Z9 at high ISO by Matt Granger

Matt has uploaded a new video (and files) comparing high ISO performance of the Z9, Z7 II and Z6 II.

The video:
From the comparisons, it looks like my Z9 will not be replacing my D5 for those dimly lit venues where I expect to be exclusively shooting at 6400 and above. In all honesty, I did not expect it to perform as well.

You have to take into account the difference in megapixels. If a Z9 image is downscaled to match the D5 20 megapixels I doubt you will see a difference.

Downscaling eliminates noise. There was a recent video by dpreviewtv where they test the noise vs megapixel hypothesis and found it not to be true.

Alex Permit said:

Matt has uploaded a new video (and files) comparing high ISO performance of the Z9, Z7 II and Z6 II.

The video:

From the comparisons, it looks like my Z9 will not be replacing my D5 for those dimly lit venues where I expect to be exclusively shooting at 6400 and above. In all honesty, I did not expect it to perform as well.

Click to expand...

Yeah, downsampling or not, I doubt that Z 9 will replace my D5 for nighttime stadium scenes. D5 files are simply magic, particularly when processed in Capture One.

I think he messed up the test a bit.

The pictures with the Z7II are shot with the Vivid picture profile, those with the Z6II with the portrait profile. To test it correctly, they should have at least all the same picture profile, but I would avoid them in such tests. For certain ISO values the picture are shot with different aperture value, I think that alters the test too. Starting from ISO 12800 the shots with the Z6II look a bit misfocused to me, he should have used manual focus instead of AF-C.

--
https://www.flickr.com/photos/149089857@N03/

And for z9 he needs to use NX beta as the rest does not treat the z9 file raw as it should at this point.

Alex Permit said:

Matt has uploaded a new video (and files) comparing high ISO performance of the Z9, Z7 II and Z6 II.

The video:

From the comparisons, it looks like my Z9 will not be replacing my D5 for those dimly lit venues where I expect to be exclusively shooting at 6400 and above. In all honesty, I did not expect it to perform as well.

Click to expand...

There is a reason why real flagship models have kept low pixel count for so long.

There are 'basic school-level' errors in what passes for his methodology. These condemn the results. Standard targets? As already pointed out.... where's the downsizing and also decent exposures i.e. not skewed as has been reported already on a RAW or two of these images...

Is not 1/6400 different from 1/4000 ? It is noted that at least one Z9 image was taken at 1/6400 vs Z7 II at 1/4000..... this report is a minefield of confounding errors trying to compare sensors that are likely to only differ slightly ie the 45mp Z7 vs 45mp Z9.

Maybe, maybe just. this video might be yet more clickbait in a youtuber channel capitalizing on early access to a prerelease camera. And said camera is running beta firmware, and there has been private feedback from beta testers that Nikon requested no detailed comments about image quality to the media after the Z9 launch.

Any first hand statements about Z9 IQ are subject to later, when we see independent measurements grounded in the basic scientific methods - comparing the Z9 against other 45mp cameras ie Z7, and also the flagship DSLRs.

And we know who can and cannot be trusted with respect to reliable tests of camera sensors.

I opened these RAWs in LR Classic 11.0 and the Z9 rendered differently to what I see in Granger's video. The colour noise of the Z9 files at 12,800 ISO (which I use a a standard these days) is the same as the Z7 II when using LR. In Granger's testing I'm not sure what colour noise reduction settings he was using, but the Z9 files had more colour noise in his video. I'm not getting that in LR with colour noise reduction set to 25. The 12800 ISO shot with the Z9 is also underexposed by about .5 stops.

Maybe the Z9 is a fraction of a stop behind the Z7 II in Luma noise. That's to be expected with a stacked sensor. Electronics is the enemy of noise.

