Ethernet energy efficiency

▼ How power-hungry are various permutations of Ethernet on modern MacBook Pros? Tests performed and written up by Jeroen van der Ham and Iljitsch van Beijnum.

For some reason, we thought it would be a good idea to make the students of the System and Network Engineering master program test whether various permutations of IPv4/IPv6, energy-efficient Ethernet (eee) and jumboframes use more or less power. While the students were busy measuring, we were bored so we decided to run the same tests ourselves using our MacBook Pros (with Retina Display®)—Jeroen the 15" mid-2012 model and Iljitsch the 13" late 2013 model. These computers are too thin to house an Ethernet port, so we used Apple's Thunderbolt - Gigabit Ethernet adapter. We looked at the number of milliamps taken from the battery to see how much power the entire system used during various tests.

We didn't realize it at first, but the most surprising result was actually the first one, looking at system power use with the adapter plugged into the computer, but without attaching a network cable. This increased system power use by 300 mA, or about 3.4 Watts. On the 15" MBP this is already significant, increasing power use from 1250 to 1550 mA. But the 13" MBP only used 370 mA without the adapter plugged in, so its power use almost doubled after plugging in the adapter!

Could it be that the secret to the long battery life of Apple's recent laptops is the fact that they come without an Ethernet port?

We continued our testing trying different possibilities with Energy Efficient Ethernet (eee) turned on or off. Students in 2012 already looked at the general impact of eee on energy usage for clusters. They found that it provides an energy saving of up to 40% in switches, depending on the amount of traffic. Energy usage goes up when traffic goes up, but it already maxes out when you are using roughly 40% of the bandwidth. We confirmed this in our testing: there is not much difference with eee when the interface is idle, or with full traffic.

More significant energy savings started to appear when we enabled jumbo-frames (MTU 9000). The amount of data that goes through the adapter remains virtually identical, but with only one-sixth as many packets per second, the impact of per-packet overhead is greatly reduced, and this shows in the energy consumption. With full traffic in one direction this saves you over 10%. These savings also add up if you have traffic in both directions, in our experiments we saw a savings 16-19% for a bidirectional fully-utilized link.

Not surprisingly, the biggest difference in energy usage is to scale down the interface speed. When we forced the interface down to 10Mbps we saw a saving of up to 49% with full utilisation. Then again, if you're sending a large file, it is going to take a lot longer, and you basically use a lot more energy, only more distributed over time.

The results of our measurements are in the table below, but note that our measurements (using the System Profiler battery current figures) weren't very precise. The numbers give a pretty rough indication. If you want to do something serious with this data, perform your own experiments.

Total system power use	MBP15	MBP13
Interface attached vs baseline	+24%	+92%
Baseline vs idle eee	+27%	+110%
Idle eee vs no eee	-6%	-1%
Idle eee vs iperf 1 Gbps TCP eee	+74%	+97%
iperf 1 Gbps TCP eee vs no eee	-3%	-1%
iperf 1 Gbps TCP jumbo eee vs no jumbo eee	-13%	-12%
iperf 0.5 Gbps UDP eee vs no eee	+1%	+1%
iperf 1 Gbps TCP bi-di eee vs unidirectional eee	-0%	-3%
iperf 1 Gbps TCP bi-di jumbo eee vs no jumbo bi-di eee	-19%	-16%
iperf 10 Mbps TCP no eee vs 1 Gbps	-38%	-49%

We did some extra tests to look at the impact of jumboframes on CPU use. Please note that our Thunderbolt-to-Gigabit Ethernet adapters don't support TCP segmentation offloading (TSO). With 1500-byte packets the CPU utilization on the MBP13 was about 16%, with 9000-byte packets about 90%. Despite being idle more than 80% of the time, the CPU still ran at its advertised 2.4 GHz clock speed:

So the difference in CPU power use was 18% (7.9 W for 1500 bytes vs 6.4 W for 9000 bytes), despite the difference in CPU utilization only being 5%.

