DxOMark scores for 5DMkIII out - total score 81, 5DMkII had 79

[sarangiman]

Re: DxO results out for 5D3

I dunno V8Beast, but the original shots are in that thread, if you follow the link.

Still remembering how awesome your car shots are, haha

 

[V8Beast]

Re: DxO results out for 5D3

I dunno V8Beast, but the original shots are in that thread, if you follow the link.

Still remembering how awesome your car shots are, haha

Thanks! Good thing I don't shoot landscapes

 

[sarangiman]

Re: DxO results out for 5D3

jrista et al.: speaking of highlight headroom...

I'll do the DR test I did with my Stouffer transmission wedge w/ my friend's D7000 this Sunday & post back. That'll give us some idea of real world DR. Since I got a number of 11.2 for my 5DIII, I'm somewhat confident this test might at least show if DXO's DR numbers are meaningful.

Also, do you know if the blinking highlights on the LCD display on the 5DII/III indicate blown highlights for a sRGB conversion? Or does it actually show pixels who's RAW values are at saturation?

In my Rizal Bridge shot tonight, I ETTR'd such that I got just very little blinking highlights -- most of those were recovered very well in LR though. So I'm wondering if I could've pushed it even more. A stop probably would've really helped clean up those dirty shadows in the buildings I was talking about...

 

[V8Beast]

Re: DxO results out for 5D3

In my Rizal Bridge shot tonight, I ETTR'd such that I got just very little blinking highlights -- most of those were recovered very well in LR though. So I'm wondering if I could've pushed it even more. A stop probably would've really helped clean up those dirty shadows in the buildings I was talking about...

There's only one way to find out. Keep pushing it to the right A lot of times I'm shooting shiny stuff like chrome or polished aluminum bits inside a black engine compartment with lots of dark nooks and crannies, which is a contrast nightmare. I gets even worse since I insist on using off-camera flashes to balance with the ambient and fill in some of the shadows. I can usually push well beyond the point where the highlights are flashing in the LCD and still recover them in post.

 

[jrista]

Re: DxO results out for 5D3

Apparently it takes Sensorgen for you to finally find someone you can trust , but now that even they say the same thing as myself, the DPR crowd and DxO and so on, I'm glad you finally see.

Um... doesn't Sensorgen just fit the data from DXO to get their numbers? Look at the methodology on Sensorgen's site...

Because at base ISO it has slightly worse read noise vs saturation well size than the 5D2. The difference is so minor though. So it has a touch worse DR down there.

Again, if Sensorgen is just fitting DXO data, and you suspect that DXO is just looking at black/white frames from the cameras, especially with 5DIII's boosted black levels of 2048... do you even trust those read noise values in e- from Sensorgen?

Here's the relevant image from that thread:

Once again, 5DIII gets pwned by D800, though not as badly as in the example LTRLI posted. Also, the sunstars are so much more pleasing on the Nikon image... even w/ a 1970s lens Incidentally that's why the Nikon image is so soft... bad old lens. Regardless, Canon really needs to stop making 8-blade aperture lenses.

Ok, so the image there is more balanced. I definitely see differences in the highlights as well...there are much nicer transitions into the bright highlights reflecting on the water in the Canon image than in the D800 image. In the D800 image, there is a rather rapid and harsh transition into the maximal levels of the reflected lights on the water. The color rendition on the Canon also seems a lot better...the Nikon has a distinct pinkish tint to it, which feels rather unrealistic to me (sodium vapor street lamps are distinctly orange like in the Canon shot.)

I think at the very least the sample here does seem to indicate an earlier cutoff for highlights in Nikon cameras than in Canon cameras, though. The harsh transition from the halo reflection to the actual light source reflection really does seem to indicate that some highlight tonal range was trades for shadow tonal range in Exmor sensors. The Canon shot lends a bit more credence to ETTR with Canon cameras. You can probably eek more shadow detail out with ETTR and end up with a more Nikon-like highlight cutoff, although I still don't see how a Canon could compete with a Sony Exmor sensor in general even with some judiciously tight right histogram shifting.

 

[jrista]

Re: DxO results out for 5D3

jrista et al.: speaking of highlight headroom...

