Click on image to enlarge
Focus Numerique (google translation here) posted the full Sony A7r ISO test. I translated their key findings:
– At lower sensitivities up to ISO 800 the performance is just remarkable! Details are clear and noise barely noticeable.
– Only from ISO 1600 you see first effects of smoothing but the details are still there and dynamic range is intact.
– At ISO 3200, the smoothing is a bit stronger, but once again the treatment due to the image analysis (solid, edge, texture) works really well. The dynamic range drops slightly, but nothing dramatic.
– The overall image is a little more dense, fine details are blunt and the rolling becomes more easily visible at 100% on screen. Chromatic noise is still well under control and images are still easily exploitable.
– At ISO 25600, the grain still looks natural, destroying a little more fine detail, but the results are still impressive, especially with a resolution of 36 megapixels.
This is a very impressive result from the Sony A7r! It looks like Sony was able to improve considerably their image processor engine. We were used to see Sony lagging behind Canon-Nikon. This now changed with he A7-A7r.
—
Specs and Preorders in US and Canada (Click on store name to see direct preorder page):
Dedicated Sony announcement page at Amazon US (Click here).
Sony A7r body at Amazon, Adorama, BHphoto, SonystoreUS. TheCameraStore Canada.
Sony A7 body at Amazon, Adorama, BHphoto, SonystoreUS. TheCameraStore Canada.
Sony A7 with 28-70mm kit lens at Amazon, Adorama, BHphoto, SonystoreUS. TheCameraStore Canada.
Sony RX10 at Amazon, Adorama, BHphoto, SonystoreUS.
Zeiss 24-70mm at Amazon, Adorama, BHphoto, SonystoreUS.
Zeiss 35mm at Amazon, Adorama, BHphoto, SonystoreUS.
Zeiss 55mm at Amazon, Adorama, BHphoto, SonystoreUS.
70-200mm A-mount at Amazon, Adorama, BHphoto, SonystoreUS.
Sony A-Mount to E-Mount Lens Adapter at Amazon, Adorama, BHphoto.
Sony A to E adapter with mirror at Amazon, Adorama, BHphoto.
A7/A7r leather case at Amazon, Adorama, BHphoto, SonystoreUS.
A7/A7r vertical grip at Amazon, BHphoto, SonystoreUS.
Sony W Series Battery Charger at Amazon, BHphoto.
Sony FA-CS1M Off-Camera Shoe at BHphoto.
Semi-Hard LCD Screen Protector for A7-A7r at Amazon, BHphoto.
Specs and Preorders in Europe:
Sony A7r body at Sony DE, UK, FR, IT, ES, NL, BE, CH, AT, SE, FI, NO, PT, WexUK, WexDE.
Sony A7 body at Sony DE, UK, FR, IT, ES, NL, BE, CH, AT, SE, FI, NO, PT, WexUK, WexDE, Amazon UK, Amazon DE, Amazon FR.
Sony A7 with 28-70mm kit Sony DE, UK, FR, IT, ES, NL, BE, CH, AT, SE, FI, NO, PT, WexUK, WexDE, Amazon UK, Amazon DE, Amazon FR.
Sony RX10 at Sony DE, UK, FR, IT, ES, NL, BE, CH, AT, SE, FI, NO, PT, WexUK, WexDE.
Zeiss 35mm at Sony DE, UK, FR, IT, ES, NL, BE, CH, AT, SE, FI, NO, PT,
Zeiss 55mm at Sony DE, UK, FR, IT, ES, NL, BE, CH, AT, SE, FI, NO, PT
Asia:
A7r at Digitalrev
A7 at Digitalrev.
A7 with 28-70mm lens at Digitalrev.
117 Responses
I think smoothing is on the higher side on the A7R which may be problematic with details on dark colors. As always High ISO noise is always problematic with dark colors and that is where Sony does a lot of smoothing to create the impression of low noise
A7R is just underexposed (faster shutter speed used). That’s all. Wait for proper studio shots from imaging-resource
Good catch. The difference is despite a 2/3 stop underexposed setting used in A7r compared to D800.