There were predictions that the Z9 sensor would not be able to beat the state of the art in all areas of performance. Nikon has designed a sensor for ultimate readout speed. It was predictable that other performance specs might not be state of the art. See Here for instance:

https://www.dpreview.com/forums/post/64969454

"...(the) Z9 sensor is unlikely to be Sony derived. Their (Nikon's) problem is that the A1 chucks in just about everything that Sony SS has, so even to beat it in one specific specification is going to be a big ask. Given the stacked sensor that Nikon has announced, I suspect that the one specification that it will beat it at is readout rate, which for a mirrorless will possibly contribute to close to DSLR VF dynamics, and possibly allow simultaneous readout of main image and VF data streams. But that will put a huge load on the peripheral circuitry, which might end up with this being a (even) more expensive solution than Sony's."

My impression is that people are expecting too much of this new sensor over and above its class leading scan rate.

I shoot sports with the D5 as well and have some concern in the back of my mind over noise levels of the Z9 - particularly as almost all of my photography is around 6400-8000 ISO on 2.8 lenses.

However, I know that in practice, printing an image at the same size as the D5 images I print now or down sampling images to the resolutions I usually deliver, noise at the ISOs I generally work in shouldn't be an issue (assuming the Z9 noise is the same as the D850).

Here's an illustration using the dpreview image comparison tool:


It includes the D5, D850, Sony A1 and Canon 1Dx mkIII. When normalized in size for print (8MP), the noise is nearly identical for all. The higher resolution cameras just produce an image that looks sharper. And this is worst case scenario - no noise reduction at all. Like you, I use CaptureOne, and the finished results would look way better than this.

Still, I really prefer (and will bend over backward when needed) to stay below 12,800. That's for sure when the three D5s all get their fixed 1.4, 1.8 and 2.0 primes attached (I'm looking at you Mandalay Bay Convention Center - you're a pit!)

---
Micheal

Alex Permit said:

Matt has uploaded a new video (and files) comparing high ISO performance of the Z9, Z7 II and Z6 II.

The video:

From the comparisons, it looks like my Z9 will not be replacing my D5 for those dimly lit venues where I expect to be exclusively shooting at 6400 and above. In all honesty, I did not expect it to perform as well.

Click to expand...

He doesn't seem to really know what he's doing there.

Anyone who has ever shot a D850 or a Z7II (or any high MP camera) in RAW and processed very high ISO files will know that they can look very noisy until the RAW processing software has done it's work on the noise! There is no way that you can get the sort of look at 12800 or 25600 that he showed with the Z7II files without very effective late generation NR!

It gets increasingly obvious as the demo progresses that the Z9 shots are being shown without any effective NR applied with obvious chroma noise showing, increasingly evident above 6400, whereas the Z6Ii and Z7II samples are shown with chroma noise effectively eliminated. That's purely down to post processing differences.

So I think that video ends up looking quite misleading.

If the Z9 can do what i saw there with almost no NR then I expect the high ISO results, especially when well processed (using the most effective 3rd party software such as DXO Photolab etc), willl come out looking great.

Frank

Atahualpa said:

You have to take into account the difference in megapixels. If a Z9 image is downscaled to match the D5 20 megapixels I doubt you will see a difference.

Downscaling eliminates noise. There was a recent video by dpreviewtv where they test the noise vs megapixel hypothesis and found it not to be true.

Click to expand...

I did, in a way.

When I first got my Z7, I compared its very high ISO performace to my D5. I downrezed the Z7 shots when doing the comparison. The D5 was noticably less noisy.

Photons to photos appears to verify this finding. The D5 tests one stop better than the Z7 at or above ISO 3200.

https://www.photonstophotos.net/Charts/PDR.htm#Nikon D5,Nikon Z

I figure, by extension, if the Z9 is slighly worse than the Z7, than the D5 will be better than the Z9 as well.

now says "Video unavailable"

"This video is private."

One has to ask if Nikon told him to remove his error riddled clickbait

From Matt Granger's website :

"Change of plans...

Hi everyone - I have removed my previous video on Nikon Z high ISO comparison.

Thank you to the viewers that pointed out there were some technical errors in the test (like not matching settings), and it seems the way the files came through on YouTube was also inaccurate. It ALSO seems like different software are applying different automatic settings to some files and not others.

Until I figure out the best way to present only accurate test findings, I will keep it offline. It maybe that I wait until the Z9 is on sale - so it is no longer pre-production, and the RAW decoder is available for all software.

