GH4, G9, RX10 IV comparison at ISO 6400

I thought I would provide a few test shots where I was comparing ISO 6400 on the GH4, G9, and RX10 IV.  This was not a scientific experiment as you can tell from the laundry room shots.  However, I was trying to test in fluorescent lighting conditions to mimic some of the conditions I encounter on a regular basis.  These are the best of several shots at each condition as the lighting causes variable exposure.

I am not sure how to upload full res images (maybe it did it) with EXIF.  However, what I was surprised at is that when I look at these on the 60 inch HDTV the G9 is no better than the GH4 and RX10 IV.  This might be considered pixel peeping, however I expected a 1 stop improvement based on reviews and the dpreview scene comparison.  In reality, to my untrained eye I do not see much difference, which tells me that panasonic really needs to improve their sensor as the RX10 IV is a 1 inch sensor.

I thought this might help others thinking of upgrading from the GH4 to the G9.  In general, the G9 has some handling improvements and the focus is a bit faster.  However, if you are thinking of upgrading don't expect too much image improvement.  I am sure under the right conditions there is improvement.  However, I shoot my kids at indoor sporting events and I was hoping to gain a stop so I could freeze action better.  That does not appear to be the case.

G9, f3.5, ISO 6400

GH4, f3.5, ISO 6400

RX10 IV

I noticed you used a faster shutter speed on the G9, which probably negates some of the advantage.

Also the GH4 has a 16MP. Did you normalize the pictures to the same size? If you are comparing per pixel level, the G9 is 20MP and that gives it another hit.

RX10 IV does have BSI, which might help it up to 1/3 of a stop (theoretically).

Them being JPEGs makes it a bit more difficult to do a pure sensor comparison (as the amounts of sharpening and noise reduction applied varies and plays a factor in JPEGs).

However, if you direct your attention to the Clorox logo on the clog remover bottle and where it says "From the makers of Clorox products" there is a very clear winner. This was the only part in the picture that is somewhat dark and has some detail that can be used to compare.

ThePhilosopher wrote:

d3xmeister wrote:

- you are not really comparing sensors but jpeg engines

Technically it is both since the jpeg engine uses the output from the sensor. I purposefully used jpeg as 95% of the time I do not use RAW.

I think you are underestimating the impact of the jpg engine and the camera's high ISO NR algorithms to a final JPG.

You could have 2 sensors (lets call it Sensor A and Sensor B). Let's say "A" is theoretically 2 stops better high ISO performance then "B". If the jpg engine in "A" mushes the high ISO jpg and "B" does a better job the final jpg output will look better for B.

If you did the RAW processing yourself you'd find "A" 2 stops better.

If you use jpg 95% of the time then you really are evaluating the jpg engine IMO.  I've read many reviews here on cameras where the jpg engine destroys high ISO shots but the RAW files are completely acceptable.

I do miss my GH4. I wish  I could have afforded to keep it when I got my E-M1 MKii. If I could afford a backup camera I think I would pick up a good used GH4 over the G9 as it fitted my hand like a glove and was a joy to use.

To my eyes there is certainly more detail in the G9 shot than the others (you can see it most clearly in the text on the label).

I try not to pixel-peep JPEGs, though, since a) I would always be disappointed and b) for me, JPEGs are only for quick sharing, not for photos that I really care about. Obviously you should tailor your tests for your specific usage, though, and if you mainly prefer the output of JPEGs, that is what matters.

However, the fact that the G9 is at 1/250s instead of 1/200s means that you are giving up 1/3 stop of light. While 1/3 stop is not particularly noticeable, when comparing sensors that are nominally 2/3 to 1 stop apart, it obviously degrades that difference significantly.

As you say, your test was not intended to be scientific, but an easy way to improve the consistency and validity of your results is to just put the cameras in M mode and shoot with the exact same settings. If the brightness is different, you can match that in post to see the ramifications on image quality. At least that way you know you are giving the cameras the same amount of light. No need for guesswork, then.

Upon closer inspection, I'm not only noticing more detail in the G9 shot, but the biggest difference is actually in the interactions between the noise reduction and the sharpening. The RX10 has quite aggressive sharpening mixed in with its noise reduction, such that flat smooth areas of the image (like the walls behind the bottles, or in the sink to the right of the Clorox) have a "gravelly" texture, where lumps of smoothed texture with sharp edges are formed. The G9 is cleaner here.

I'd argue that the G9's noise reduction is actually too strong for my tastes. That's a general criticism I have of Panasonic's JPEG engine. But the fact that it retains the same or better detail than the RX10 while having noticeably smoother flat areas is where that 2/3 EV of theoretical difference is showing up.

Obviously, the question is whether 2/3 EV is different enough to merit using an ILC instead of a fixed-lens camera, if everything else is up to your satisfaction. Only you can answer that question!

Androole wrote:

Upon closer inspection, I'm not only noticing more detail in the G9 shot, but the biggest difference is actually in the interactions between the noise reduction and the sharpening. The RX10 has quite aggressive sharpening mixed in with its noise reduction, such that flat smooth areas of the image (like the walls behind the bottles, or in the sink to the right of the Clorox) have a "gravelly" texture, where lumps of smoothed texture with sharp edges are formed. The G9 is cleaner here.