I'll do the DR test I did with my Stouffer transmission wedge w/ my friend's D7000 this Sunday & post back. That'll give us some idea of real world DR. Since I got a number of 11.2 for my 5DIII, I'm somewhat confident this test might at least show if DXO's DR numbers are meaningful.

Looking forward to the results...

Also, do you know if the blinking highlights on the LCD display on the 5DII/III indicate blown highlights for a sRGB conversion? Or does it actually show pixels who's RAW values are at saturation?

The blinking highlights are based on JPEG conversion, be it AdobeRGB or sRGB (whatever you configure). They are fairly inaccurate as well...sometimes only small areas blink, but they turn out to be completely blown and entirely unrecoverable in post...often with that ugly gray halo around them of nearly-blown highlights. Other times, when processing RAW in post, there might not be any blown highlights at all despite the camera blinking half the image. Its kind of a useless feature, although personally whenever I see blinking highlights in the momentary preview after a sequence of shots, I'll bump exposure down by 1/3 to 2/3 of a stop and take another sequence just to be safe (assuming the moment isn't already gone...when it comes to birds in flight, you have to be pretty conservative with your ETTR...which often means none at all.)

In my Rizal Bridge shot tonight, I ETTR'd such that I got just very little blinking highlights -- most of those were recovered very well in LR though. So I'm wondering if I could've pushed it even more. A stop probably would've really helped clean up those dirty shadows in the buildings I was talking about...

I'd try pushing pretty far. I spent a lot of time photographing the moon, as it was the only subject I was really able to photograph for a while during the winter (we get some pretty short days here in Colorado during winter, and at the time my work schedule was pretty bad for afternoon photography.) I've taken photographs of the moon where it appeared almost entirely white, and the camera would blink quite a bit of it...but in post every last scrap of luminance data would be recoverable. Its pretty amazing how you can push highlight data around with a Canon camera. I think the moon might be a fairly unique subject though...it has a pretty low DR in general, so its not quite the same as a step wedge. (And I'm still getting used to my 7D's iFCL metering...when focus and color are taken into account, the histogram distributes differently. There are often very small amounts of disproportionately bright highlights that you can't really see on the histogram, and if you do see them, they might only register as a single red, green, or blue pixel in the corresponding channel very near the far right edge. If you ETTR without looking for that little bit of highlight, you can blow a surprising amount of highlights way beyond recovery.)

 

[Former Nikonian]

Re: DxO results out for 5D3

read but never totally trust any reviews online, especially no-no with reviews just give words or graphs without illustrating samples. Anybody, including you guys and me, can put those words and graphs online without a camera in hand.
Does it make sense when DxO puts D800 over D4 and Markii over Markiii? It's the business (not charity), stupid.
Pls use your commom sense to think. The oddity is very clear, but why they purposefully do that? It's a big question.
I guess DxO is playing a huge Black Humor game to extract those stupid guys, who believe on their nonsense conclusion, out of commom photographers. And they are laughing and enjoying reading what is going on here. Have fun, DxO, you are a true comedy writers.
If you have time, with your own naked eyes and non biased mind, should take looks on: http://www.cameralabs.com/reviews/Canon_EOS_5D_Mark_III/Canon_5D3_vs_Nikon_D800_noise.shtml;
http://www.imaging-resource.com/;
http://www.dslrnewsshooter.com/2012/04/01/video-shootout-nikon-d800-vs-canon-5d-mkiii/;
http://nofilmschool.com/2012/04/5d-mark-iiid800-hands-on-review-part-3/;
Of course you need powerfull PC, especially video card, and high resolution monitor. If not, it would be tough. Have fun.

 

[LetTheRightLensIn]

Re: DxO results out for 5D3

Dude, why didn't you post this earlier ?

Because Fred only just posted his review tonight.

Fortunately, I've only encountered one or two situations in the last 10 years where an image would require the amount of shadow recovery in this test. Also, while a white building casting deep shadows during the middle of the day is a great way to illustrate the need for DR strictly for testing purposes, the way the shot was framed I wouldn't really care to see what's in the shadows of that scene anyways, unless is was some scantily clad blonde standing there. In that case, you'd just compose the image so that the blonde fills up the frame, which would allow reducing the contrast in the scene and exposing accordingly ;D

Yes but what about all the times there turns out to have been a super model lurking in the shadows and you didn't realize it until you got back home and went to process!
That's 90% of my shots!

