Sony Alpha 7S Review

Give me that cube! Now!

But then you'll see DIY mods to turn it into a half-normal camera.

Complaints about Sony nomenclature always fascinate me.

Somehow, people get lost in their 4 mounts, even though I think it's as clear as it gets: FF/crop for DSLRs + FF/crop for mirrorless. AND crop lenses are full compatible with FF mounts, with automatic detection and framing.

Say, Canon has "only" three mounts, but why? Because they do not have mirrorless FF yet. Plus their crop lenses are not compatible with FF, at all. And no-one is complaining. Go figure...

It was a long time ago longer now than it seems that Sony made DSLR's

@Vadim whilst I agree with you ...
Sony only have two mounts. On top of each mount there is FF or cropped lenses.
As you say a differentiator from Canon is that you can use the cropped lenses on an FF body. That is valid for both mounts.

But I've seen alot of Canon users complaining about Sony tempo when releasing bodies. Maybe the Canon tempo is a better fit for them?

Can you use a Canon crop lens on a Canon FF body
that's news for me

@Eleson

> That is valid for both mounts.

Not true for Canon, unfortunately. You cannot use EF-S lens with EF-mount (FF) camera.

Speaking as someone who has used small-bodied interchangeable lens cameras since their inception with m4/3rds, you will never have the shooting comfort of a full-sized DSLR. You save the weight, but you gain the pain!!

For example, btw, we'll be doing 1 EV less pushes for APS-C (normalizing brightness to ISO 3200), and 2 EV less pushes for mFT (normalizing brightness to ISO 1600), b/c otherwise you're just comparing really noise ISO 100 pushed images to ISO 6400 images (which, on mFT, will be significantly noisier to begin with than ISO 6400 on full-frame).

So we definitely understand your concern, just that we're designing the test to benchmark cameras against the best, b/c we don't have infinite time & resources... for now, anyway.

Also, just FYI (I think you know this), the ISO-invariant test isn't about whether a 6 EV push is reasonable as much as it is about determining the ISO above which there's little to no benefit to amplifying in-camera vs. in-post (for most tones, anyway).

What this shows you is how you can gain back many stops of highlight detail in high ISO shooting conditions, for certain cameras. For certain cameras, 4, 5, 6 EV pushes are not unreasonable, and gaining 6 EV highlight headroom is not trivial - and is something of massive potential benefit to the very people here claiming they want higher DR at higher ISOs.

I know you've said that extra highlight headroom isn't worth anything if you don't use it, which is true, but since it's mostly bright lights blowing at high ISOs, it most likely will be useful b/c it'll likely bring those back (underexposing via decreasing ISO amp).

Also, just correct the vignetting of a 24/1.4 lens then push your shot just 1 EV & you've already got a 4 EV push - something that easily kept happening in a set of wedding photos I processed this weekend from the 5D Mark III. I got unacceptable banding well before the 4EV push was completed, at ISO 100.

Now, you know that a 4 EV push at ISO 400 will be better than that 4 EV push at ISO 100, so you can extrapolate from our ISO-invariance widget that the results of a 4 EV ISO 100 push will be roughly somewhere in between our ISO 100 + 6EV and ISO 400 + 4EV push.

We'd ideally do what you're suggesting so you can see exactly what this is, but it's important to design tests that ensure comparable results, to maximize their utility. And a one-off 2 EV push test doesn't tell you anything about the massive amounts of highlight range you can reclaim by underexposing ISO-invariant sensors by decreasing ISO, nor does it actually show differences in Raw DR between current generation cameras.

""ISO-invariance" is a different thing, and guess it's useful to know about that too if you want to shoot ISO-less, but don't think it's necessary to mix it up with the tests of base ISO DR."

Yes, agree w/ you here, which is why we split the pages into 'Exposure Latitude' & 'ISO-invariance'.

"To get an idea about [base ISO DR] it's just necessary to underexpose for example 2, 4 and 6 stops and lift the shadows afterwards."

Well, it's not just about how many stops base ISO shadows can be pushed to get a reasonable result, b/c ultimately it depends on where those tones were to begin with. If they were originally exposed to be above a certain acceptable SNR, then you'll be able to push them however many stops you want.

So it's hard to design a proper Raw DR test.

But you are correct that what you're asking for would be useful, just more useful were we shooting a high dynamic range scene, or wedge (as opposed to our limited contrast ratio studio scene).

