy. 4.6k might be just the first.

I haven't seen nearly as gigantic a leap in drive speed. Even with fairly fast drives under (say) Thunderbolt 2, it still takes 4 hours to copy 1TB. I would bet if you had fast fiber channel drives, you might be able to cut this in half. But if you're shooting 15TB of material a day, that's still 30 hours just in backups. And how long to make dailies and editable copies? Certainly, SSD RAIDs might push the data through faster, but just how much money do you want to spend, given that a 2TB SSD RAID is $2500? If you fill up 7 of these a day, how many are you going to need for a 30 or 40-day production? What if you go six months, like ​Captain America?

Maybe a bigger question is, how fast is your computer going to be and how well can any software handle it? Michael Cioni, who is as much of an expert on post as anybody around here, is doing a demo at Cinegear showing how Pablo can do real-time 8K and playback on an 8K display. But even the Pablo was struggling a bit at times with handling 4GB a second. Still amazingly impressive and impossible on most comparable systems. But how practical will this be for the real world?

There are still $200 million movies coming out this year that were not made with an end-to-end 4K workflow. They can barely handle 2K for a massive VFX pipeline. 4K will happen within another year or two. But 8K? You're dreaming. It absolutely could be done, given enough time and money, but I think it's unrealistic to expect that it'll be practical for 97% of users here.

It's all relative Marc, on the codec used. The original redcode was a hog compared to conventional codecs. Now FX if they use less resolution FX in higher resolution films today, they can do the same with 8k. But again, it's relative to.the level of effects you use. Unless you use, say raytracing the FX processing power does not have to so directly relative to the number of pixels. I could probably get a commodore 64 to render vector based pong game on an 8k screen (a actual guess I haven't done the calculations yet) but to basic an effect to be too meaningful.

Now as L has mentioned, stacked memory for storage, but also for processing. As I mentioned resistive dram as an alternative to flash. Is cheap to make, low powered and as fast as normal dram compared to flash, and non-volatile. Once it is finally in we might have a real ternative to flash for high performance applications.

Processing is the big issue. Node process reduction is shortly topping put for power savings due to leakage. But the big issue is the monolithic design of PC processors. GPU processing is one area where we can squeeze more out, but also other processing arrays. So using a suitable lower overhead codec we should be able to do 8k next year as far as video goes on such equipment.. As for PC's better Arms ate coming. So there is some wiggle room there but stay sway from jpeg2000 based codecs if you don't have the right hardware.

 

[Rick Lang]

Just a rumor... from a reliable source.

Dunkan, can you post the context in which your quote was taken? Was it a 'comment' from an anonymous source or part of a detailed article with a known author who was protecting the source? Or is this just someone who likes attention and isn't above fabricating material for fun?

The new 4.6K sensor announced at NAB 2015 was at least two years in development and cost BMD at least $2M. I'm not making up those numbers as they were stated at NAB. The 16:9 aspect ratio sensor uses 5.5 micron photosites for a horizontal dimension of 25.344mm (4608 x 2592) and an angle of view of 1.42x Vistavision. It has a diagonal exceeding 29.07mm. There are at least seven recording formats including more than one that uses the full sensor width. The ProRes UHD and HD outputs are downscaled in camera from the full sensor if you select that option. ProRes 444 XQ 12bit is supported, and all flavours of 10bit ProRes. Raw can be mathematically lossless (about 1.27x compression of uncompressed raw) or visually lossless 3:1 or 4:1 (lossy but in motion pictures you are very unlikely to see the loss at 3:1 or 4:1). To save some storage, there is a 4608 x 1920 (2.4:1 aspect ratio) window for raw 'Scope. To support 2x anamorphic, there is a 3072 x 2560 (6:5 aspect ratio) raw window that desqueezes to 6144 x 2560 (2.4:1 aspect ratio). Just a brief summary of the factual strengths, (not even mentioning the dynamic range and global/rolling shutter) as relates to resolution.

BMD cameras record raw and ProRes in camera to dual CFast 2 cards. The fastest reliable cards that are ideal for recording BMD raw are the Lexar 3400x 256GB cards which write up to 450 MB/s on a single card or up to 900 MB/s in dual recording mode. And the fastest reliable cards for recording ProRes are the Wise 3400x 128GB which write up to 350 MB/s. There is no dual recording of ProRes of course. When you do the math, it's easy to see that you already can run into some brick walls trying to record the most demanding data rates on the CFast 2 cards. Newer versions of the cards may be able to approach the theoretical write speed of 750 MB/s, but who knows?

