Archive 2020 · EOS-R5....will it live up to the hype?
p.35 #1EOS-R5....will it live up to the hype?
EverLearning wrote:
Again, the reason for higher DR is not that you can "see" or print the wide DR that the camera can capture. It is so that you an capture image data that you can flatten in post (bright up shadow lightness, drop highlight brightness) in order to fit in a range that we can reproduce and view.We can debate your first point (the president of IBM decades ago said “I think there is a world market for about five computers.” Many said that man would never go to the moon. The "impossible" has been made possible again and again with time and determination), but it is the second point, the one in your quoted paragraph, that says it all. There absolutely is value to higher DR in a camera. Just think of the DR on digital cameras 10 years ago and the tools we had to manipulate DR and compare it to now.
Perhaps your shooting style...Show more →
I agree. We can't foresee what in-camera and post process computational engines will be able to do with a more encompassing data-set (image) in the future. The more granular we can digitally record the analog world, the better!
gdanmitchell wrote
EverLearning wrote
Lastly, even if print and screens today can't show all of the DR, those of tomorrow will. So again, I would rather have more than less.Perhaps your shooting style...Show more →
I agree. We can't foresee what in-camera and post process computational engines will be able to do with a more encompassing data-set (image) in the future. The more granular we can digitally record the analog world, the better!
Apr 23, 2020 at 09:19 AM
p.35 #2EOS-R5....will it live up to the hype?
I will say that I can record raw video on both the 64 and 1258 even though it shows as not supported. The largest card I have is 128, so I know it works. I am headed off tomorrow to try some video. I will see how much each actually does.
Apr 23, 2020 at 09:19 AM
p.35 #3EOS-R5....will it live up to the hype?
Nikon shooter and heavily considering the EOS-R5 as of lately I'm shifting toward more videos. Not a big fan of beta testing, so I'll wait a few months after release.
Apr 23, 2020 at 10:00 AM
p.35 #4EOS-R5....will it live up to the hype?
EverLearning wrote:
i have seen quite a number of posts where you repeat your three scenario treatise on why improved DR is of no practical value, so you have certainly been emphatic and consistent about it.
.I didn't see where Dan said that ...
i have seen quite a number of posts where you repeat your three scenario treatise on why improved DR is of no practical value, so you have certainly been emphatic and consistent about it.
.I didn't see where Dan said that ...
Apr 23, 2020 at 10:16 AM
p.35 #5EOS-R5....will it live up to the hype?
EverLearning wrote:
The incremental costs of adding video capability to these still cameras is almost certainly very small by comparison to the base price of producing the units.
...
Video is just another feature that some users find attractive, but not all care about — think of features like face-detection, ultra-fast burst modes, some of the esoteric AF modes, the ability to work with electronic flash, etc.
...
The bottom line here is that I'm baffled that folks are so up in arms about the fact that their great still camera also has some additional features that don't interfere with their still photography, might potentially have some use, and which enlarge the market of the cameras, allowing the development and production costs to be spread over a larger number of units.
DanPerhaps the incremental costs of adding video are small compared to base price of producing the unit. Perhaps not. How much extra R&D went into designing the camera to offer top of class video rather than video like a 7D MII or a 5D IV? Were more expensive materials used within the camera than otherwise necessary to aid in heat dissipation, or tolerating the higher heat? What other considerations might there be that I, not an engineer, wouldn't even think about?
I did mention two considerations; the other being perceived value, which you actually allude to as well. If you think...Show more →
I am cautiously excited about the R5. I would absolutely love 120fps in 4k. Though I think when it comes out, I'll rent one and compare it to what Nvidia can do with a GPU generating slow motion from my Z6's footage.
But you mentioned heat. I didn't read this entire thread so I apologize if I'm bringing up a topic that's been mentioned before, but I actually only saw this morning that there will be an HEVC option. I can't fathom how they will pull this off. My desktop computer, with a full custom water cooling loop, with a 360mm radiator, and a total of 6 120mm and 2 140mm fans in an air conditioned room, with a Ryzen 3950x (that's 16 cores and 32 threads) can, with an overclock, encode 4k at about 9fps, and the CPU touches 90C. I get about 8fps without the overclock, and the CPU hits about 75-80.
If I hand the encode over to the NVENC, I get about 50fps encoding, but that's a full desktop GPU being directly fed cool air, 3 120mm intake fans blow directly in to the thing, and it hits about 70C during an encode. Also, during an encode with that GPU, Davinici Resolve hits around 90GB of RAM usage.
I don't think that Canon has defied the laws of physics, so there is SOMETHING going on here. 8k footage to HEVC, if it can be done at these frame rates for a sustained amount of time, they must have really come up with something special. Though I'll remain skeptical until I see it in action.
Edited on Apr 23, 2020 at 11:47 AM · View previous versions
gdanmitchell wrote
EverLearning wrote
gdanmitchell, i agree size of market influences price, but if your means of increasing 'market' is increasing price by adding all kinds of extras, the subsequent reduction in price due to enhanced market reach doesn't offset that increase. If it did, everything - cars, TVs, furnaces, cameras, computers, etc - would just come fully spec'ed out. Such is not the case....