I'm curious to see how the 1Dx will perform in similar situations. With the 5DII, Canon didn't hold back on the sensor technology, but rather the rest of the feature set (AF, build quality, etc) to distinguish it from the 1DsIII. Perhaps their strategy this time around is to cripple the 5DIII's sensor instead, especially since the AF system between it and the 1Dx are so similar. I wouldn't expect Sony Exmor performance, but maybe the DR will be competitive with the D4.

It will be interesting to see if the 1DX is old tech or new.


(on that other board some guys who have a certain degree of in with Canon got the impression Canon got caught with their pants down, they though Nikon and SOny would dribble sensor stuff out much more slowly and thought they could just sit around on the sensor tech and not bother as much and more just focus on fixing body performance and then they saw the D800 and were like.... oh. Guess they they brought what they had right away instead of dribbling it out slowly. And then some were like we told you to stop being so complacent and we needed to get going sooner. More disturbingly it seemd like some think MP and SNR matter and they have no comprehension or feeling that they need to seriousl address DR. Myself I'd rather they fix up DR more than add more MP if it really had to be one of the other, especially since more DR won't threaten stuff like fps)

(does anyone always notice how so many rumors are like well we see no real need yet for the 7D2 because the D400 hasn't been announced yet or well the 5D3 has been an option for quite a while but we saw no need to rush when we could just sit and wait to see when the D800 arrives. I mean come on man, Canon didn't used to act like that. If you have something bring it and stop waiting until it seems like old hat. Be the old Canon. The leader in all.)

 

[LetTheRightLensIn]

Re: DxO results out for 5D3

read but never totally trust any reviews online, especially no-no with reviews just give words or graphs without illustrating samples. Anybody, including you guys and me, can put those words and graphs online without a camera in hand.
Does it make sense when DxO puts D800 over D4 and Markii over Markiii? It's the business (not charity), stupid.
Pls use your commom sense to think. The oddity is very clear, but why they purposefully do that? It's a big question.
I guess DxO is playing a huge Black Humor game to extract those stupid guys, who believe on their nonsense conclusion, out of commom photographers. And they are laughing and enjoying reading what is going on here. Have fun, DxO, you are a true comedy writers.
If you have time, with your own naked eyes and non biased mind, should take looks on: http://www.cameralabs.com/reviews/Canon_EOS_5D_Mark_III/Canon_5D3_vs_Nikon_D800_noise.shtml;
http://www.imaging-resource.com/;
http://www.dslrnewsshooter.com/2012/04/01/video-shootout-nikon-d800-vs-canon-5d-mkiii/;
http://nofilmschool.com/2012/04/5d-mark-iiid800-hands-on-review-part-3/;
Of course you need powerfull PC, especially video card, and high resolution monitor. If not, it would be tough. Have fun.

I think you missed post #79 on the previous page and http://www.fredmiranda.com/5DIII-D800/index_controlled-tests.html

 

[LetTheRightLensIn]

Re: DxO results out for 5D3

If you want something more cheerful though, how about comparing video instead (now the exposure don't seem to be so well balanced here so it's not quite properly carried out and the canon is so soft it needs to be sharpend in post to make it closer to D800 which would increase 5D3 noise more than shown, but anyway it sort of gives an impression):

Canon 5D Mark II vs. Mark III vs. Nikon D800 - Candlelight: High-ISO


And for kicks a short film done with the 5D3:
Canon 5D Mark III - Copelandia

Don't watch on vimeo though since it compresses both videos too much, you need to log in to vimeo and then download and watch the original uploads.

 

[Aglet]

Re: DxO results out for 5D3

I'm glad Fred got the stuff posted, saves me some time. ;D

NOW will some of you have a little more confidence in some of the numbers published and those silly-looking tests shots with the lens caps on and pushed in post or other methods we may use?..

These simple tests can demonstrate a lot of what goes on at the dark end of the DR range so very easily.

It's been interested to watch the progression of this thread.

Congrats to LTRLI for hitting the DR nail on the head so early and not losing patience in defending his position.

Congrats to those of you who now see and understand what some of us have been going on about for so long. Canon's engineering is not keeping up with its marketing when a $600 consumer grade Nikon can even kick the low ISO pants off a midrange Canon costing nearly 6x as much.