...

...

We'll consider adding what you're asking for as a part of the Exposure Latitude test. To get back to your original point of not quite seeing why we chose a 6 EV push, we did so for full-frame b/c it's where the best sensors just start showing some differences between the best ones out there. So if we can differentiate those, we can definitely differentiate the rest.

And starting at ISO 6400 levels of noise (b/c we picked an ISO 6400 exposure for the ISO 100 shot... 6 EV underexposed) for full-frame means we're somewhat future-proof... future FF cameras that significantly improve dynamic range will most likely have to also improve ISO performance (they'll have to collect more light), & so these tests will show levels of noise comparable to < ISO 6400... in other words, the improvements should hopefully be visible.

For APS-C, we'll start with a 5EV ISO 100 push, b/c ISO 3200 levels of noise on APS-C are about equivalent to ISO 6400 levels of FF noise. And so on...

We'll try & incorporate what you're asking, but for now, we held the Canon FF DSLRs to the same standard as the other FF DSLRs. Lowering it to a 4 EV push, e.g., would've lowered it to the 'standard' test we use for mFT in this test (unpublished so far, but forthcoming).

So it's just a matter of whether we can invest the resources to show the +5, +4, +3 +2, +1 EV pushes for FF, then the +4, +3, +2, +1 EV pushes for APS-C, etc.

Or whether for the particular test you're asking for, we just shoot a wedge with DR greater than the DR of the camera, and show +1 all the way up to +6 (or more) EV pushes of that. That way, you cover all tones and show where noise creeps in.

Of course at this point, with the lowered visual utility of a wedge test, you might wonder how this'd be any more valuable over DXO's measurements... :)

Point being: you're right, just that we optimized to differentiate class-leaders, which also allows us to then show how other cameras on the spectrum fall in relation.

Yes, understand it's hard to design a proper RAW DR test, but I'm sure that you (and the rest of the team) will come up with something that makes the reviews even better than they already are. Thanks for listening.

Thanks for providing valuable feedback.

Ah, okay. I thought it might be related to DR in the highlights, but I guess not. Maybe that doesn't even make sense. :-)

Well, while it's not technically related to camera DR, the lens can have very minor local effects... for example a lens with a lot of flare or poor contrast might actually help w/ captured DR if it elevates dark tones above the noise floor.

But this is so rare, or so trivial, that it's really not worth considering, as far as I am aware.

Decimation is actually superior to downsampling. The NX1 can already be compared in its 1080p mode. Having a camera perform decimation is a waste of battery life...capture the highest sampling you can RAW, decimate in post.

In the video world it will soon be considered that 1080p modes are akin to the jpegs of stills vs RAWs.

Really when we post process stills often we are decimating to a lower resolution output for display and print as well. Same thing for video...best practice is capture as much data you can up front, process off the camera. I think Andrew Reid's comment of a GH4 battery lasting some 5x the amount of the A7S is rather condemning of the 'do everything in camera' angle. Does not deserve gold IMO. People make a big deal about battery life for stills...battery life is even more sacred for video! Not the best stills camera...not even the best video camera if you actually pay attention to Andrew Reid...but it gets gold cause its 1080p looked good...

I define decimation as the act of reducing the sampling rate (i.e. 4k -> 1080p) with the inclusion of low-pass filter for anti-aliasing, etc. And downsampling as just reducing the sampling rate. My argument is very simple...when you capture as much sample as you can up front, you can take what you have and post process therefore 1. Enabling you to save battery life on the capture device itself. 2. Allow you to customize and select whatever decimation/downsampling techniques *you* creatively want for your own purpose and taste.

I never mentioned lineskipping and do not believe decimation = lineskipping. If you think about it lineskipping is throwing away samples now and then on the vertical axis, and throwing away EVERY SINGLE SAMPLE on the horizontal axis normal to the vertically skipped sample...thats why its complete garbage. Also the 5DMKIII I'm not sure if it line skips like say the 5DMKII/6D but whatever it is is clearly out of date.

Whatever the case may be in terms of pedantics...the methodology of capturing as many samples as you can in the capture device, and doing what you want with it in post...is pretty sound. I suggest that DPreview do some research on what some 'best-practice' video tools are, and how well they work at downsampling 4k to 1080p, and just like you provide studio samples showing 'ACR' rendering of RAW, provide industry standard video down-sampling of 4k to 1080p. Because if we are talking video cameras with 2k+ price tag, it warrants more than amateur video analysis.