Video is just another feature that some users find attractive, but not all care about — think of features like face-detection, ultra-fast burst modes, some of the esoteric AF modes, the ability to work with electronic flash, etc.
...
The bottom line here is that I'm baffled that folks are so up in arms about the fact that their great still camera also has some additional features that don't interfere with their still photography, might potentially have some use, and which enlarge the market of the cameras, allowing the development and production costs to be spread over a larger number of units.
Dan
I did mention two considerations; the other being perceived value, which you actually allude to as well. If you think...Show more →
I am cautiously excited about the R5. I would absolutely love 120fps in 4k. Though I think when it comes out, I'll rent one and compare it to what Nvidia can do with a GPU generating slow motion from my Z6's footage.
But you mentioned heat. I didn't read this entire thread so I apologize if I'm bringing up a topic that's been mentioned before, but I actually only saw this morning that there will be an HEVC option. I can't fathom how they will pull this off. My desktop computer, with a full custom water cooling loop, with a 360mm radiator, and a total of 6 120mm and 2 140mm fans in an air conditioned room, with a Ryzen 3950x (that's 16 cores and 32 threads) can, with an overclock, encode 4k at about 9fps, and the CPU touches 90C. I get about 8fps without the overclock, and the CPU hits about 75-80.
If I hand the encode over to the NVENC, I get about 50fps encoding, but that's a full desktop GPU being directly fed cool air, 3 120mm intake fans blow directly in to the thing, and it hits about 70C during an encode. Also, during an encode with that GPU, Davinici Resolve hits around 90GB of RAM usage.
I don't think that Canon has defied the laws of physics, so there is SOMETHING going on here. 8k footage to HEVC, if it can be done at these frame rates for a sustained amount of time, they must have really come up with something special. Though I'll remain skeptical until I see it in action.
Edited on Apr 23, 2020 at 11:47 AM · View previous versions
Apr 23, 2020 at 11:43 AM
p.35 #6EOS-R5....will it live up to the hype?
Custom ASICs. A GPU is still general purpose. A FPGU would be better. An ASIC is best.
Apr 23, 2020 at 11:46 AM
p.35 #7EOS-R5....will it live up to the hype?
stanj wrote:
Custom ASICs. A GPU is still general purpose. A FPGU would be better. An ASIC is best.
Dude, what ASIC can make that enormous of a difference? I mean, yes, they are inherently specialized. But that isn't a tiny improvement over what a GPU can do, that's orders of magnitude better than what a GPU on a top of the line process node can do, which I can't see Canon being able to afford for this application. I can't find anything below about 30-40W TDP and they can't do 4k past 60fps, and they're the size of a full size PCIe card.
Custom ASICs. A GPU is still general purpose. A FPGU would be better. An ASIC is best.
Dude, what ASIC can make that enormous of a difference? I mean, yes, they are inherently specialized. But that isn't a tiny improvement over what a GPU can do, that's orders of magnitude better than what a GPU on a top of the line process node can do, which I can't see Canon being able to afford for this application. I can't find anything below about 30-40W TDP and they can't do 4k past 60fps, and they're the size of a full size PCIe card.
Apr 23, 2020 at 11:49 AM
p.35 #8EOS-R5....will it live up to the hype?
reggieb wrote:
Dude, what ASIC can make that enormous of a difference? I mean, yes, they are inherently specialized. But that isn't a tiny improvement over what a GPU can do, that's orders of magnitude better than what a GPU on a top of the line process node can do, which I can't see Canon being able to afford for this application. I can't find anything below about 30-40W TDP and they can't do 4k past 60fps, and they're the size of a full size PCIe card.
Well, dude, I don't know because I work in the software and not hardware realm, but at one of my prior camera companies we used FPGAs over GPUs because of their huge performance gain for the task at hand. But I'm not the hardware guy.
Either they're lying, which I find unlikely, or their chips (ASICs) are indeed better (for this task) than your GPU. I have a hunch that the image quality will be decent, too, or they'll be the laughing stock.
Dude, what ASIC can make that enormous of a difference? I mean, yes, they are inherently specialized. But that isn't a tiny improvement over what a GPU can do, that's orders of magnitude better than what a GPU on a top of the line process node can do, which I can't see Canon being able to afford for this application. I can't find anything below about 30-40W TDP and they can't do 4k past 60fps, and they're the size of a full size PCIe card.
Well, dude, I don't know because I work in the software and not hardware realm, but at one of my prior camera companies we used FPGAs over GPUs because of their huge performance gain for the task at hand. But I'm not the hardware guy.
Either they're lying, which I find unlikely, or their chips (ASICs) are indeed better (for this task) than your GPU. I have a hunch that the image quality will be decent, too, or they'll be the laughing stock.
Apr 23, 2020 at 12:20 PM
p.35 #9EOS-R5....will it live up to the hype?
reggieb wrote:
The incremental costs of adding video capability to these still cameras is almost certainly very small by comparison to the base price of producing the units.
...
Video is just another feature that some users find attractive, but not all care about — think of features like face-detection, ultra-fast burst modes, some of the esoteric AF modes, the ability to work with electronic flash, etc.
...
The bottom line here is that I'm baffled that folks are so up in arms about the fact that their great still camera also has some additional features that don't interfere with their still photography, might potentially have some use, and which enlarge the market of the cameras, allowing the development and production costs to be spread over a larger number of units.