Splitting hairs about the quality of JPEGs are we? *sigh*
JPEG or RAW issue aside, a logical comparison should try to match the exposure for any evaluation. A 2/3 stop difference is significant (and Leica is used stopped down, to top it off).
Please check these comparisons, all the description can be found in my message at the bottom
In short these are crops from raw files processed thru same software, no NR, no sharpening, no lens correction
exposure was matched between cameras. Auto white balance instead of “As shot”
http://d.pr/i/xwCX
http://d.pr/i/Yf6i
Was to be expected now that the pellicle mirror is removed. Nothing surprising.
Actually there are no big differences with the D800. After matching the A7r brightness with the D800, the noise will become more apparent.
Sony has surpassed Canon’s image quality a long time ago, and Nikon used Sony sensors for years. Only Jpeg processing was weak in the first camera generations.
Actually, the A7r shots are underexposed by 2/3 stops compared to D800. That should help D800 but it still appears to trail.
A7 results should be even more impressive
Not at 36MP ;P
could be same or even worse on print.
really? I thought it would have less noise because there are less photosites per unit area of sensor
Hardly any correlation in practice between pixel density/pixel size and SNR for a given output size/unit area.
then why does Nex 7 have a bit more noise than Nex 5n
@erotavlas – Simple: that fisrt generation of 24mp sensors weren’t so good. This new generation, coming with the Pentax K3 I guess results will be the same (perhaps even marginally better in favor of the newer 24mp sensor).
Zohee is being tricky there. Denser sensor would have more noise at the pixel level (i.e. like when you pixel peep at 100%) compared to a sparser sensor of the same generation. However, when you normalize both outputs to a certain size, they would perform similarly. So if you downsize the higher MP picture to the lower MP picture, you’re averaging the pixel level noise; it would be less than before. If you upsize the lower MP, you pretty much magnify the pixel level noise. This comes with the assumption that you have resized them correctly (proper sharpening/interpolation).
The only small difference between equally sized pictures from the 7 and 5n usually comes from different physical exposures since the 7 understates ISO’s by at least a third of a stop.
Nothing tricky about making output size equal. If you wouldn’t, that would mean that for example the D800 and D7000 perform equally for having equal pixel performance. Ignoring the obvious fact that sensor area/pixel count counts and apples vs apples involves scaling to equal size, regardless whether print or screen and regardless which size.
For non BSI-sensors, more pixels translate to less sensitivity due to wiring in front of the photosites. Since Sony doesn’t claim this sensor to be Exmor-R, this sensor probably isn’t BSI.
I am very interested in seeing the a7r performs in low light.
Here in Sweden it gets pretty darn dark during most of time of the year and fast lenses only won’t cut. I shoot mainly street.
can’t wait for proper reviews on the focus speed on the a7r
is it only me or the d800 still has a better image than the Sony?
Yes, I’ts only you.
I prefer D800 results also.
No.. D800 seems to be better.
Based solely on those two images, the Sony won the 12800 battle but the Nikon won the 25600… and sadly for Sony their win was slim, the Nikon win was by a large margin. Not sure the test didn’t mess something up to get result that swap so decidedly.
Yes, D800 holds onto more detail. The E has about a 3rd stop advantage on top of the D800.
Dont put too much stock in Nikons ISO numbers. They have been fudging them for a while now.
D800 isdefinitely better. I’ve been seeing this over processing by Sony and thought it was just the JPEG’s which in this case it could be. If it turns out to be inherent in the raw files it will be a disaster for the camera.
I think Sony have done similar Jpg processing as Canon’s 5diii etc. They have just smudged everything to hide noise. This ‘hopefully’ shouldn’t be present in the raw files however.
Is it possible to open the raw files yet, does any software support them?
Just head over to DXO. They measure everything, including ‘stated’ and ‘real’ ISO. For some reason Sony just seems to be behind. I’m not talking about SLT’s, but the RX1 should be upto par as the D600 but it isn’t quite.