I'd argue that the G9's noise reduction is actually too strong for my tastes. That's a general criticism I have of Panasonic's JPEG engine. But the fact that it retains the same or better detail than the RX10 while having noticeably smoother flat areas is where that 2/3 EV of theoretical difference is showing up.

Obviously, the question is whether 2/3 EV is different enough to merit using an ILC instead of a fixed-lens camera, if everything else is up to your satisfaction. Only you can answer that question!

You have to factor in the lens aperture too. The RX10 has a maximum aperture of 2.4. You'd pay a lot more for a G9 with such a fast lens. So you should be able to offset the smaller sensor of the RX10 with a lower ISO. In this case you have (understandably) kept the same aperture. You could, of course have fitted an much faster prime on the G9 and give it the ISO advantage.

bcr5784 wrote:

Androole wrote:

Upon closer inspection, I'm not only noticing more detail in the G9 shot, but the biggest difference is actually in the interactions between the noise reduction and the sharpening. The RX10 has quite aggressive sharpening mixed in with its noise reduction, such that flat smooth areas of the image (like the walls behind the bottles, or in the sink to the right of the Clorox) have a "gravelly" texture, where lumps of smoothed texture with sharp edges are formed. The G9 is cleaner here.

I'd argue that the G9's noise reduction is actually too strong for my tastes. That's a general criticism I have of Panasonic's JPEG engine. But the fact that it retains the same or better detail than the RX10 while having noticeably smoother flat areas is where that 2/3 EV of theoretical difference is showing up.

Obviously, the question is whether 2/3 EV is different enough to merit using an ILC instead of a fixed-lens camera, if everything else is up to your satisfaction. Only you can answer that question!

You have to factor in the lens aperture too. The RX10 has a maximum aperture of 2.4. You'd pay a lot more for a G9 with such a fast lens. So you should be able to offset the smaller sensor of the RX10 with a lower ISO. In this case you have (understandably) kept the same aperture. You could, of course have fitted an much faster prime on the G9 and give it the ISO advantage.

Yes, of course.

For all intents and purposes, the RX10 IV is basically the same as a G85 or G9 kit with a 12-60/f3.5-5.6 (or 14-140/f3.5-5.6) and a 100-300/f4-5.6.

The obviously advantage is that you don't need to change lenses with the RX10 IV to achieve that entire range.

The obvious disadvantage is that you can never use a lens that is wider than 24mm, and the M4/3 camera can use fast f1.7 (or even faster) primes that will give it 2 full stops of image quality advantage in low light.

There is no bad option there, they are merely different compromises. The right choice depends solely on the user.

I think the results speak for themselves. The G9 is still the best, then the GH4 and the RX10IV. If you can't see the lack of details then I think you're looking in the wrong places. It's pretty obvious on the blue caps that you can see the ridges on the G9 whereas on the RX10IV it becomes mush. Now that is a result of NR from a cooked jpg, but that seems to be what the OP wants to compare. You can even look at the droplets inside the bottle to see the detail there or just the text/lettering on several bottles.

Now I wouldn't say there is a drastic difference, but there is a difference nonetheless. The shutter speed difference doesn't favor the G9. I wouldn't shoot past 3200 ISO unless need be.

ThePhilosopher wrote:

Androole wrote:

To my eyes there is certainly more detail in the G9 shot than the others (you can see it most clearly in the text on the label).

What I see (my untrained eyes) in most of the images is just trade offs between keeping detail versus NR. However, no real big difference in sensor capability. Hence, if one uses RAW (which I did some basic tests) and the tuning is done right one will get equivalent images. I expected the G9 to be a bit better than it is based on the reviews I read and that was all I was really trying to let others know in case they are considering an upgrade from the GH4. There are other tech features like the 6K photos, additional function buttons, etc.. that obviously make a difference, but image quality will be similar.

I try not to pixel-peep JPEGs, though, since a) I would always be disappointed and b) for me, JPEGs are only for quick sharing, not for photos that I really care about. Obviously you should tailor your tests for your specific usage, though, and if you mainly prefer the output of JPEGs, that is what matters.

The problem is if you do a slide shows on a large screen HDTV sometimes the image can look really bad. Yes, it can be cleaned up with processing, but I was surprised it was so similar.

Yes, RAW is used for my photos I really care about or in challenging conditions in which I am unsure if I can get "exposure" right. However, I have found that modern JPG engines are pretty good if the "exposure" is right for my 90% casual photos.

However, the fact that the G9 is at 1/250s instead of 1/200s means that you are giving up 1/3 stop of light. While 1/3 stop is not particularly noticeable, when comparing sensors that are nominally 2/3 to 1 stop apart, it obviously degrades that difference significantly.

True. But I also have shots at 1/200 and they look identical.

If the shots at 1/200s look identical to the shots at 1/250s, the only conclusion can be that small differences in image quality - real though they absolutely are - are not important to you.