DanPerhaps the incremental costs of adding video are small compared to base price of producing the unit. Perhaps not. How much extra R&D went into designing the camera to offer top of class video rather than video like a 7D MII or a 5D IV? Were more expensive materials used within the camera than otherwise necessary to aid in heat dissipation, or tolerating the higher heat? What other considerations might there be that I, not an engineer, wouldn't even think about?
I did mention two considerations; the other being perceived value, which you actually allude to as well. If you think back to the 80s, it didn't cost Sony anything to put 'Sony' on its products but it sure added to the cost. That was due to reputation and therefore perceived value. In this case, making this camera probably the best ML around AND the best non-cinamatic video device around is definitely going to cost the buyer more because you are getting 'more' so there is more perceived value.
You say that video is just another feature. True, but the other features you mention in the same breath are all related to stills, not video. A photographer who only shoots photos is very likely to use an assortment of those features but will not use the video feature.
Another point supporting this perspective is products that have been crippled in lower versions. So a product actually can do 10 things for example, but the manufacturer cripples seven things for the base version (via firmware or software) and cripples four things for the mid version. Obviously, production costs of all three versions of the product is exactly the same yet the three products sell for three very different prices.
So while I will readily agree that the video doesn't 'interfere' with my use of the camera as a tool for actual photography, I still firmly hold the believe that it adversely impacts price; possibly to the extent I will not be able to buy what is certainly looking to be an impressive tool for photography. To me, that would be very sad.I am cautiously excited about the R5. I would absolutely love 120fps in 4k. Though I think when it comes out, I'll rent one and compare it to what Nvidia can do with a GPU generating slow motion from my Z6's footage.
But you mentioned heat. I didn't read this entire thread so I apologize if I'm bringing up a topic that's been mentioned before, but I actually only saw this morning that there will be an HEVC option. I can't fathom how they will pull this off. My desktop computer, with a full custom water cooling loop, with a...Show more →
Because you aren't just moving raw data. The camera is doing the type of stuff you are doing when you just take photos from your memory card and put it on the hard drive, only without all the extra cables and connections. I don't see how anyone would think that there would be more than a few minutes of 8k record time at once. Does anyone think they're going to give 30 minutes a pop?.
EverLearning wrote
gdanmitchell wrote
EverLearning wrote
gdanmitchell, i agree size of market influences price, but if your means of increasing 'market' is increasing price by adding all kinds of extras, the subsequent reduction in price due to enhanced market reach doesn't offset that increase. If it did, everything - cars, TVs, furnaces, cameras, computers, etc - would just come fully spec'ed out. Such is not the case....
Video is just another feature that some users find attractive, but not all care about — think of features like face-detection, ultra-fast burst modes, some of the esoteric AF modes, the ability to work with electronic flash, etc.
...
The bottom line here is that I'm baffled that folks are so up in arms about the fact that their great still camera also has some additional features that don't interfere with their still photography, might potentially have some use, and which enlarge the market of the cameras, allowing the development and production costs to be spread over a larger number of units.
Dan
I did mention two considerations; the other being perceived value, which you actually allude to as well. If you think back to the 80s, it didn't cost Sony anything to put 'Sony' on its products but it sure added to the cost. That was due to reputation and therefore perceived value. In this case, making this camera probably the best ML around AND the best non-cinamatic video device around is definitely going to cost the buyer more because you are getting 'more' so there is more perceived value.
You say that video is just another feature. True, but the other features you mention in the same breath are all related to stills, not video. A photographer who only shoots photos is very likely to use an assortment of those features but will not use the video feature.
Another point supporting this perspective is products that have been crippled in lower versions. So a product actually can do 10 things for example, but the manufacturer cripples seven things for the base version (via firmware or software) and cripples four things for the mid version. Obviously, production costs of all three versions of the product is exactly the same yet the three products sell for three very different prices.
So while I will readily agree that the video doesn't 'interfere' with my use of the camera as a tool for actual photography, I still firmly hold the believe that it adversely impacts price; possibly to the extent I will not be able to buy what is certainly looking to be an impressive tool for photography. To me, that would be very sad.
But you mentioned heat. I didn't read this entire thread so I apologize if I'm bringing up a topic that's been mentioned before, but I actually only saw this morning that there will be an HEVC option. I can't fathom how they will pull this off. My desktop computer, with a full custom water cooling loop, with a...Show more →
Because you aren't just moving raw data. The camera is doing the type of stuff you are doing when you just take photos from your memory card and put it on the hard drive, only without all the extra cables and connections. I don't see how anyone would think that there would be more than a few minutes of 8k record time at once. Does anyone think they're going to give 30 minutes a pop?.
Apr 23, 2020 at 12:55 PM
p.35 #10EOS-R5....will it live up to the hype?
stanj wrote:
Well, dude, I don't know because I work in the software and not hardware realm, but at one of my prior camera companies we used FPGAs over GPUs because of their huge performance gain for the task at hand. But I'm not the hardware guy.
Either they're lying, which I find unlikely, or their chips (ASICs) are indeed better (for this task) than your GPU. I have a hunch that the image quality will be decent, too, or they'll be the laughing stock.
