DxO - 70-200 IS Comparison - Is the original really better?

I searched a few times and didn't see this really discussed here so I wanted to see how this was taken by the community.

Link here

Before I purchased my IS II, I read TONS of reviews which all seemed to indicate that while the orginal was a great lens, the new version is better is most categories.

This DxO review seems to contradict those claims.

Any thoughts?

tats wrote in post #13419096
I searched a few times and didn't see this really discussed here so I wanted to see how this was taken by the community.

Link here

Before I purchased my IS II, I read TONS of reviews which all seemed to indicate that while the orginal was a great lens, the new version is better is most categories.

This DxO review seems to contradict those claims.

Any thoughts?

DxO ratings consistently do not match reality on a number of fronts. I have used some of the cameras (OK, several of the cameras) that they have rated and I can assure you that they are very, very, very wrong on many occasions. Whatever it is that they like to measure sure as hell has nothing to do with the kinds of things that actual, real photographers care about.

As for their measurement of the 70-200/2.8 IS II. Well, huge surprise, they are wrong again.

And I should mention, I too owned both the 70-200/2.8 IS and the 70-200/2.8 IS II. At the point that I had both in my posession I did of course test them against each other.

My assessment was this. For most prints, the actual differences between these two lenses is almost negligible at almost any normal sizes. At a 1:1 view from a body like the 1D Mark IV a difference can be seen, the Mark II is sharper.

Was the switch really worth the extra cost to me.....probably not really in prints. But I know me....those sharper 1:1 views would bug me had I not kept the Mark II.

I don't know anything about their lens testing, but their sensor ratings are based entirely on various measurements (i.e. lab data). It's good because they are objective, and their raw data is accurate. It's bad because the ratings they've come up with have nothing to do with reality.

I find my mk2 better in every way apart from price.

JeffreyG wrote in post #13419334
And I should mention, I too owned both the 70-200/2.8 IS and the 70-200/2.8 IS II. At the point that I had both in my posession I did of course test them against each other.

My assessment was this. For most prints, the actual differences between these two lenses is almost negligible at almost any normal sizes. At a 1:1 view from a body like the 1D Mark IV a difference can be seen, the Mark II is sharper.

Was the switch really worth the extra cost to me.....probably not really in prints. But I know me....those sharper 1:1 views would bug me had I not kept the Mark II.

Agree. I had both for a while. My ver I was real good. I would guess not much change till 11x14 this using 5d but I got rid of ver I and my 135L for the ver II.

tats wrote in post #13419096
I searched a few times and didn't see this really discussed here so I wanted to see how this was taken by the community.

Link here

Before I purchased my IS II, I read TONS of reviews which all seemed to indicate that while the orginal was a great lens, the new version is better is most categories.

This DxO review seems to contradict those claims.

Any thoughts?

This is has been on DPR for the last few days.
Many find their sensor tests to be pretty decent, the individual graphs, not the overall camera ratings, but their lens results to all too often be simply bizarre even on the individual charts (their overall lens ratings just beyond beyond absurd, far worse than even the overall camera ratings, they seem to like to pick out the single sharpest f-stop and focal length at the single sharpest part of the frame and rate the entire lens by that alone?).

Try these on for size:

at 300 f/5.6:
70-300 IS non-L [much better than] 70-300L IS [Better than] 300 f/4L IS

(hmmm)

or 16-35 2.8L II:
f/2.8 corners [much sharper than] f/5.6 corners

(it is true, that if you refocus for edges they don't necessarily peak at f/8+ as some think and many times will peak at not much beyond where the center does but I'd be stunned that an ultra-wide zoom would peak corners at f/2.8 over f/5.6!!!!)

or as you found (plus adding more) 200mm f/2.8:
70-200 2.8L IS [Better than] 70-200 2.8L non-IS [Better than] 70-200 2.8L IS II

(the exact reverse of by far most user reports)

or:
70-200 2.8 IS [easily beating over the entire range] 70-300L IS

(never mind that my 70-300L beats my 70-200 f/4 IS near 70mm and 200mm and that lens itself beat my 70-200 2.8 non-IS over the entire range and that beat a borrowed 70-200 2.8 IS, easily, at 200mm and was same or better across the rest of the way)

and there are so many more, I think

wombatHorror wrote in post #13419968
This is has been on DPR for the last few days.

or 16-35 2.8L II:
f/2.8 corners [much sharper than] f/5.6 corners

I don't know if DxO is simply unable to explain themselves in a way that humans can understand, or if they are totally stupid. It's like, aliens vs. imbeciles, and either explanation is plausible.