The error was mine - I was rushing to do the test - and I don't want to present bad information. Thanks for your understanding."

FrankG said:

Alex Permit said:

Matt has uploaded a new video (and files) comparing high ISO performance of the Z9, Z7 II and Z6 II.

The video:

From the comparisons, it looks like my Z9 will not be replacing my D5 for those dimly lit venues where I expect to be exclusively shooting at 6400 and above. In all honesty, I did not expect it to perform as well.

Click to expand...

He doesn't seem to really know what he's doing there.

Anyone who has ever shot a D850 or a Z7II (or any high MP camera) in RAW and processed very high ISO files will know that they can look very noisy until the RAW processing software has done it's work on the noise! There is no way that you can get the sort of look at 12800 or 25600 that he showed with the Z7II files without very effective late generation NR!

It gets increasingly obvious as the demo progresses that the Z9 shots are being shown without any effective NR applied with obvious chroma noise showing, increasingly evident above 6400, whereas the Z6Ii and Z7II samples are shown with chroma noise effectively eliminated. That's purely down to post processing differences.

So I think that video ends up looking quite misleading.

If the Z9 can do what i saw there with almost no NR then I expect the high ISO results, especially when well processed (using the most effective 3rd party software such as DXO Photolab etc), willl come out looking great.

Frank

Click to expand...

High ISO comparisons in RAW is essentially a comparison of the RAW converter's interpretation of the particular RAW file (and perhaps an indication of the user's skill level).

If one starts with what the camera outputs, in terms of camera defaults, one gets a comparison something like this:





Rather than RAW:

mapgraphs said:

FrankG said:

Alex Permit said:

Matt has uploaded a new video (and files) comparing high ISO performance of the Z9, Z7 II and Z6 II.

The video:

From the comparisons, it looks like my Z9 will not be replacing my D5 for those dimly lit venues where I expect to be exclusively shooting at 6400 and above. In all honesty, I did not expect it to perform as well.

Click to expand...

He doesn't seem to really know what he's doing there.

Anyone who has ever shot a D850 or a Z7II (or any high MP camera) in RAW and processed very high ISO files will know that they can look very noisy until the RAW processing software has done it's work on the noise! There is no way that you can get the sort of look at 12800 or 25600 that he showed with the Z7II files without very effective late generation NR!

It gets increasingly obvious as the demo progresses that the Z9 shots are being shown without any effective NR applied with obvious chroma noise showing, increasingly evident above 6400, whereas the Z6Ii and Z7II samples are shown with chroma noise effectively eliminated. That's purely down to post processing differences.

So I think that video ends up looking quite misleading.

If the Z9 can do what i saw there with almost no NR then I expect the high ISO results, especially when well processed (using the most effective 3rd party software such as DXO Photolab etc), willl come out looking great.

Frank

Click to expand...

High ISO comparisons in RAW is essentially a comparison of the RAW converter's interpretation of the particular RAW file (and perhaps an indication of the user's skill level).

If one starts with what the camera outputs, in terms of camera defaults, one gets a comparison something like this:

Rather than RAW:

Click to expand...

Also compare at similar resolutions.

Alex Permit said:

mapgraphs said:

FrankG said:

He doesn't seem to really know what he's doing there.

Anyone who has ever shot a D850 or a Z7II (or any high MP camera) in RAW and processed very high ISO files will know that they can look very noisy until the RAW processing software has done it's work on the noise! There is no way that you can get the sort of look at 12800 or 25600 that he showed with the Z7II files without very effective late generation NR!

It gets increasingly obvious as the demo progresses that the Z9 shots are being shown without any effective NR applied with obvious chroma noise showing, increasingly evident above 6400, whereas the Z6Ii and Z7II samples are shown with chroma noise effectively eliminated. That's purely down to post processing differences.

So I think that video ends up looking quite misleading.

If the Z9 can do what i saw there with almost no NR then I expect the high ISO results, especially when well processed (using the most effective 3rd party software such as DXO Photolab etc), willl come out looking great.

Frank

Click to expand...

High ISO comparisons in RAW is essentially a comparison of the RAW converter's interpretation of the particular RAW file (and perhaps an indication of the user's skill level).