Indeed, when a Samsung smartphone can record 8K video on the fly in H.265 at 24 fps, I think there is a decent chance that Canon can pull this off too. The quality would have to be higher of course. As a matter of fact, I wouldn’t be surprised if Samsung might be involved in the manufacture of these ASIC, they have a very advanced process tech. I think Samsung is also using Dual Pixel AF, so.... maybe som cooperation?
reggieb wrote
Dude, what ASIC can make that enormous of a difference? I mean, yes, they are inherently specialized. But that isn't a tiny improvement over what a GPU can do, that's orders of magnitude better than what a GPU on a top of the line process node can do, which I can't see Canon being able to afford for this application. I can't find anything below about 30-40W TDP and they can't do 4k past 60fps, and they're the size of a full size PCIe card.Either they're lying, which I find unlikely, or their chips (ASICs) are indeed better (for this task) than your GPU. I have a hunch that the image quality will be decent, too, or they'll be the laughing stock.
Indeed, when a Samsung smartphone can record 8K video on the fly in H.265 at 24 fps, I think there is a decent chance that Canon can pull this off too. The quality would have to be higher of course. As a matter of fact, I wouldn’t be surprised if Samsung might be involved in the manufacture of these ASIC, they have a very advanced process tech. I think Samsung is also using Dual Pixel AF, so.... maybe som cooperation?
Apr 23, 2020 at 12:58 PM
p.35 #11EOS-R5....will it live up to the hype?
As good as the R5 sounds for video, I can't wait to see what the specs are like for still photos.
And what additional features are built into the body?
I'd love to see focus stacking an integral feature like it was for the RP.
And what additional features are built into the body?
I'd love to see focus stacking an integral feature like it was for the RP.
Apr 23, 2020 at 01:06 PM
p.35 #12EOS-R5....will it live up to the hype?
Based on the 1dx 5.5k raw bit rate, which is ~2800 Mbps, or about 1 GB every 3 seconds. So about 6 min of record time for 128GB card or 24min for 512GB. Even if there is no software limit, one would run into storage limits.
Apr 23, 2020 at 01:55 PM
p.35 #13EOS-R5....will it live up to the hype?
I worked in DC for many years where the culture was rife with the use of the of what was called a peeing contest. I see where it has evolved to prime time tv. You guys are amateurs. I began flying in a WWII T-6 and wound up flying using glass cockpit displays. I used Bronicas and Canon film in a wedding business that I opened after I left DC. What I learned from my tenure in DC was to do the best I could with what I had in hand whether it be a camera or an airplane to produce professional results and to stay out of fruitless and useless attempts at trying to prove my manhood. .
Apr 23, 2020 at 02:12 PM
p.35 #14EOS-R5....will it live up to the hype?
AJSJones wrote:
.I didn't see where Dan said that ...
We have a winner. You are correct. :-)
I always wonder about these folks who argue with something the other person didn't say. Arguing with phantoms must be fun, I guess.
EverLearning wrote
i have seen quite a number of posts where you repeat your three scenario treatise on why improved DR is of no practical value, so you have certainly been emphatic and consistent about it..
We have a winner. You are correct. :-)
I always wonder about these folks who argue with something the other person didn't say. Arguing with phantoms must be fun, I guess.
Apr 23, 2020 at 08:22 PM
p.35 #15EOS-R5....will it live up to the hype?
gdanmitchell wrote:
Print will never show it. There isn't enough range between paper white and print black.
ilkka_nissila wrote:
A large print can show it. As you spread the photo over a larger area, the dynamic range of the original per unit area is reduced.
Um, you aren't showing the full dynamic range when "the dynamic range of the original per unit area is reduced."
Like I said.
Plus, you can't actually see the full actual scene of subject with the DR you are thinking of in the way a camera can. In order to do so you visual system actually adjusts to the light of what it is looking at
[— this is why you can see details in shadow and then look at a bright snow field and see detail there. But you can't actually do both at once.
The brain nevertheless perceives the scene details in each area even if this is based on scanning the scene and some kind of processing the data. When you're looking at e.g. brightly lit backround with your fovea, you see its details, and when you look at an animal in the shadows, the brain sees the object, and forms a memory of the scene parts as when they were looking at them directly (in their adjusted form). A good print accomplishes something similar and some regional adjustment of the tones is necessary to accomplish this. It's not "radical post-processing" but a basic part of print-making that has been practiced for 100 years. Now, if you start with a high-quality original, you can make these adjustments and the final print will appear higher quality than if you started with an original with lower signal-to-noise overall and in the shadows in particular.
You are really missing the point here. In order to do what the eye does with a high dynamic range scene you must do precisely what I said you'll do — reduce the overall dynamic range of the original scene by making the darkest areas lighter and/or the lightest areas darker.
If you made a print with the actual dynamic range of the original hight dynamic range scene, it would not work because your eyes aren't going to adapt to the different areas of the print the way they do to the original scene.
Remember, you eyes cannot actually see the entire dynamic range of the scene at once when you are there looking at it. If you look into the dark shadows, your pupils dilate to send more light to your retina. You don't notice that the bright areas are getting brighter due to the way our visual system works. (You can notice this if you move your gaze quickly between very light and vary dark areas as it takes our pupils an observable amount of time to change. ) When you move your gaze to the bright areas your pupils close down a bit and at that moment you would no longer be able to see the same details in the dark areas... until you move your vision back to the shadows and your pupils dilate again.