For the record, here is my 16-35/2.8 II corners at f/2.8 and then f/5.6. Does anyone agree that the second image (f/5.6) is better than f/2.8?

Clearly, DxO is not worth reading, ever.

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

JeffreyG wrote in post #13420050
I don't know if DxO is simply unable to explain themselves in a way that humans can understand, or if they are totally stupid. It's like, aliens vs. imbeciles, and either explanation is plausible.

For the record, here is my 16-35/2.8 II corners at f/2.8 and then f/5.6. Does anyone agree that the second image (f/5.6) is better than f/2.8?

Clearly, DxO is not worth reading, ever.

Bang on!

You know, your comments on that 16-35 II made me go look it up on DxO. Not only are they wrong about the lens for corner performance vs. aperture, they are also wrong relative to focal length. They show the lens wide open f/2.8 is capable of ~50 lp/mm in the corner at 16mm but only ~20 lp/mm in the corner at 35mm.

For starters, this should immediately make most people think "Hmmmmm.....an UWA zoom with better corners at the ultra wide end? Kind of hard to believe."

So then a person like me might check it. And (results below) anyone with this lens and a pair of eyeballs can see that once again DxO is wrong. I don't know about you, but when a website claims infalliable highly technical performance data and is then demonstrably wrong over and over, well, I tend to not believe anything on the site.

It's not like DxO was wrong once. They said the D90 had less noise than the 5D. I tested my 5D with my brother-in-law's D90 and we were both left scratching our heads. Then there was the non-sensical 70-200/2.8 IS II results and now these ridiculous 16-35 II results.

We should have a contest to find the most ridiculous and unsupportable DxO claims.

Anyway - here's the f/2.8 corners from my 16-35 II at 16mm and then 35mm. DxO says the 16mm corner is better. What do you think?

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

DXO is even worse then Ken Rockwell KR, at least we know, he just writes things to raise controversy.

Seriously, I think the guys at DXO must be stoned all the time. I doubt any of them are actually photographers since if they were they would realize that their conclusions have little to do with reality... the image.

There are other useless sites out there but I rate DXO pretty much at the bottom of review sites.

bohdank wrote in post #13422335
DXO is even worse then Ken Rockwell KR, at least we know, he just writes things to raise controversy.

Seriously, I think the guys at DXO must be stoned all the time. I doubt any of them are actually photographers since if they were they would realize that their conclusions have little to do with reality... the image.

There are other useless sites out there but I rate DXO pretty much at the bottom of review sites.

It makes one wonder about their DxO Optics Pro software. How can it correct for lens aberrations if this is the sort of result they get when they measure lenses? Some seem to say DxO OP works ok so maybe they have a different team on it or something and just tossed off the lens tests to some summer interns or something. It's almost like they are relying on camera AF for the tests or only doing on or two manual focus attempts. Although MTF is actually a lot trickier to measure than distortion and CA, for MTF comparisons you need utterly exact focus while for distortion you can be at any even semi-reasonable tolerance level and just need flat alignment.

Ignoring any of their overall sensor ratings, I do find their sensor plots to mostly pretty reasonable results. The Canon sensor data certainly matches pretty well for the most part to the independent tests that I and others have done. But yeah their overall sensor and camera rankings are just kinda silly.

JeffreyG wrote in post #13421710
It's not like DxO was wrong once. They said the D90 had less noise than the 5D. I tested my 5D with my brother-in-law's D90 and we were both left scratching our heads.

hmm but
It seems to show low light rating 1368 for 5D and 977 for D90 though.
It also shows 5D better at SNR.

It does show better lowest ISO DR for the D90, but that actually isn't necessarily shocking, if the D90 was the first SONY column ADC sensor (not sure, but I think it was).

OTOH it does have a very curious kink in the DR plot for the D90, was the D90 the one Nikon that was found to be doing something funky to the RAW I think???

Canon probably didn't slip the DXO folks the "good review" bonus.

Apart from attacking DxO Labs, Lloyd Chambers sent back several (iirc five) 70-200 IS Mark II lenses before he received one that was sharp over the field. This suggests that there's at least some possibility of a quality control problem.

On the other hand...
I bought a 70-200 Mk II from the first batch and it on my 7D. I mounted it piggy-back on my telescope. Manually focussed it resolves stars to near pixel level from about f/3.2 to f/6.3 suggesting that it's essentially diffraction limited. There was some evidence of violet fringing - this might have been very weak CA or sensor overload.

I also noted that when disabled, the IS sometimes switched off in a strange state. This lead to apparently weird performance. It hardly matters since this lens is supposed to work fine when tripod mounted with the IS turned on.