You are your own biggest critic and the only reviewer that matters for your photography. If you can't notice the quality difference, that's an important observation and should usefully inform the future choices that you make with your gear.

But make no mistake, there is a 1/3 stop difference between exposures made with 1/200s and 1/250s. It is exactly the same as saying that ISO 6400 and ISO 8000 look identical.

As you say, your test was not intended to be scientific, but an easy way to improve the consistency and validity of your results is to just put the cameras in M mode and shoot with the exact same settings. If the brightness is different, you can match that in post to see the ramifications on image quality. At least that way you know you are giving the cameras the same amount of light. No need for guesswork, then.

Absolutely. However, for my quick test I was also trying to see what SS the cameras would chose. I typically shoot in A mode and let the camera choose the appropriate shutter, hence my initial intent in this test was really just to see how the cameras behaved in my typical shooting method. Doing my test this way helps me to confirm that the ISO being reported is actually comparable. For some cameras you find there is a large difference, which typically leads me to believe a bit of cheating on the ISO reporting. For instance, if ISO 6400 and A3.5 gave a SS of 100 you might conclude that ISO 6400 is really ISO 3200 or ISO 1600. What matters to me is given an aperture of 3.5 what is the maximum shutter speed I can achieve with acceptable noise level. In fairness I was also testing a few other lenses like the adapted Tamron 16-300 to see if the reported fstop with the adapter was correct.

For a true comparison (in my opinion) it becomes more complicated then just using M mode with the same "settings". For instance, if you want to get the same DOF so you can really compare that the image will be identical, then you need to adjust the aperture. Of course this leads to needing a different SS to get the same light. Is equivalence the same settings or the same image? What about my comment above in which camera manufacturers cheat on ISO? Ultimately for me, it comes down to how the camera behaves in real conditions.

ThePhilosopher wrote:

JakeJY wrote:

I noticed you used a faster shutter speed on the G9, which probably negates some of the advantage.

Yes, I was in A mode. The SS was fluctuating between 200-250. I have several images at 200 that look the same. However, those I had different settings for the picture profile so wanted to match that up. My conclusion was exactly the same.

Also the GH4 has a 16MP. Did you normalize the pictures to the same size? If you are comparing per pixel level, the G9 is 20MP and that gives it another hit.

Nope, these are SOOC. However, to me the 20 should be able to give equivalent or better results SOOC.

RX10 IV does have BSI, which might help it up to 1/3 of a stop (theoretically).

So far in casual use I have found the RX10 IV to 98% give equivalent performance to the GH4. The BSI clearly performs better as it enables the smaller sensor to behave similarly to the larger sensor. Every torture test I throw at it and compare to the GH4 the RX10 IV seems to either come out equal or win. If I can find the image I'll post my comparison of a test board I made between it, the GH4, and the A7III. I was surprised at how well it did.

For what I shoot I need about another 2 stops to be happy as I need to freeze action. There isn't a camera available today with the right size that meets my needs. Clearly I could go FF, but then I need a larger lens for zoom. I am guessing if Sony releases a high end APSC with the BSI sensor I will finally be able to hit my performance/size specification. However, I am guessing they don't do this until September and I am not a big fan of their APSC form factor.

Them being JPEGs makes it a bit more difficult to do a pure sensor comparison (as the amounts of sharpening and noise reduction applied varies and plays a factor in JPEGs).

However, if you direct your attention to the Clorox logo on the clog remover bottle and where it says "From the makers of Clorox products" there is a very clear winner. This was the only part in the picture that is somewhat dark and has some detail that can be used to compare.

I was looking at general noise profile as it mimics what I see for skin tones and general quality under some of the lighting conditions I shoot in. Believe it or not the gray background is a good predictor and tells me how much I will be able to push the sensor if I decide to shoot RAW and post process. If I look at that background I see lots of grain that looks equivalent. I also have several images in which the G9 seems to perform worse, but I chose not to publish as the fluorescent lighting and timing of the photo plays a role.

Sony a73 with the new Tamron 28-75 2.8 sure seems to fit the bill: https://www.bhphotovideo.com/c/product/1393332-REG/tamron_a036_28_75mm_f_2_8_di_iii.html

ThePhilosopher wrote:

I was looking at general noise profile as it mimics what I see for skin tones and general quality under some of the lighting conditions I shoot in. Believe it or not the gray background is a good predictor and tells me how much I will be able to push the sensor if I decide to shoot RAW and post process. If I look at that background I see lots of grain that looks equivalent. I also have several images in which the G9 seems to perform worse, but I chose not to publish as the fluorescent lighting and timing of the photo plays a role.

Sorry, this just reminds me so much of the iPhone vs GX85 comparison someone posted (where most of the picture was just a wall with no detail). Most half-way decent JPEG engines can use NR to reduce the appearance of noise for such uniform areas. To really look at noise performance of the sensor, you have to look at areas with detail.

As others point out, the worse camera tends to turn those areas into mush, while the better camera will keep detail.