When we do with a print (or over presentation) of such a scene is we lighten the dark areas and/or darken the bright areas so that the entire image, now with a compressed dynamic range, fits within our visual dynamic range without having to rely on the active changes in pupil dilation.
Apr 23, 2020 at 08:32 PM
p.35 #16EOS-R5....will it live up to the hype?
My prediction (and desire for a bit) has been that prints will be displaced for a number of use cases by Screens. Oleds and other technologies as they get more cheaper are a very viable replacement for prints. My main room tv is now a few years old, 1080p 65'' and just having screen saver be photos has been a gift. Over time as budget allows, I expect to replace all prints with screens.
stanj wrote:
This was a good and thoughtful post, but it highlights the need of what I would (sadly) describe as a dying breed: large prints. I mean I loved it when Calypso was still in business (in my home town no less), when I could print 80" wide. I have a 60x40 RA-4 print hanging over my home office desk, made with a 1DS no less. But I have exactly two of such prints in my home, because I don't have that much wall.
What I do have however is three laptops, and one iMac Pro with an attached 6K XDR display....Show more →
stanj wrote:
gdanmitchell wrote
- we were looking at crops out of larger images that had been sized to print at 30" x 40"What I do have however is three laptops, and one iMac Pro with an attached 6K XDR display....Show more →
Apr 23, 2020 at 08:59 PM
p.35 #17EOS-R5....will it live up to the hype?
mikeinctown wrote:
The incremental costs of adding video capability to these still cameras is almost certainly very small by comparison to the base price of producing the units.
...
Video is just another feature that some users find attractive, but not all care about — think of features like face-detection, ultra-fast burst modes, some of the esoteric AF modes, the ability to work with electronic flash, etc.
...
The bottom line here is that I'm baffled that folks are so up in arms about the fact that their great still camera also has some additional features that don't interfere with their still photography, might potentially have some use, and which enlarge the market of the cameras, allowing the development and production costs to be spread over a larger number of units.
DanPerhaps the incremental costs of adding video are small compared to base price of producing the unit. Perhaps not. How much extra R&D went into designing the camera to offer top of class video rather than video like a 7D MII or a 5D IV? Were more expensive materials used within the camera than otherwise necessary to aid in heat dissipation, or tolerating the higher heat? What other considerations might there be that I, not an engineer, wouldn't even think about?
I did mention two considerations; the other being perceived value, which you actually allude to as well. If you think back to the 80s, it didn't cost Sony anything to put 'Sony' on its products but it sure added to the cost. That was due to reputation and therefore perceived value. In this case, making this camera probably the best ML around AND the best non-cinamatic video device around is definitely going to cost the buyer more because you are getting 'more' so there is more perceived value.
You say that video is just another feature. True, but the other features you mention in the same breath are all related to stills, not video. A photographer who only shoots photos is very likely to use an assortment of those features but will not use the video feature.
Another point supporting this perspective is products that have been crippled in lower versions. So a product actually can do 10 things for example, but the manufacturer cripples seven things for the base version (via firmware or software) and cripples four things for the mid version. Obviously, production costs of all three versions of the product is exactly the same yet the three products sell for three very different prices.
So while I will readily agree that the video doesn't 'interfere' with my use of the camera as a tool for actual photography, I still firmly hold the believe that it adversely impacts price; possibly to the extent I will not be able to buy what is certainly looking to be an impressive tool for photography. To me, that would be very sad.I am cautiously excited about the R5. I would absolutely love 120fps in 4k. Though I think when it comes out, I'll rent one and compare it to what Nvidia can do with a GPU generating slow motion from my Z6's footage.
But you mentioned heat. I didn't read this entire thread so I apologize if I'm bringing up a topic that's been mentioned before, but I actually only saw this morning that there will be an HEVC option. I can't fathom how they will pull this off. My desktop computer, with a full custom water cooling loop, with a 360mm radiator, and a total of 6 120mm and 2 140mm fans in an air conditioned room, with a Ryzen 3950x (that's 16 cores and 32 threads) can, with an overclock, encode 4k at about 9fps, and the CPU touches 90C. I get about 8fps without the overclock, and the CPU hits about 75-80.
If I hand the encode over to the NVENC, I get about 50fps encoding, but that's a full desktop GPU being directly fed cool air, 3 120mm intake fans blow directly in to the thing, and it hits about 70C during an encode. Also, during an encode with that GPU, Davinici Resolve hits around 90GB of RAM usage.
I don't think that Canon has defied the laws of physics, so there is SOMETHING going on here. 8k footage to HEVC, if it can be done at these frame rates for a sustained amount of time, they must have really come up with something special. Though I'll remain skeptical until I see it in action.Because you aren't just moving raw data. The camera is doing the type of stuff you are doing when you just take photos from your memory card and put it on the hard drive, only without all the extra cables and connections. I don't see how anyone would think that there would be more than a few minutes of 8k record time at once. Does anyone think they're going to give 30 minutes a pop?.
Waving hand over here...... yes it is possible.
