The thing I find a bit odd is that the pixels are pretty darn small. It seems Canon may have made major improvements in noise levels, but lost it by using such a small pixeled sensor. For general photography, isn't the pixel size (and how it relates to noise)more important then the crop factor?
This is a common misunderstanding. For general photography, the biggest factor is the total light gathered. There have been some extensive discussions about this topic, and some great writups, as it's called "Equivalence".
Assuming your framing a person for a portrait. If you have an APS-C and a FF camera at your disposal, and your goal is to make the same photo with both. The FF has several advantages (regardless of it's pixel size). Because the frame is larger, you have to get closer to your subject. That reduces DOF, increasing background blur (desirable for portraits specifically, but for many other things as well.) Once you have identical framing, the larger frame of the FF camera is gathering more total light than the APS-C camera.
On an equal output magnification basis, the FF camera, when downsampled to the same image dimensions as the APS-C image, is still better. The increased pixel count is averaged together, which increases SNR. If you have the same pixel size as a smaller sensor, it also gathered more spatial information, so that additional detail results in a sharper, crisper downsampled image than the natively sized APS-C image. On the other hand, you could shoot the same scene with a FF and APS-C camera, both of which have the same pixel count. The FF camera has larger pixels, so the per-pixel SNR is still higher. You don't gain in resolution (and you could well suffer the consequences of larger pixels...aliasing and moire)...but the image is still better. The larger sensor gathered more light. It doesn't really matter how big the pixels are...on a normalized basis, SNR is still going to be higher with the larger sensor.
Few photographers seem to understand this, and that's always confused me. But in the grand scheme of things, bigger sensor == better pictures when it comes to general photography. That is why medium format is still so desired by so many studio photographers. Despite the fact that for some time, after the advent of the Pentax K-5 and Nikon D7000 (the first couple of cameras to use Sony's then-new high DR Exmor sensors), medium format sensors did not have the same per-pixel IQ, the significantly larger sensor area (as well as more stringent manufacturing tolerances, to some degree) still produced better results.
On an equal-dimensions basis, there is very little to fear from smaller pixels. Assuming read noise is low (which, sadly, is probably NOT the case with the 5Ds, since it's Canon and Canon just seems to refuse to address their read noise issues), smaller pixels can never really be worse than larger pixels when the sensor is the same size or larger than the sensor of whatever camera you are comparing it to. Just about anything with a Sony Exmor sensor is going to have a flat read noise curve (average 3e- across all ISO settings, all sensor sizes, etc.) Comparing Exmor-based cameras is pretty easy, as read noise tends to be meaninglessly low relative to the image signal. A 5Ds should produce better quality than a 5D III or even a 1D X when the images are normalized (at low ISO, this may not be completely true, due to the high read noise). A 5Ds will certainly produce better quality than a 7D II on a normalized basis, even though the two have nearly identical pixel sizes. Because the larger frame of the 5Ds, with identical framing, will gather more total light.
The one situation where this is not true is when you are reach-limited. If you are photographing a bird or a deer, and you cannot get closer, you are going to use the same absolute area of the sensor for your subject. A bigger frame is just going to capture more space around your subject. In the reach-limited scenario, bigger pixels mean higher per-pixel SNR, but lower spatial resolution. Smaller pixels mean lower per-pixel SNR, but higher spatial resolution. For the most part, bird and wildlife photographers who are reach limited choose smaller pixels, despite their potential issues...because you can always downsample to the same dimenstions as the sensor with larger pixels, and most of those issues are averaged out and detail is still crisper.
-----
When it comes to astrophotography, there pretty much isn't anything more reach limited.
As such, or us, it's all about sampling rate and/or field of view. I think the 5DsR could be a great camera for short focal length, wide field imaging that would give us a much better image scale than existing DSLRs. I'd take 4.14µm pixels over 6.5µm pixels for wide field imaging every time.
My biggest concern so far are the statements that Canon has still not moved off their 500nm fab (in contrast, Sony uses 180nm fabs and Samsung is apparently using a 65nm fab!!), and that the 5Ds/R have the same read noise as the 5D III (!!!...more than the 6D???). I'm quite dismayed by the statements that the 5Ds has the same read noise as the 5D III. I know my 5D III very well. It's read noise is really high at lower ISOs (and doesn't have the greatest characteristic at higher ISOs), and it's the thing I hate the most about it.
I was hoping the 5Ds would have read noise more like the 6D or 7D II at the very least, which is lower, not a lot but still lower, than all their predecessors. The 6D has something around 25e- RN, where as the 5D III has closer to 35e- RN, at ISO 100. I honestly cannot see how that could possibly lead to better low ISO IQ than the D810 or A7r, both of which have ~4.5e- RN at ISO 100 (and about two stops or more dynamic range). All the rumors and the press release kept touting the low ISO IQ of the 5Ds, but without any improvement in read noise (over the 6D even, let alone the 5D III), I think the 5Ds is going to be a flop...at least from a competitive IQ standpoint. Personally, I'd still take the D810 with it's 36.3mp FF sensor over the 5Ds. The significantly lower noise/significantly higher dynamic range is going to mean each and every pixel is better quality...I could probably upsample D810 images to 5Ds size and barely be able to tell the difference, and most of the difference would be in terms of spatial resolution. That...is a really disappointing thought...at least to me. Quite, quite dismayed at the early reports of the 5Ds read noise.