Thats because many of you didn’t notice slower shutter speed at same ISO, Aperture on D800, meaning that A7r images were underexposed compared to D800.
Or it could be the difference in lens T stop. Aperture does not actually measure the amount of light that passes through a lens, and thus all lenses will pass more or less light at the same aperture.
Nothing to do with t-stop which would have to imply that the lens on A7r has a t-stop nearly 2/3 faster than the lens on D800, with both at f/5.6. And that doesn’t explain f/6.8 as the choice for aperture for Leica.
There is no excuse for not matching exposures properly… but yes, same lens would make the comparison even better.
I’m looking forward to A7 iso test… everything is always A7r this and A7r that.
That’s mainly due to the A7r sensor being of such a new design and therefore pushing the boundaries, theoretically, somewhat further than the standards that came before it. It’s really the litmus test that people will use to measure Sony’s photoimaging prowess against other manufacturers. That it holds its own against the D800 at a “meets or exceeds” level is a nice treat.
+1
Sony Pro/broadcast is using a new “3D noise reduction technology” for video. They might have ported that over to Sony Alpha teams.
Do you write science fiction for a living?
You should.
I meant to write from my phone “maybe they have ported it over…”
Yes…Sony does have what they call 3D noise reduction….its not fiction.
Why SLT design for the pro then? Wedding photographers need this type of clean high ISO.
Because the AF tracking is poor on mirrorless. I’d take fast AF over the third-stop loss on my SLT any day, the SLT gives me what I want where mirrorless and dSLR does not. Others have other priorities, good for them and I couldn’t care less.
High ISO shooters should be using Nikon, if I wanted to be a lamer high ISO noob impressing people in forums I wouldn’t be a Sony shooter.
Weird – at 12800 the Sony looks clearly better, but at 25600 the Nikon…..
Yeah, it’s like somebody mixed up 6400 and 12800, but I doubt that was the case.
Why these good people would compare the A7r to a D800 and not a D800E will remain a mystery. lol.
Yuck. In my opinion, its still the classic mushy smoothing typical of Sony. “Monet” look, as someone recently said. The Nikon looks so much more natural and realistic to me.
1600 and 3200 look pretty identical. This is great.
I do find it odd that ISO 12800 actually looks sharper than 6400. Haven’t checked out the full sample, but the “PN” is significantly sharper.
Fingers-crossed production model with final firmware will have even better performance.
If you look at the whole test scene, then Sony JPEGs are by far the worst and RAW files–between M and D800 (at ISO 6400). Can you read “PONT NEUF” on the Sony chart? I can’t.
The most impressive, easily, is the M. A shame it is way too expensive.
It is also 24MP.
Odd, isn’t it? That could deliver more details than the 36MP. 😉
Three cameras, three different settings… A7r gets underexposed (2/3 stops compared to D800) and Leica gets stopped down treatment.
Sony uses unethical moves now with cheating with those tests. Stupid to lie like this when everyone is already hooked up. Images are not exposed equally. I hate to say it but I lose faith in Sony. If they feel they need to lie to us then it means those cameras are not really good.
What are you talking about? Sony didn’t do these tests. It was done by a third party website. They probably set shutter speed to auto when everything should have been done in manual to keep the same base tests. It’s not Sony’s fault that the tests were done wrong.
Since when Focus numerique is owned by Sony? You moron.
The M test results look completely different at all ISO settings. It makes the a7r/D800 look muddy and unrefined. Is it the lens they are using or the sensor?
I think you have been caught up in the Leica myth,ignoring the couple of blog reports from the likes of Steve Huff.The actual sites that do controlled testing have a very different view.
http://www.dxomark.com/Cameras/Compare-Camera-Sensors/Compare-cameras-side-by-side/%28appareil1%29/865|0/%28brand%29/Nikon/%28appareil2%29/844|0/%28brand2%29/Leica/%28appareil3%29/814|0/%28brand3%29/Nikon
The leica barely outdoes the best APS models and certainly not the D800/E
It’s photos vs maths argument. 😉
I prefer photos 😉
Nope guys its reality versus myth , as anyone looking at the respective results from these cameras will be able to tell.The reason why those people who defend Lecia so fiercely never talk about things that can be measured is that their delusions are to quickly reveled. Better to talk about ephemeral things its about the “look” or the soul of the images seriously guys ?