So it turns out you didn't know the meaning and use of the word. Decimation is a word for line-skipping! Once we correct your usage, it turns out we agree.

Please provide a trusted reference that clearly shows that decimation is the definition of line-skipping! Until then you are thread(or comment) crapping big time. I use decimation and downsampling in the most generic term in respect to digital sampling. Please provide reference that in the video world decimation = line-skipping because I have not seen it.

5D3 doesn't seem to line skip. It gets like 2 stops better SNR than 5D2 for video and doesn't have all the aliasing and moire of the 5D2. It IS soft, but seemingly not straight off the sensor since ML can intercept the early video and it looks pretty sharp, only at some later stage does it turn to waxy mush.

I believe "downsampling" is widely accepted in the photography world to mean resizing to smaller dimensions (i.e. numerically smaller pixel grid). Data is not necessarily simply "thrown away", but typically compressed to such an extent as to not be recoverable to its original state. (For demosaiced Bayer filter array images, since 2/3 of data is interpolated in the first place, actual loss of detail would not be expected to occur until approximately the same reduction in pixel dimensions.)

You can do downsampling in several ways. By decimation is one naive form of it. Some methods of downsampling are called "resampling". If one were deriving a video frame from a downsampled sensor readout (as on the NX-1), one would prefer something very clean, such as bicubic, or Lanczos windowing.

resampilng = interpolation + decimation by some rational factor. Yes this is the best method and the type of interpolation you use provides different results. This is an expensive operation best done in post. Yes I should have used the term resampling.

However the term downsampling was just as wrong to use as well. Resampling = interplation (upsampling) + decimation (downsampling). So it is just as wrong to consider downsampling as the 'best' operation if you want to be on point.

Back to topic...considering that now we agree that *resampling* using various algorithms produces best output...I think we agree this is best done in post to save on battery life and to have the best IQ in the end result.

Bicubic, Lanzcos, etc are *interpolation* and upsampling algorithms and not downsampling ones.

Although again 4k is already an oversampling of 1080p by a rational factor so can't we get right to decimation without a big penalty?

At a very mathematical level, interpolation is involved in downsampling as well, so the mentioned algorithms can be used for both up- and downsampling.

Thanks for the reply guys.

From the updated information it's pretty clear why you use it as the technical point to compare DR. It would be interesting though to see what the a7s and a7r files look like though as you progressively let the highlights clip.

Maybe someone with both cameras and enough time on their hands could try it by finding the exposure that preserves the highlights, dialling it down by 2 EV, and then using the bracketing feature to show how the images compare. Then one could resolve once and for all whether there's any real world advantage to the A7s.

We have all those Raws shot already, as we do for every Raw DR test, since we bracket a ton to choose the proper ETTR exposure. If you'd like the Raw files for yourself to give it a try, PM me & I can send them to you.

We just don't wish to spend the time demonstrating this since we can quite easily see in our Raw converter that there's no difference in how they blow... :)

I don't really care so much about JPG's. But where that breaking point is, in RAW is what I'm interested in. And not only that, what it looks like. I found the shadow pushing to be quite interesting, comparing what it looks like between cameras (thanks). I figured it would also be interesting to see what the highlight retention looked like.

You say you exposed to the right, what did you base that exposure on? Hopefully not the on-camera histogram, because that is typically based off the internal JPG processing. Unless you take the RAW file into an editor, and pull back to see where the actual breaking point (highlight leeway), "exposing to the right" could potentially be limiting your dynamic range potential.

I find that my cameras typically will have another 1.5 stops of highlight potential, with full color retention, beyond what my in-camera histogram shows me (I can't wait to try out the D750's raw histogram, soon).

Perhaps, we would find nothing in the test, but maybe not.

Thanks guys, I wasn't aware that you were looking at the actual raw data while setting up the exposures, knowing how many absolute values are in 14bit, you would know how close you are to the "ETTR", and you're right middle grey becomes in-consequential.