It happens probably ever day.... well before the covid19 thing.
also not 8K but at least 6K.
Blackmagic pocket CC 6K will out put to a 1 TB or 2 TB SSD. granted these are fast (580 mb/s) as to not drop frame rates.
reggieb wrote
EverLearning wrote
gdanmitchell wrote
EverLearning wrote
gdanmitchell, i agree size of market influences price, but if your means of increasing 'market' is increasing price by adding all kinds of extras, the subsequent reduction in price due to enhanced market reach doesn't offset that increase. If it did, everything - cars, TVs, furnaces, cameras, computers, etc - would just come fully spec'ed out. Such is not the case....
Video is just another feature that some users find attractive, but not all care about — think of features like face-detection, ultra-fast burst modes, some of the esoteric AF modes, the ability to work with electronic flash, etc.
...
The bottom line here is that I'm baffled that folks are so up in arms about the fact that their great still camera also has some additional features that don't interfere with their still photography, might potentially have some use, and which enlarge the market of the cameras, allowing the development and production costs to be spread over a larger number of units.
Dan
I did mention two considerations; the other being perceived value, which you actually allude to as well. If you think back to the 80s, it didn't cost Sony anything to put 'Sony' on its products but it sure added to the cost. That was due to reputation and therefore perceived value. In this case, making this camera probably the best ML around AND the best non-cinamatic video device around is definitely going to cost the buyer more because you are getting 'more' so there is more perceived value.
You say that video is just another feature. True, but the other features you mention in the same breath are all related to stills, not video. A photographer who only shoots photos is very likely to use an assortment of those features but will not use the video feature.
Another point supporting this perspective is products that have been crippled in lower versions. So a product actually can do 10 things for example, but the manufacturer cripples seven things for the base version (via firmware or software) and cripples four things for the mid version. Obviously, production costs of all three versions of the product is exactly the same yet the three products sell for three very different prices.
So while I will readily agree that the video doesn't 'interfere' with my use of the camera as a tool for actual photography, I still firmly hold the believe that it adversely impacts price; possibly to the extent I will not be able to buy what is certainly looking to be an impressive tool for photography. To me, that would be very sad.
But you mentioned heat. I didn't read this entire thread so I apologize if I'm bringing up a topic that's been mentioned before, but I actually only saw this morning that there will be an HEVC option. I can't fathom how they will pull this off. My desktop computer, with a full custom water cooling loop, with a 360mm radiator, and a total of 6 120mm and 2 140mm fans in an air conditioned room, with a Ryzen 3950x (that's 16 cores and 32 threads) can, with an overclock, encode 4k at about 9fps, and the CPU touches 90C. I get about 8fps without the overclock, and the CPU hits about 75-80.
If I hand the encode over to the NVENC, I get about 50fps encoding, but that's a full desktop GPU being directly fed cool air, 3 120mm intake fans blow directly in to the thing, and it hits about 70C during an encode. Also, during an encode with that GPU, Davinici Resolve hits around 90GB of RAM usage.
I don't think that Canon has defied the laws of physics, so there is SOMETHING going on here. 8k footage to HEVC, if it can be done at these frame rates for a sustained amount of time, they must have really come up with something special. Though I'll remain skeptical until I see it in action.
Waving hand over here...... yes it is possible.
It happens probably ever day.... well before the covid19 thing.
also not 8K but at least 6K.
Blackmagic pocket CC 6K will out put to a 1 TB or 2 TB SSD. granted these are fast (580 mb/s) as to not drop frame rates.
Apr 23, 2020 at 09:02 PM
p.35 #18EOS-R5....will it live up to the hype?
reggieb wrote:
Dude, what ASIC can make that enormous of a difference? I mean, yes, they are inherently specialized. But that isn't a tiny improvement over what a GPU can do, that's orders of magnitude better than what a GPU on a top of the line process node can do, which I can't see Canon being able to afford for this application. I can't find anything below about 30-40W TDP and they can't do 4k past 60fps, and they're the size of a full size PCIe card.
I think you are talking about 2 different applications.
one is moving data from a cmos or type of sensor for light gathering, processing the analog signal into digital information followed by saving to a digital media.
This is being done on a closed loop, highly controlled electrical path.
The other is taking digital information, manipulating (editing) + adding special effects and then transcoding to another format.
These are apple to oranges comparisons.
Also... why do you think company that produce pro/cinema rendered footage have large server farms ?
Edited on Apr 24, 2020 at 09:35 AM · View previous versions
stanj wrote
Custom ASICs. A GPU is still general purpose. A FPGU would be better. An ASIC is best.I think you are talking about 2 different applications.
one is moving data from a cmos or type of sensor for light gathering, processing the analog signal into digital information followed by saving to a digital media.
This is being done on a closed loop, highly controlled electrical path.
The other is taking digital information, manipulating (editing) + adding special effects and then transcoding to another format.
These are apple to oranges comparisons.
Also... why do you think company that produce pro/cinema rendered footage have large server farms ?
Edited on Apr 24, 2020 at 09:35 AM · View previous versions
Apr 23, 2020 at 09:10 PM
p.35 #19EOS-R5....will it live up to the hype?