While I agree that any camera in the right hands will be capable of wonderful images. I have to ask why would those who profess to be only interested in the “photo” be wasting time on a gear forum dedicated to the latest and greatest ? 🙂
You cannot quote DXO to support your position and expect to be taken seriously. Anybody can read that junk.
Have you got lengthy practical experience? Any photos to show? Real life stories that happened to you? I mean to you, not to some turd on the web, whom you’ve never met.
Witam
A7r na 12800 ma przewagę nad D800 ale już nie ma na 25600 , a D800 ma prawie dwa lata . Super że w końcu jest postęp i sony w końcu umie dobrze oprogramować dobrze matrycę . Ale jak zobaczę że sony walczy jak równy z równym z następcą D800 i Canonem 5Dmark IV wtedy naprawdę będę się cieszył . A dwa cały czas czekamy na nowe obiektywy z mocowaniem A !!!! Są pilnie potrzebne by móc się równać z Canonem i Nikonem !!!!!
Too much wanking over jpegs. Want to see what you really get? wait until DxO, Adobe and C1 support the RAW files and you got yourself a good comparison field there.
I am not disrespecting the efforts. I am just calling into attention that the “performance of the sensor” in these tests is not representative of what the sensor could really deliver in the maximum allowed performance that the camera delivers before being chucked into a RAW developer.
And darktable too–probably the best free raw converter in the world.
Seriously, A7R raws are so different that generic tonal curve isn’t even close. It feels like something’s wrong with the dynamic range of the 7 line…
The raw files open with the Sony image data converter software no need to enjoy yourself with JPEGs
Yes, let’s wait so we can wank on DXO figures. Much better throw.
I consider the samples provided as excellent advertising for the M.
That and any defenders using swear words. 🙂
They just traded visual noise for shutter noise. I hope those guys doing the test has some ear protection! 🙂
Sony should ask Nikon if they have any leftover oil from their D600 shutter.
That might quiet it down some.
😉
The D800 isn’t even that good. Get it to 6D quality and then you can use the word “impressive”.
Ha, I take that back. It looks better than the D800 and pretty close to the 6D/5D3. Pretty good.
You obviously haven’t used the D800. Otherwise you would know that is nonsense. The D800 has set the bar in 35mm image quality. I work with all digital camera formats on a daily basis. The D800 is a fucking good camera!! So if the A7r is close in quality that is immensely impressive!!
@ GAV: And obviously you don’t know how to read because I said I take it back because it looks better than the D800. The 6D is known for having the best ISO. I’ve owned it so I can say that with confidence. Do you’re homework and don’t base your fanboy opinion on the only camera you use. Thanks for trolling. Come again.
If all the settings were the same than the Nikon images appear to have been pushed a half-stop, hence the brighter image and additional noise in the shadows. There isn’t much latitude in post at those ISOs, so this test seems pretty unfair.
No two cameras types replicate the same image brightness at the same ISO, shutter, f/stop. Even between lenses in the same range there are variations…No conspiracy here…
Sounds like you’re getting overly excited about potential variations or you’ve been praying to DxO before going to bed every night.
Cameras produce nearly identical exposures at the same settings when used in daylight because that’s the same light source all manufacturers use to calibrate things 😉 When test sites report variations between sensitivities, most of the time they’re using some kind of artificial light. So it’s wrong to say that these variations hold true all the time as if manufacturers are sloppy.
What’s funny is I use digital cameras and film cameras side by side and they still produce virtually identical exposures. So I think what you’re doing is mixing up different tone curves with exposure.
And no… different samples of the same lens don’t transmit light differently, if that’s what you meant by “range”. If there’s such a difference between glass batches the entire formula needs to be adjusted, making them different lenses. So what you’re saying is a massive joke and I don’t know where you pulled that one out from in your excitement.