Yeah, exactly. And we can't look at Raw values at the time of exposure (one of our major gripes to camera manufacturers, to whom we ask 'Raw histograms, please?'), so we just bracket like monkeys, then throw them, one after the other, into RawDigger & then start doing some pixel statistics.

mosc: thanks, that's very kind of you. And you're absolutely right - the A7 series ends up being like a case study in DR and noise across similar (yet not) sensors of the same size. We're able to correlate the differences in pixel count, upstream read noise, and downstream read noise (including what I feel are major impacts of quantization error w/ the a7S, which goes against 'bigger pixels are better', with respect to DR anyway) with image quality impacts. Especially in the 'High ISO Compared' article's second page, where we can even look at how different tones are shot- vs. read-noise limited.

Another fun thing we found was that downsizing a 12, 24, and 36MP camera to 8MP still shows sharpness benefits to the higher resolution cameras, w/ the 36MP a7R still showing a benefit against the 24MP a7 II, even when viewing at 8MP normalized viewing size.

See here.

I mean, I'd always known this, but it's pretty cool to see visually, in a controlled way.

At lower ISOs, the A7s is not that good. Low DR compared to less expensive FF And APS-C cameras. Resolution is low too making issues more pronounced. DPR did a great job exposing this.

It is not a matter of it being bad. And no one have said it is. Using the BAD argument is just dishonest, or maybe just clueless.

It is just that it is not that fantastic good as some claim. Take a look at the comparison at dxomark.

http://www.dxomark.com/Cameras/Compare/Side-by-side/Sony-A7S-versus-Sony-A7R___949_917

Then look at SNR, DR and TR. SNR and TR are the same for both cameras. No important difference at all.

For low ISO the A7R is better and for high ISO the A7S, regarding DR. Regarding resolution, the A7R is always better.

Now - which do you think is more important, TR or DR?

Good to know. Thanks guys!

So the criticism is that the camera's 1080p in "super 35" mode is not as good but still nicely oversampled but the 4k mode for that crop which no other manufacturer offers is not very pretty. If they had just left off the option of a 4k output size from the crop mode it doesn't seem like you would have called it out on the "cons" list...

This negative is left over from it's comparison against the GH4 which has it's own issues with 4K capture from "super 35" lenses, it doesn't use the whole lens!

Neither should be a negative on either camera.

I still use my SX1-IS on a tripod for recording 1080p video at kids events in the school gym or other indoor venue. Bought it before they were born and they're 6 now. 1080p is still the standard and it has plenty of zoom range + fully articulated screen.

Though the file format hasn't aged so gracefully. Serious pain.

In the UK Panasonic are currently offering a free 5 year warranty on the internally recording tiny LX-100 and the internally 4K recording FZ1000.

You need an entire additional Shogun-type rig to maximize 4K from a Sony A7S, making it so expensive you may just as well go pro and buy a decent camera.This doubles its cost, and then you have to have lenses, so.......

It is also unfortunately unclear from this review what area of the full-frame is used for 4K output, so I do not know if wide-angles are cropped etc, but it is dispiriting to note the limitations endemic to using off-brand lenses via adaptors here.

By limitations on off-brand lenses I mean they have no image stabilization, unlike the new A7 MkII. This deters us from buying it for both stills and video, as it does the A7R.

We simply cannot go on buying more and more bodies to get the spec one body can provide. A new camera now should have both 4K and HD with a full-frame image-stabilized sensor, and the ability cameras normally have to work with any lens taking advantage of the in-built IS as Oly 4/3rds bodies have since the year dot.

If Sony/Olympus "upgraded" the E5 to output 4K internally they could rival the GH4! (Just a thought!)

For 3:2, it requires fewer MP in that ratio to get the same 4K width. The A7-S could be just a 3840x2560 camera (just under 10mp) instead of 4240 x 2832 (12mp) and supplied the same width 4K video it does.

That said, there are benefits to a little extra width from the senosr for getting the corners better. I would say 4K really limits your quest to reduce resolution to the 12mp range.

"An APS-C camera w/ the a7S' pixel pitch would produce 5.3MP images. "
Yes I understand that. That is why I stated I'd be happy with 8MP.
Pixels wont be as large as 7S but a hell lot better than the willie-wow BS megapixels of current APS-C now hitting 24MP. I really have not use for 24MP and all that bloat and slowness added specially when shooting Raw. It really irritates me.

(It would have helped to have shown highlight clipping in the scenes as well as a cherry-picked piece of shadow and then provided exposure info too.)