Why bore you with this stuff? Well I don't think the opening quote is credible for NAB 2016. 6k, yes, possibly. Why am I so myopic? I doubt BMD will move to smaller photosites than 5.5 microns in the near future as those wells are needed to collect light and give them the 15 stops of dynamic range. Their market is dominated by indie shooters although there is some movement to using their cameras at the next level. That market is going to shoot with lenses that are typically going to cover either Super 35 (24.89mm horizontally) or Vistavision (36mm). 6K can be kept within that 36mm limit with 5.5 micron photosites but 8K cannot. So the argument doesn't need to go into the need to record very high data rates which are already a challenge for the highest quality codecs. Their market limitations will put the maximum resolution at 6K for the next year or longer. And I suspect they may not even feel any need to increase the new 4.6K sensor in the short term. Rather than propogate the 4.6K sensor to greater resolutions, they may find ways to put it into smaller cameras (URSA Micro) or shrink it to smaller resolutions for other cameras. That would make post easier and help DITs retain some of their hair.

There is a way that the OP could have merit of course and that would require new sensor technology to pull off but I don't think BMD will have that rabbit to pull out of their hat that quickly. Nothing wrong with taking a baby step up to 6K first, because for their cinema cameras, high quality high dynamic range high frame rate 8K is a giant banana step. "You forgot to say, 'May I?'"

 

[Elsie N]

Dunkan, can you post the context in which your quote was taken? Was it a 'comment' from an anonymous source or part of a detailed article with a known author who was protecting the source? Or is this just someone who likes attention and isn't above fabricating material for fun?

The new 4.6K sensor announced at NAB 2015 was at least two years in development and cost BMD at least $2M. I'm not making up those numbers as they were stated at NAB. The 16:9 aspect ratio sensor uses 5.5 micron photosites for a horizontal dimension of 25.344mm (4608 x 2592) and an angle of view of 1.42x Vistavision. It has a diagonal exceeding 29.07mm. There are at least seven recording formats including more than one that uses the full sensor width. The ProRes UHD and HD outputs are downscaled in camera from the full sensor if you select that option. ProRes 444 XQ 12bit is supported, and all flavours of 10bit ProRes. Raw can be mathematically lossless (about 1.27x compression of uncompressed raw) or visually lossless 3:1 or 4:1 (lossy but in motion pictures you are very unlikely to see the loss at 3:1 or 4:1). To save some storage, there is a 4608 x 1920 (2.4:1 aspect ratio) window for raw 'Scope. To support 2x anamorphic, there is a 3072 x 2560 (6:5 aspect ratio) raw window that desqueezes to 6144 x 2560 (2.4:1 aspect ratio). Just a brief summary of the factual strengths, (not even mentioning the dynamic range and global/rolling shutter) as relates to resolution.

BMD cameras record raw and ProRes in camera to dual CFast 2 cards. The fastest reliable cards that are ideal for recording BMD raw are the Lexar 3400x 256GB cards which write up to 450 MB/s on a single card or up to 900 MB/s in dual recording mode. And the fastest reliable cards for recording ProRes are the Wise 3400x 128GB which write up to 350 MB/s. There is no dual recording of ProRes of course. When you do the math, it's easy to see that you already can run into some brick walls trying to record the most demanding data rates on the CFast 2 cards. Newer versions of the cards may be able to approach the theoretical write speed of 750 MB/s, but who knows?

Why bore you with this stuff? Well I don't think the opening quote is credible for NAB 2016. 6k, yes, possibly. Why am I so myopic? I doubt BMD will move to smaller photosites than 5.5 microns in the near future as those wells are needed to collect light and give them the 15 stops of dynamic range. Their market is dominated by indie shooters although there is some movement to using their cameras at the next level. That market is going to shoot with lenses that are typically going to cover either Super 35 (24.89mm horizontally) or Vistavision (36mm). 6K can be kept within that 36mm limit with 5.5 micron photosites but 8K cannot. So the argument doesn't need to go into the need to record very high data rates which are already a challenge for the highest quality codecs. Their market limitations will put the maximum resolution at 6K for the next year or longer. And I suspect they may not even feel any need to increase the new 4.6K sensor in the short term. Rather than propogate the 4.6K sensor to greater resolutions, they may find ways to put it into smaller cameras (URSA Micro) or shrink it to smaller resolutions for other cameras. That would make post easier and help DITs retain some of their hair.

There is a way that the OP could have merit of course and that would require new sensor technology to pull off but I don't think BMD will have that rabbit to pull out of their hat that quickly. Nothing wrong with taking a baby step up to 6K first, because for their cinema cameras, high quality high dynamic range high frame rate 8K is a giant banana step. "You forgot to say, 'May I?'"

A well thought out and rational post. Thanks.

 

[Mike Garrick]

Just remember 2020 Japan Toyko olympics will be shot & broadcast in 8K.

So it may be possible if you not looking at it thru a cinema prism.

 

[Gunleik Groven]

Like Phil, I don't doubt 8k is happening.

Still I have yet to have issues where 6k has been the limiting factor.

I would love to have the larger sensor, though

 

[Gunleik Groven]

After a quick search with your quote, it seems this thread is the source of the rumor...

https://www.google.no/webhp?sourcei...+version+of+their+4.6k+fairchild+sCMOS+sensor.

I guess then this thread has run its cause...