If one starts with what the camera outputs, in terms of camera defaults, one gets a comparison something like this:

Rather than RAW:

Click to expand...

Also compare at similar resolutions.

Click to expand...

What these examples show is basic affects of noise from smaller number of photons captured due to twice the number of pixels (-half the number of photons captured) compared with noise sources. You can't get around it. Small improvements and perhaps on-chip noise reduction. The breakthrough is in the backside sensor in terms of signal/noise. The higher resolution should offset any increase in noise. A better comparison would be with a good noise reduction software such as DxO Prime Topaz AI. Noise is going to be present irreguardless at high ISO. Can it be minimized with software? More consistant focus is the real breakthrough in the Z9 which gives perfect infocus images.

No clue why

Glad to see he has some integrity.

I peaked at a video posted over in the Sony Alpha Rumors site of (gasp) Tony Northrup reviewing the 70-200 lens choices out there from Fujifilm, Nikon, Canon, and Sony. There's a point in the video where he is still putting forth the false information that the Z9 has an AA filter.

Spoiler alert... the Sony 70-200 wins Tony's comparison. I was surprised he actually gave 2nd place to Nikon. The Canon was third.

Philip Eihuyar said:

Alex Permit said:

mapgraphs said:

FrankG said:

He doesn't seem to really know what he's doing there.

Anyone who has ever shot a D850 or a Z7II (or any high MP camera) in RAW and processed very high ISO files will know that they can look very noisy until the RAW processing software has done it's work on the noise! There is no way that you can get the sort of look at 12800 or 25600 that he showed with the Z7II files without very effective late generation NR!

It gets increasingly obvious as the demo progresses that the Z9 shots are being shown without any effective NR applied with obvious chroma noise showing, increasingly evident above 6400, whereas the Z6Ii and Z7II samples are shown with chroma noise effectively eliminated. That's purely down to post processing differences.

So I think that video ends up looking quite misleading.

If the Z9 can do what i saw there with almost no NR then I expect the high ISO results, especially when well processed (using the most effective 3rd party software such as DXO Photolab etc), willl come out looking great.

Frank

Click to expand...

High ISO comparisons in RAW is essentially a comparison of the RAW converter's interpretation of the particular RAW file (and perhaps an indication of the user's skill level).

If one starts with what the camera outputs, in terms of camera defaults, one gets a comparison something like this:

Rather than RAW:

Click to expand...

Also compare at similar resolutions.

Click to expand...

What these examples show is basic affects of noise from smaller number of photons captured due to twice the number of pixels (-half the number of photons captured) compared with noise sources. You can't get around it. Small improvements and perhaps on-chip noise reduction. The breakthrough is in the backside sensor in terms of signal/noise. The higher resolution should offset any increase in noise. A better comparison would be with a good noise reduction software such as DxO Prime Topaz AI. Noise is going to be present irreguardless at high ISO. Can it be minimized with software? More consistant focus is the real breakthrough in the Z9 which gives perfect infocus images.

Click to expand...

No it does not. Assuming the same technology, you capture the same number of photons in the entire image or a comparable crop (as shown above). Twice the number of pixels, each captures half the number of photons. Total photons captured over the entire image is the same. 2 x 1/2 = 1.

The Nikon D5 does better than the Z7 at high ISO because it uses different technology than the Z7. This technology is optimized to capture the most photons with the least amount of noise noise in low light situations. It does so by sacrificing relative performance at low ISO's.

Valladolid said:

From Matt Granger's website :

"Change of plans...

Hi everyone - I have removed my previous video on Nikon Z high ISO comparison.

Thank you to the viewers that pointed out there were some technical errors in the test (like not matching settings), and it seems the way the files came through on YouTube was also inaccurate. It ALSO seems like different software are applying different automatic settings to some files and not others.

Until I figure out the best way to present only accurate test findings, I will keep it offline. It maybe that I wait until the Z9 is on sale - so it is no longer pre-production, and the RAW decoder is available for all software.

The error was mine - I was rushing to do the test - and I don't want to present bad information. Thanks for your understanding."

Click to expand...

Respect.