Ziffl3 wrote:
Dude, what ASIC can make that enormous of a difference? I mean, yes, they are inherently specialized. But that isn't a tiny improvement over what a GPU can do, that's orders of magnitude better than what a GPU on a top of the line process node can do, which I can't see Canon being able to afford for this application. I can't find anything below about 30-40W TDP and they can't do 4k past 60fps, and they're the size of a full size PCIe card.I think you are talking about 2 different applications.
one is moving data from a cmos or type of sensor for light gathering, processing the analog signal into digital information followed by saving to a digital media.
This is being done on a closed loop, highly controlled electrical path.
The other is taking digital information, manipulating (editing) + adding special effects and then transcoding to another format.
These are apple to oranges comparisons.
Also... why do you thing company that produce pro/cinema rendered footage have large server farms ?
It's still HEVC encoding, I'm not talking about adding effects, of course something that intensive slows the encoding, because that's not all it's doing. HEVC compression is a very compute heavy task.
Stoffer wrote:
Well, dude, I don't know because I work in the software and not hardware realm, but at one of my prior camera companies we used FPGAs over GPUs because of their huge performance gain for the task at hand. But I'm not the hardware guy.
Either they're lying, which I find unlikely, or their chips (ASICs) are indeed better (for this task) than your GPU. I have a hunch that the image quality will be decent, too, or they'll be the laughing stock.Indeed, when a Samsung smartphone can record 8K video on the fly in H.265 at 24 fps, I think there is a decent chance that Canon can pull this off too. The quality would have to be higher of course. As a matter of fact, I wouldn’t be surprised if Samsung might be involved in the manufacture of these ASIC, they have a very advanced process tech. I think Samsung is also using Dual Pixel AF, so.... maybe som cooperation?
That's a possibility. And yeah, Samsung has about the best process node out there. I don't think Canon would have the volume to put in an order with someone like TSMC, but if they have a partnership with Samsung, that could be the ticket.
mikeinctown wrote:
The incremental costs of adding video capability to these still cameras is almost certainly very small by comparison to the base price of producing the units.
...
Video is just another feature that some users find attractive, but not all care about — think of features like face-detection, ultra-fast burst modes, some of the esoteric AF modes, the ability to work with electronic flash, etc.
...
The bottom line here is that I'm baffled that folks are so up in arms about the fact that their great still camera also has some additional features that don't interfere with their still photography, might potentially have some use, and which enlarge the market of the cameras, allowing the development and production costs to be spread over a larger number of units.
DanPerhaps the incremental costs of adding video are small compared to base price of producing the unit. Perhaps not. How much extra R&D went into designing the camera to offer top of class video rather than video like a 7D MII or a 5D IV? Were more expensive materials used within the camera than otherwise necessary to aid in heat dissipation, or tolerating the higher heat? What other considerations might there be that I, not an engineer, wouldn't even think about?
I did mention two considerations; the other being perceived value, which you actually allude to as well. If you think back to the 80s, it didn't cost Sony anything to put 'Sony' on its products but it sure added to the cost. That was due to reputation and therefore perceived value. In this case, making this camera probably the best ML around AND the best non-cinamatic video device around is definitely going to cost the buyer more because you are getting 'more' so there is more perceived value.
You say that video is just another feature. True, but the other features you mention in the same breath are all related to stills, not video. A photographer who only shoots photos is very likely to use an assortment of those features but will not use the video feature.
Another point supporting this perspective is products that have been crippled in lower versions. So a product actually can do 10 things for example, but the manufacturer cripples seven things for the base version (via firmware or software) and cripples four things for the mid version. Obviously, production costs of all three versions of the product is exactly the same yet the three products sell for three very different prices.
So while I will readily agree that the video doesn't 'interfere' with my use of the camera as a tool for actual photography, I still firmly hold the believe that it adversely impacts price; possibly to the extent I will not be able to buy what is certainly looking to be an impressive tool for photography. To me, that would be very sad.I am cautiously excited about the R5. I would absolutely love 120fps in 4k. Though I think when it comes out, I'll rent one and compare it to what Nvidia can do with a GPU generating slow motion from my Z6's footage.
But you mentioned heat. I didn't read this entire thread so I apologize if I'm bringing up a topic that's been mentioned before, but I actually only saw this morning that there will be an HEVC option. I can't fathom how they will pull this off. My desktop computer, with a full custom water cooling loop, with a 360mm radiator, and a total of 6 120mm and 2 140mm fans in an air conditioned room, with a Ryzen 3950x (that's 16 cores and 32 threads) can, with an overclock, encode 4k at about 9fps, and the CPU touches 90C. I get about 8fps without the overclock, and the CPU hits about 75-80.
If I hand the encode over to the NVENC, I get about 50fps encoding, but that's a full desktop GPU being directly fed cool air, 3 120mm intake fans blow directly in to the thing, and it hits about 70C during an encode. Also, during an encode with that GPU, Davinici Resolve hits around 90GB of RAM usage.
I don't think that Canon has defied the laws of physics, so there is SOMETHING going on here. 8k footage to HEVC, if it can be done at these frame rates for a sustained amount of time, they must have really come up with something special. Though I'll remain skeptical until I see it in action.Because you aren't just moving raw data. The camera is doing the type of stuff you are doing when you just take photos from your memory card and put it on the hard drive, only without all the extra cables and connections. I don't see how anyone would think that there would be more than a few minutes of 8k record time at once. Does anyone think they're going to give 30 minutes a pop?.