You should try taking more than one photo each when comparing images between cameras/lenses.
From my observation and reading, there’s some brightness variation between cameras(even with identical settings). Possible of potential causes – identical F numbers doesn’t means identical T numbers, photosite vignetting, etc.
Wow genotypewriter…that an easy statement from someone who know nothing of what I do for a job! 🙂 I have had 2 canon 70 – 200mm lenses on the bench give me 1/2 stop difference. Same lens type, same light source, same camera, poor manufacturing tolerances…
So please refrain from being a dick…
You probably tested it wrong.
Yes, agreed – the f-stop does not speak to the light transmitted to the sensor (hence the more accurate t- stop). In theory, if using the same lens and settings the ISOs should give the same exposure to adhere to the international standard.
Judging the samples alone, the Nikon appears to have more gain. If the exposures were all identical I expect the noise advantage of the 7r at that ISO to diminish.
This is great news for all A-Mounts. The SLTs constantly get hammered on by the camera reviewers for bad ISO performance, and poor battery life. Hopefully the next generation SLTs take advantage of this new tech in order to the get market share, and reviews they deserve. I can’t wait to see the SLTs hammer their competition for once.
For ultimate single pixel resolution, the Sigma Foveon tech is still tops. I heard they are prepping their FULL FRAME SD2 as we speak. Imagine the resolution on that mother fooobar.
Imagine the price.. hehehehe
Same test with RX-1r & a99:
a99:
http://www.focus-numerique.com/test-1567/reflex-sony-alpha-99-bruit-electronique-12.html
rx1r:
http://www.focus-numerique.com/test-1729/compact-test-sony-rx1r-bruit-electronique-12.html
a7r:
http://translate.google.com/translate?sl=auto&tl=en&js=n&prev=_t&hl=en&ie=UTF-8&u=http%3A%2F%2Fwww.focus-numerique.com%2Ftest-1782%2Fcompact-test-sony-a7r-bruit-electronique-12.html
——
At ISO 1600 the RX1r has the best combination of detail and noise control with the a99 a close 2nd and the a7r a distant third.
36Mpx resize to 24Mpx, noise will reduce too.
This test shows nothing about these cameras’ abilities when used at their full RAW potential. It’s like judging race cars by how they look driving 25MPH.
Quality and latitude will be better in Raw than Jpeg. So if it is good in Jpeg it will be great in RAW…
Agreed-except until someone answers why the M240 is wiping the floor with the a7r/D800 at 100-800 ISO I say this test is inaccurate. Photographer Huff owns the M240 and says the a7r outperforms his M. According to this test that would be physically impossible. I prefer to believe someone with real world experience.
Huff isn’t really the guy to test anything beyond “first impressions” (but he can makre really nice FI tests with some well thought opinions)
Don’t get me wrong, I appreciate the test and efforts. I just question the real world application of them.
I am sorry that I did not translate and read the entire script. I did not realize that there were RAW files available for download. Unfortunately, they will not open for me in LR5 or Capture One.
Just ordered my A7r. Will be using with Nikon Ai Primes. Will post images…
Which primes? I have the 50mm f1.8 AIS Nikkor, it was only ok on the a3000’s 20.1mp and Nikon E Seies 135mm f2.8 which is still sharp as.
awesome, I have the 105mm 2.5 p.c and may get the 28mm ais. I can’t wait to try the 105 on the A7.
Should have waited for nikon DF – it will have native support for these lenses.
I looked at the Google translation, and though they use the word ‘dynamic(s)’, they don’t use the term ‘dynamic range’. So I’d be careful in writing: “Only from ISO 1600 you see first effects of smoothing but the details are still there and dynamic range is intact.”
