From personal experience and use (always shooting raw).
From worst to best.
C2005 Canon 350D 22.2x14.8mm sensor size - 8MP (**)
C2009 Canon G11 P&S - 10MP (not a huge ammount of difference between these two. (**)
C2007 Canon 40D 22.2x14.8mm sensor size - 10.1MP
C2005 Canon 5D 35.8x23.9mm sensor size - 12.8mp - Full frame and a big step up image wise from the above cameras.
C2015 Olympus OM-D E-M5 II 17.3 x 13 mm (M4/3) - 16MP
C2013 Canon 70D 22.5 x 15mm (APS-C) - 20MP (I feel it is about 1 stop better in noise performance than the Olympus body above), however there are other factors that increase usability, in the real world, for a lot of my shooting needs (not all) - by using the Olympus bodies.

As the above shows a smaller MP camera is not necessarily better noise wise than a larger sensor camera (from using my own cameras).
(see my first post in this thread)

I tend to think my lowly 12MP Canon EOS Rebel T3 does a better job of decreasing the noise issue, than ANY of my four 24MP Nikons, Richard … and that the one 16MP Nikon I have - withholds the onset of noise much sooner than any of the other DSLRs I possess - including the equal sensor Pentax K-50.

But, this isn't meant to be an expose of my cameras, Richard. It's meant to be a reaching out for a more clear explanation of why lesser-megapixel cameras (12MP-16MP) seem to do a better job of controlling the noise issue, than more recent 24MP-42MP/46MP/50MP cameras. Since the pixel-sites are smaller, for each newer generation of hi-res sensors - it would seem to be MORE difficult to control noise, as the high density pixels - are more apt to show noise at a sooner point. Thus, if one pursues the MP race - one is actually wounding oneself, in the process, and, perhaps, might've been better off, saving one's money.

My Canon 1D MkIII is 10mp and several generations older than my 1Dx MkII which is twice the megapixels. But while both are low noise leaders in their generation/build era there is no comparison between the two. The MkII is vastly more capable of high ISO shooting than the MkIII. I would say the MkIII is good to ISO 800 or perhaps 1,600 while the MkII is easily capable of ISO 25,000.

Technology marches forward. It is best to compare cameras of the same era and generation in my view. Also note that the MkII is full frame sensor and the MkIII is an APS-H sensor, 1.3x crop factor. The 1Ds was the same era as the MkIII but probably had slightly better ISO performance. I have not owned one nor compared the specs, but the MkIII was built for speed and the 1Ds not as much. They merged into one 1 series body with the 1Dx I believe.

Nearly all noise in a RAW image fille is thermal noise from the sensor.
It is relate to two things sensitivity (ISO) and sensor temperature.
Cool the sensor cold enough: no more thermal noise.

In electronics, noise is measured by signal-to-noise ratio.
There is no fundemental reason why higher MP sensors should
have more noise. HOWEVER, there may be an engineering reason:
inadequate cooling. This could cause slightly more noise when
shooting a high MP sensor at high frame rates.

Sensors located inside of cameras are difficult to cool adequately
But so far as I am aware, this has been a problem mainly in
lower-end cameras with "global shutter" sensors, not DSLRs
or high end mirrorless cameras with focal plane shutters.

Phase One makes 100MP backs for medium format film cameras
(real "pro" cameras, like the Miyama and Hasselblad).
These backs have cooling fans or thermoelectric
(Peltier effect) cooling systems.

The worst case for heat would be what al ot of peoople
are running out and buying:

* Mirroless camera with no mechanical shutter
* "Global shutter" sensor
* Fast frame rate
* EVF screen located inside the camera

Think about it: the camera is a small box, probably made
out of plastic. How you gonna cool it?

There are only four ways to move heat:
* Conduction
* Convection
* Radiation
* Active cooling (refrigeration)

None of them work well nside a small plasic box: How well
does plastic conduct heat, compared to say, copper? How's
the air flow throgh that box? If the sensor radiates heat, can
the IR light escape the box? And where you going to put a
refrigeration system?

The bottom of the sensor is covered with contacts--surface-mount
pads that get soldered to the board, so there's not even anywhere
to put a heat sink.

Phase One covers the entire back of the board in a heat ink. That's
not going to fit inside a camera.

Technology isn't about buzzwords--"The Cloud", "nanocrystals", etc---
it's about engineering. You can't get four pounds of crap into a
one-pound box no matter what the marketing department decides
to call it.

Maybe having a tiny, plastic camera wasn't really such a great idea
after all. Maybe that's one reason why Phase One's sensors are so
much superior to the ones the consumer camera companies are stuffing
inside of cameras.

Be careful what you wish for--because Nikon, Canon, Sony, etc. will
give it to you. Phase One, Leica, Hasselblad, Linhof, Sinar , Lockheed
Martin will not--they customers demand top performance, not top styling.

I was curious what DxO testing showed for signal to noise ratio of your cameras. Their testing seems to match what you're saying about your cameras. I drew wider, solid lines over their dotted lines on the graph because their dotted lines are hard to see when small like this. You can still see part of their dotted lines over on the right side.

DxOMark has three dotted lines on the graph, a green line, a yellow line, and a red line. Above the green line, photos taken at those ISOs will be relatively free of noise. Under the green line noise will be noticeable. Under the yellow line noise will be quite noticeable, but not catastrophic. And under the red line noise will be dramatic.

One reads a great deal on the Net - vis a vis this problem, but one is left with a certain uncertainty. It would seem - the higher res one attains - the less the threshold, before the onset of noise. In other words, you are better off with a 12MP camera, than you are with a 24MP camera. And, conversely, you're better off with a 24MP camera, than you are with a 50MP camera, as - the higher you go up the ladder, the higher in density are the photo-sites - which, in turn - makes the onset of noise - that much sooner. Please comment, if you would like. And if you have a clear explanation of this phenomenon - please, elaborate, if you've a mind.

It is sort of like packing a lot of people in a room. if you put 100 people in a small room there will be a lot of noise. Put the same number in a larger room, they will probably irritate each other less, and there will be less noise. Double the size of the room and the number of people you will get a lot of noise, but with the same hundred you started with, put them in the big room = less noise. The noise comes from them talking to each other. Same with electronic noise. Put a lot of electronic elements in a small space - they talk to each other = noise. Fewer elements in the same space, or same number in a larger space, = less noise.