Well, no, it's not like copying files at all. But anyway, if the only limitation is that record time is a few minutes at once, that's fine with me.
reggieb wrote
stanj wrote
Custom ASICs. A GPU is still general purpose. A FPGU would be better. An ASIC is best.one is moving data from a cmos or type of sensor for light gathering, processing the analog signal into digital information followed by saving to a digital media.
This is being done on a closed loop, highly controlled electrical path.
The other is taking digital information, manipulating (editing) + adding special effects and then transcoding to another format.
These are apple to oranges comparisons.
Also... why do you thing company that produce pro/cinema rendered footage have large server farms ?
It's still HEVC encoding, I'm not talking about adding effects, of course something that intensive slows the encoding, because that's not all it's doing. HEVC compression is a very compute heavy task.
Stoffer wrote:
stanj wrote
reggieb wrote
Dude, what ASIC can make that enormous of a difference? I mean, yes, they are inherently specialized. But that isn't a tiny improvement over what a GPU can do, that's orders of magnitude better than what a GPU on a top of the line process node can do, which I can't see Canon being able to afford for this application. I can't find anything below about 30-40W TDP and they can't do 4k past 60fps, and they're the size of a full size PCIe card.Either they're lying, which I find unlikely, or their chips (ASICs) are indeed better (for this task) than your GPU. I have a hunch that the image quality will be decent, too, or they'll be the laughing stock.
That's a possibility. And yeah, Samsung has about the best process node out there. I don't think Canon would have the volume to put in an order with someone like TSMC, but if they have a partnership with Samsung, that could be the ticket.
mikeinctown wrote:
reggieb wrote
EverLearning wrote
gdanmitchell wrote
EverLearning wrote
gdanmitchell, i agree size of market influences price, but if your means of increasing 'market' is increasing price by adding all kinds of extras, the subsequent reduction in price due to enhanced market reach doesn't offset that increase. If it did, everything - cars, TVs, furnaces, cameras, computers, etc - would just come fully spec'ed out. Such is not the case....
Video is just another feature that some users find attractive, but not all care about — think of features like face-detection, ultra-fast burst modes, some of the esoteric AF modes, the ability to work with electronic flash, etc.
...
The bottom line here is that I'm baffled that folks are so up in arms about the fact that their great still camera also has some additional features that don't interfere with their still photography, might potentially have some use, and which enlarge the market of the cameras, allowing the development and production costs to be spread over a larger number of units.
Dan
I did mention two considerations; the other being perceived value, which you actually allude to as well. If you think back to the 80s, it didn't cost Sony anything to put 'Sony' on its products but it sure added to the cost. That was due to reputation and therefore perceived value. In this case, making this camera probably the best ML around AND the best non-cinamatic video device around is definitely going to cost the buyer more because you are getting 'more' so there is more perceived value.
You say that video is just another feature. True, but the other features you mention in the same breath are all related to stills, not video. A photographer who only shoots photos is very likely to use an assortment of those features but will not use the video feature.
Another point supporting this perspective is products that have been crippled in lower versions. So a product actually can do 10 things for example, but the manufacturer cripples seven things for the base version (via firmware or software) and cripples four things for the mid version. Obviously, production costs of all three versions of the product is exactly the same yet the three products sell for three very different prices.
So while I will readily agree that the video doesn't 'interfere' with my use of the camera as a tool for actual photography, I still firmly hold the believe that it adversely impacts price; possibly to the extent I will not be able to buy what is certainly looking to be an impressive tool for photography. To me, that would be very sad.
But you mentioned heat. I didn't read this entire thread so I apologize if I'm bringing up a topic that's been mentioned before, but I actually only saw this morning that there will be an HEVC option. I can't fathom how they will pull this off. My desktop computer, with a full custom water cooling loop, with a 360mm radiator, and a total of 6 120mm and 2 140mm fans in an air conditioned room, with a Ryzen 3950x (that's 16 cores and 32 threads) can, with an overclock, encode 4k at about 9fps, and the CPU touches 90C. I get about 8fps without the overclock, and the CPU hits about 75-80.
If I hand the encode over to the NVENC, I get about 50fps encoding, but that's a full desktop GPU being directly fed cool air, 3 120mm intake fans blow directly in to the thing, and it hits about 70C during an encode. Also, during an encode with that GPU, Davinici Resolve hits around 90GB of RAM usage.
I don't think that Canon has defied the laws of physics, so there is SOMETHING going on here. 8k footage to HEVC, if it can be done at these frame rates for a sustained amount of time, they must have really come up with something special. Though I'll remain skeptical until I see it in action.
Well, no, it's not like copying files at all. But anyway, if the only limitation is that record time is a few minutes at once, that's fine with me.
Apr 23, 2020 at 09:57 PM
p.35 #20EOS-R5....will it live up to the hype?
If you're short on time you might want to just start listening in at about the 4:30 mark. 
Well except if you're from Canada.
$3499 baby!
Well except if you're from Canada.
$3499 baby!
Apr 24, 2020 at 09:53 AM