It is a given as the camera has been out for a while & DXO results have been known for ages.

Ok you find the comparisons at higher ISOs less relevant, but what about those who don't?
I shoot often between 12800 & 51200 ISOs with an A7s.
You knock the A7s for dynamic range but I think you should make it clear that is for dynamic range at base ISO.

The A7s curve for dynamic range is very different from just about all other cameras. -many that start higher go lower very early.
Can you tell me what cameras there are that ARE better for dynamic range at ISO 12800 if that is what someone needs?

I just don't get how a low light camera that clearly has MUCH better dynamic range at the high ISOs it is mostly used at can be marked down for having a slightly lower (but still ok) dynamic range at base ISO.
Again, I also think the camera is rated higher than it should be for a general use camera.....but dynamic range WHERE IT IS USED, like video is not an area it should be marked down.

No, it's not at all argumentative. I was genuinely curious if neil holmes had a reason to expect lower DR a priori.

Your assumption that I was being argumentative is unfortunately probably a sad reflection of the argumentative tone of many people on this site.

neil holmes - thanks for the clarification. Yes, given what DXO has reported, I can see perfectly why it was a 'given'.

Also, the camera wasn't knocked down for base ISO DR - we scored it primarily as a video camera. Though even there, many DPs who are in constant search for more DR would benefit from more base ISO DR - as long as you can preserve it in the output format (which we haven't verified).

Furthermore, you keep mentioning that it's higher ISO DR that's more important for this camera. I'd then ask - if that were true, why did Sony quote 15.3 EV DR, presumably a measure of base ISO DR (and not even correct, at that)?

Also, I should've been more clear in my 'less relevant' comment- I meant it's less relevant for dynamic range concerns, b/c most people requiring higher DR will be shooting at lower ISOs to get closer to capturing all that DR. And if you really did need more DR with the a7R, you could underexpose, say, an ISO 6400 exposure by dialing back ISO amplification by 2 EV to 1600, where the a7R has the same DR as the a7S, which you can see in DXO data here. Which is also more, btw, than either camera at ISO 6400. This assumes, of course, that you can give the camera enough exposure to make use of that extra 2 EV on the highlight end.

As you get to higher ISOs, the DR becomes more a reflection of ISO performance, & that was my bigger point. So by weighting high ISO DR in a DR score, you'd, overall, be counting ISO performance twice, since the camera already scores well in the 'ISO performance' category.

There are few who would buy a A7s to shoot for maximum DR at low ISOs. It is ok though. DR becoming more a reflection of ISO performance may be true somewhat EXCEPT less so for the A7s.

A7R, 5Diii, D750 all fall away similarly to be about a stop behind at ISO 12800 for DR and gets further ahead as you go up.
D4s and 1Dx I think are the two cameras that can keep up for DR AT ISO 12800 the most (very close even) but even they both fall a stop away by ISO 25600.....and at low ISOs the Nikon D4S is a fraction better for DR and the Canon 1DX a stop worse.
Noise may not be as much of a difference for some cameras but add DR into the mix ....no.

Rishi, thanks, you did clarify a little but ..... I thought about this a bit at work today and I am quite annoyed!

For a start, dpreview have, for a very long time, provided a dynamic range page. Yes, there are problems with that as it is based on jpg processing but the ability to see how things change in different modes gave a good pointer, especially as we could see the "wedge" and how it exposed. Now that has been replaced by contrived (yes, I'll use that word again and explain why shortly) "tests" trying to push a point which does not inform us clearly about the camera being tested. The test does not actually give any number for dynamic range.

Read on...

...cont'd

For one thing, you say "- I spelled out that the clouds above the sun were just short of clipping, & that 1/3 EV higher exposure clipped the clouds" but you don't show us and we don't know how much the highlight exposure changed between shots. We don't know how highlight recovery affects the different colours, because you don't say (which is very important - for example, the Canon 60D I once had is not renowned for having a high dynamic range, yet it had an uncanny ability to recover blown blue in skies). We know nothing about highlight recovery from your "review" at all. Even if I hunt through your hidden information to find the exposure parameters, I don't know what the "correct" exposure for the scene was or where on the tone curve you really exposed. It's not objective, because it's been CONTRIVED to prove a point.

read on...