5D3's a much improved, low-light kind of machine.
But it don't do stink for what I need.

The next scrap's gonna be that 24MP, 12-bit little nugget they just announced yesterday. D3200.
I'm REALLY looking forward to how it will perform as it's likely foreshadowing (ooh, good pun) the abilities of a D7100 or D400.

BTW; I now have about 10 Nikon/Nikkor lenses.

 

[sarangiman]

Re: DxO results out for 5D3

jrista:

I definitely see differences in the highlights as well...there are much nicer transitions into the bright highlights reflecting on the water in the Canon image than in the D800 image.

If you look at the original D800 & 5DIII images (follow the link here: http://forums.dpreview.com/forums/read.asp?forum=1021&message=41012886), I don't think that's the case. I think the non-linear processing when lifting shadows makes comparisons of the highlights in the images where shadows been lifted meaningless. In the original images, the highlights on the D800 look just fine... actually, they look less intense.

Any comments, btw, on the link between sensorgen & DXO? Basically, what I'm saying is that you guys are saying that the higher read noise of the 5DIII at low ISO is the reason for the low DR, and I'm saying that the higher read noise calculated by sensorgen is due to DXO's numbers themselves... so it's all a bit of circular logic, no? In order to trust sensorgen, you must first trust DXO...?

 

[V8Beast]

Re: DxO results out for 5D3

Yes but what about all the times there turns out to have been a super model lurking in the shadows and you didn't realize it until you got back home and went to process!
That's 90% of my shots!

You mean to tell me there are super models lurking in the shadows in 90% of your shots? I need to change my specialty to whatever it is that you shoot ;D

 

[Fishnose]

In response to 'former nikonian' and many others here who seem to assume that DxoO consists of a couple of guys playing the fool in their living room and then publishing 'results' to make fun of Canon owners, here's a little text I wrote in another thread on this site, explaining what this is about.
-------------------------
DxO are a software development company that uses extremely precise testing of cameras and lenses as a way to maximise the usefulness of their software (such as DxO Optics Pro) by creating the best possible profiles for each respective model.

It just so happens that they have an interesting by-product of their testing: the test results are a very powerful marketing tool as they attract attention to the DxO brand - and of course they make the most of that by publishing their findings on the web.

Being an independent French company, one can hardly say they are best pals/neighbors with Nikon or Canon or any other manufacturer.
They spend days and days testing each camera body and lens model, using very well-defined test methodology based on scientific principles. They take thousands (yes, THOUSANDS) of photos with each camera body and lens in all possible situations in a controlled lab environment and work only from RAW format so the effects of jpg 'cooking' and other artefacts are not included in their results.

Because their methodology demands precisely controlled and repeatable results, they can't base their findings on 'real world' photography - like going downtown and shooting on the street.

I can't imagine what they would gain by INTENTIONALLY pissing off Canon owners who then are angry at them and refuse to take them or their products seriously.
They can't 'fix' their Canon results to make them look better just to please people like you - they are forced to be objective, otherwise they would be accused of bias and their reputation (and their software) would be destroyed.

Has it occurred to you that Canon (and other manufacturers) themselves never complain or call DxO idiots? Why? Because they know they're not likely to win in the laboratory. They've probably already seen similar results in their own lab tests and know full well what DxO is likely to find.

Canon's decision to release the 5DMkIII with it's present sensor is in the end a marketing decision. And it may turn out to be a bad move.

If you don't like DxO findings, that's your business. If camera and lens manufacturers don't like their findings, that's their business. But all of it generates more traffic for DxO and more exposure for their software products.
--------------------------

 

[smithy]

Now I'm not emotionally attached to this conversation at all, but to be honest, FishNose, with posts like this, you sound like you work for DxO... a full disclosure would be appreciated.