As long as the camera is using hardware-level ISO amplification, the ‘dynamic range’ can never remain ‘intact’ (same as ISO 100). Every time you raise the ISO, you lose dynamic range b/c higher signals off the sensor are clipped.
so what is the dynamic of a photo saramgiman ?
i think the web site wanna speak about the dynamic range even they dont mention the word range.
http://www.dpreview.com/galleries/reviewsamples/albums/sony-alpha-7r-samples-gallery
That would give me a hint of how well the new A7 24MP is going to perform in terms of noise & DR. WOOHOO~~~~!!! 😀
An executive at the photo plus expo told me that the a7 used the same sensor as the rx1 & a99.
That is totally consistent with Sony definition of all new sensor….not a used sensor….
I love what Sony is doing, but I just can’t stand the high amounts of smoothing typical in their high ISO jpg images. Everything looks mushy.
bravo sony your not afraid to explore! this could be my next camera!
If the exposure was the same and Sony’s images look darker, it doesn’t mean the test was “unfair”. It would mean that Sony’s ISO scale is shifted. (E.g. They’re claiming it’s ISO 6400 when in reality it’s not.)
But the setup is not properly described here. Better wait for a reliable test like DxO.
ISO is just fine. They simply underexposed in test.
Actually, if the focal plane exposures were equivalent, technically it would mean that:
(1) Perhaps the T-stop of the lens is different from the F-stop (lets in less light than it should)
-or-
(2) The quantum efficiency (QE) of the sensor is lower
-or-
(3) What you said, the processing is such that the resultant file is darker.
One must remember that it’s a whole system here, including analog capture & signal processing (both analog & digital) that lead to the final file.
The best way to really compare all these cameras is to look at the signal:noise ratios (SNR), across the entire dynamic ranges of the sensors, of different cameras for equivalent exposures (which, yes, ignores T-stop/F-stop differences, but these are typically minor).
DxO doesn’t quite do this (they only provide SNR for 18% grey), but they’re definitely a huge step in the right direction, and the best we have in this regard thus far.
Nikon ISO 6400 looks little better than A7r. In Sony sample there is some low frequency noise, missing in Nikon.
But Sony sample is very underexposed. Shutter time is 1/400 vs 1/250 on Nikon.
Basically this is very unreliable test. We still have to wait for proper test.
+1
It’s great that people upload all those sample images but making a comparison is totally pointless.
For a meaningful comparison, they should use:
1. same lens
2. same exposure in seconds
3. make sure that the same demosaicing algorithm is used (there are raw converters that allow explicit control)
1. and 2. are already possible, and 3. too as dcraw already is able to read a7 raw files (if I’m not mistaken). The third point is arguably a little bit of work but step 1. and 2. are easy to implement (if you’ve got the gear).
Actually that’s just taking it a bit too far. Yes great if someone can do all those things, but a simple test in manual mode at the same settings using something like lightroom is more than adequate.
It really depends if you want accurate answer or quick answer.
From this test all I can tell is that both cameras are comparable in terms of IQ. And I never expected anything different.
But I would be far from drawing any conclusions like Admin just did.
The only conclusion Admin did is that Sony isn’t lagging anymore.
I like! I’ve waited a long time for a real FF replacement for my Contax 139!
Indeed I’ve ordered an A7r body from Sony that should arrive 15-20 Nov. My lens choices are the basic 18-55 from my NEX-5N or (mainly!) some Zeiss (Contax T*) which I use with a Contax->NEX adaptor.
The NEX lens will only work in cropped mode, though should still offer more performance than the 5N, but what about my adaptor for the Contax lenses? Won’t I need a new one to cover the full frame, or will it work just fine(?)
Thanks!
I have converted the raw files and made a comparison between D800 and A7r using the same software that supports both cameras.
Noise reduction was off / Sharpening was off / White Balance was set to auto for both cameras (not to As Shot) / Lens Correction was disabled also
Along with it I have tried to match the exposure of both cameras by increasing the exposure switcher for all Sony shots
Unfortunately they didn’t offer a raw at ISO6400 for Sony, so I didn’t compare for that range
please check these comparisons
http://d.pr/i/xwCX
http://d.pr/i/Yf6i
Both are pretty close up to 800, after that D800 wins. IMHO 🙂
thx for the effort.