...cont'd

Yet that point fails! You show results from pushing exposure a huge amount, resulting in both cameras producing results which are acceptable for small prints but neither of which are acceptable for any critical use (e.g. A2 print, magazine page, anything for a paying client). Plus, you've had to use a much bigger file and spend more time and resources downscaling it to make the 36 MP sensor match the 12.

Finally, the HDR shot from a Nikon 810 you posted recently (same sensor as A7r), roughly comparable to the HDR sunset posted in this review, shows a level of blotchiness in the skies which is visible even when the image is downsized a lot. The shadow murkiness and noise are pretty horrible. The pushed A7S file is much, much cleaner, hence my comment it is a forgiving sensor.

deep7: "For a start, dpreview have, for a very long time, provided a dynamic range page."

You mean like page 14 of our review here, which was actually published many months ago?

"[The D810 HDR shot you posted recently in the Tamron gallery] shows a level of blotchiness in the skies which is visible even when the image is downsized a lot. The shadow murkiness and noise are pretty horrible. The pushed A7S file is much, much cleaner, hence my comment it is a forgiving sensor."

There's something very wrong going on there if that's you're analysis :) Please try downloading the original full-size JPEG, and downscaling yourself in ACR or Photoshop. It's nothing of the sort, and significantly better than the a7S, when normalized.

Ta. That's bizarre - the full list of pages didn't show here yesterday. Sorry for my error.

I found this bit telling: "You'll note that the highlight range gained at the higher DRO settings can be so extensive (up to 1.7 EV) that the highlight cutoff falls off the right side of the chart. Despite our chart not having enough highlight range to show you where the highlight cutoff occurs, you can expect to gain back in highlights whatever negative exposure shift you apply - while still maintaining proper midtone brightness." That's nice.

It seems you can play that curve (DRO or Slog) to push the curve off both ends of the charts. Maybe that's where Sony get their DR number from? I think I'll borrow an A7S and try it out for myself (I have tested both the A7 and A7r and came very close to buying the A7, hence my interest here).

Rishi,
Thanks.
I am just going by DXO mark and from my using a few cameras over time.
For what I shoot a lot in low light, there is nothing better than the A7s.
I don't use base ISO to shoot but use auto ISO and have it set up to 25600 sometimes and 51200 at others.
All I can do is say that for me at least, the dynamic range of the A7s is clearly a PLUS and certainly not a Con no matter how it does it.
Still a better than average camera at ISO 100 and better than anything at ISO 25600.
How or why is way beyond my limited knowledge.
I just know I can mount a lens like 17mm or 24mm Canon TS-E L lens to the A7s and shoot away at a concert and get photos at f4 or 5.6 (even f8 if careful) hand held at ISOs I would have thought impossible even five or six years ago that the bands love.

deep7: "It seems you can play that curve (DRO or Slog) to push the curve off both ends of the charts. Maybe that's where Sony get their DR number from? I think I'll borrow an A7S and try it out for myself (I have tested both the A7 and A7r and came very close to buying the A7, hence my interest here)."

No, b/c no matter what you do w/ DRO or SLog2, you're never going to get more DR than the Raw DR at that ISO (measured by DXO). Since SLog2 appears to increase hardware-level ISO amplification to ~ISO 640 levels, you're not going to get more than ~12.2 EV DR w/ SLog2. We actually measured this as well, we just don't publish it, b/c it's not a standard methodology yet.

DRO only affects the JPG, btw. SLog2 affects the Raw in that it changes the actual amplification off the sensor, but doesn't do anything non-linear. We haven't seen any nonlinearity in sensor response to tones yet - if/once manufacturers can do that, we may see a huge increase in DR.

deep7: "Even if I hunt through your hidden information to find the exposure parameters, I don't know what the "correct" exposure for the scene was or where on the tone curve you really exposed... it's been CONTRIVED to prove a point."

It's only contrived if you're confused about the point we're trying to make, and then assume bad bias on our part.

It'd be kind of odd for us to be brand biased and biased toward particular models within a brand, no? :) Btw, the a7R which pulled ahead in our contrived tests - we then knock it when we say 'but we'd hesitate to recommend the a7R b/c of its real shutter shock issues'... man we are a fickle bunch!

Truth is, nothing was contrived. We did a proper Raw DR test, which you can read about in our original iteration of the test with the 7D Mark II here

As for highlight recovery, no camera shows appreciable recoverability over another, w/ clipping determined by your choice of exposure.