 

[psolberg]

I can't imagine what they would gain by INTENTIONALLY pissing off Canon owners who then are angry at them and refuse to take them or their products seriously.
They can't 'fix' their Canon results to make them look better just to please people like you - they are forced to be objective, otherwise they would be accused of bias and their reputation (and their software) would be destroyed.

exactly. DXO has their parameters which they consider to be the most valuable way to evaluate image quality. Anybody that dissagrees and their work consists of selling in other mediums where their evaluation methos don't apply, then should feel free to disregard their conclusions. DXO tries to appeal to the broadest audience which is why they test the way they do including disregard for 100% pixel crops. If I sold 100% pixel crops for a living, I wouldn't pay much attention to them either. If Canon had pushed the bounds of resolution and DR in their 5D3 sensor then they would score better. It is as simple as that. It isn't DXO's fault one OEM is lagging and nor should they modify their test to help it. You don't see car magazines re-defining acceleration tests to help a slow car that happens to have been popular. That would defeat the entire purpose of the test period.

 

[jpms]

Come on people....be honest to yourself...if the results came out with the 5d3 scoring higher than the D800 I'm sure this thread would be all about how great the 5d3 is and that DxOMark testing is very accurate. I shoot Canon but understand that Canon won't always have the best gear in the market. Stop being in denial and deal with it. This thread is all about cry babies having a difficult time dealing with reality.

 

[jrista]

Re: DxO results out for 5D3

jrista:

I definitely see differences in the highlights as well...there are much nicer transitions into the bright highlights reflecting on the water in the Canon image than in the D800 image.

If you look at the original D800 & 5DIII images (follow the link here: http://forums.dpreview.com/forums/read.asp?forum=1021&message=41012886), I don't think that's the case. I think the non-linear processing when lifting shadows makes comparisons of the highlights in the images where shadows been lifted meaningless. In the original images, the highlights on the D800 look just fine... actually, they look less intense.

They look about the same to me. They are just smaller in scale since the full-size images are scaled down. The lighthouse highlight demonstrates it well...it looks like a smooth falloff, then it suddenly jumps to blown. The canon shot has a continuous falloff that eventually results in a blown highlight.

Any comments, btw, on the link between sensorgen & DXO? Basically, what I'm saying is that you guys are saying that the higher read noise of the 5DIII at low ISO is the reason for the low DR, and I'm saying that the higher read noise calculated by sensorgen is due to DXO's numbers themselves... so it's all a bit of circular logic, no? In order to trust sensorgen, you must first trust DXO...?

I guess the Sensorgen.info page does claim they use DxO measurements for their curve fitting. I thought they also used information from Clarkvision.com, but perhaps that was only in the past and no longer. So, I dunno. I am not really sure that DxO takes the Canon bias offset into account correctly...and if they do not, that very well could account for the higher read noise/lower DR at ISO 100. A 33e- read noise is just friggin insane, in my opinion. I think the average is around 15e-, and newer cameras seem to have less than 10e-.

 

[LetTheRightLensIn]

Re: DxO results out for 5D3

jrista:

I definitely see differences in the highlights as well...there are much nicer transitions into the bright highlights reflecting on the water in the Canon image than in the D800 image.

If you look at the original D800 & 5DIII images (follow the link here: http://forums.dpreview.com/forums/read.asp?forum=1021&message=41012886), I don't think that's the case. I think the non-linear processing when lifting shadows makes comparisons of the highlights in the images where shadows been lifted meaningless. In the original images, the highlights on the D800 look just fine... actually, they look less intense.

They look about the same to me. They are just smaller in scale since the full-size images are scaled down. The lighthouse highlight demonstrates it well...it looks like a smooth falloff, then it suddenly jumps to blown. The canon shot has a continuous falloff that eventually results in a blown highlight.

That can also just go down to a slightly different tone curve applied or a different camera color profile applied, it's hard to say. Anyway the D800 has so much more room at the bottom all you'd need to do is expose it 1/3 stop less or something than in that photo and there you go, highlights fixed.

I am not really sure that DxO takes the Canon bias offset into account correctly...and if they do not, that very well could account for the higher read noise/lower DR at ISO 100. A 33e- read noise is just friggin insane, in my opinion. I think the average is around 15e-, and newer cameras seem to have less than 10e-.

I think they do since I, as well as others, have taken the 1024 to 2048 change into account for the 7D and 5D3 and gotten the same results DxO did. I didn't look at sensorgen at all so whether they converted things properly I don't know.

 

[smithy]

Come on people....be honest to yourself...if the results came out with the 5d3 scoring higher than the D800 I'm sure this thread would be all about how great the 5d3 is and that DxOMark testing is very accurate.

Probably.