Permalink - posted 2014-03-03

Testing cameras on the oil rig

▼ As you may know by now, there's an oil rig 12 km off the coast of The Hague. 12 kilometers is pretty far away, so it's not easy to photograph. Even with a long zoom lens it shows up pretty tiny:

Today the weather was clear and sunny, so I tried once again with both my new Nikon D7100 with the Nikon 70-300 mm lens and the not-so-new Canon SX280 point-and-shoot. I mounted them both on a tripod and tried 300, 240 and 200 mm on the Nikon and the full 90 mm (about equivalent to 300 mm on the Nikon) with and without digital zoom on the Canon. I cropped the images to show just the oil rig.

The results (click the links to load the different images):

It looks like the Canon is slightly sharper than the Nikon combo. And although the 70-300 mm lens is supposed to be a bit soft at 300 mm but shaper at below 200 or 240 mm, it still shows a tiny bit more detail at 300 mm than at 240 mm and certainly more than at 200 mm. And as expected, the results for using digital zoom or simply cropping the photo on the computer later are virtually identical.

Permalink - posted 2014-03-02

Pentacon 135/2.8 "bokeh monster" lens

▼ I originally planned to get a D3200 or a D3300, which are part of Nikon's entry level DSLR lines. But I couldn't resist the extra features that are part of the more advanced D7100. One of those features is that you can use very old lenses that don't have a built-in CPU. So with a small adapter ring, I was even able to mount my 35-year-old (give or take a few years) Pentacon 135/2.8 135 mm f/2.8 lens, also known as the "bokeh monster".

Unlike modern lenses, the Pentacon 135/2.8 is manual-everything. Obviously the focusing is done manually, and the same thing for setting the aperture (the size of the lens opening). Nikon has been moving away from even having an aperture ring on the lens itself for about a decade now. Instead, you set the aperture through the camera. Or the camera sets the aperture automatically. But even when you set the aperture on the lens, the aperture doesn't actually close—that only happens when you press the shutter to take a photo. Not so on the Pentacon 135/2.8. Here, if you turn the aperture ring, you see the viewfinder image getting darker in real time. Which makes manually focussing a lot harder. Fortunately, the 135/2.8 has an additional preset ring that sits on top of the aperture ring. This allows you to set the minimum aperture. With that set, you can move between the maximum and minimum aperture without taking your eye off of the viewfinder by turning the aperture ring as far as it will go.

Surprisingly, the 135/2.8 holds its own on the D7100's 24 megapixel sensor as far as sharpness goes:

However, the lens has a number of other problems, including color fringing and the contrast is really poor in backlight, such as in this photo of my bicycle handlebar with the sun reflecting in the water behind it:

What you can see here too is what earned this lens its nickname "bokeh monster". Bokeh is a Japanese term used by photographers to describe the quality of the out of focus elements in a photo. The reflections in the water turn into very large discs here, which is often desirable. The edges of the discs are a bit sharper than they could be, though. In this case, I made it easy on the lens by shooting with the lens "wide open" at its maximum f/2.8 aperture, which makes for maximum background blurriness. When closing down the aperture, the shape of out-of-focus highlights is determined by the shape of the aperture, which is normally created by 5 to 9 blades. The Pentacon 135/2.8, on the other hand, has no fewer than 15 aperture blades, so the out-of-focus highlights remain round.

It's too bad that when using this lens with an adapter (Amazon link) on a Nikon DSLR makes it impossible to focus beyond about 10 meters...

If you want to read more about this lens, have a look at this review. I especially enjoyed the "hipster appeal" section:

The best feature: "Made in GDR" printed in large letters on the side of the lens. This is a two-pronged benefit: A. it's clearly not an American lens, and B. you will often get the opportunity to explain to people what "GDR" means and make a thoughtfully ironic comment about the Berlin Wall.

It's a lot of fun to be able to use a very modern camera with such an old lens, and having the focus indicator confirm that I nailed the focus manually 80% of the time—if those damn ducks would stay in one place for more than half a second!—is a nice little ego boost. And it makes you appreciate all the photographic automation we've gained over the past decades.

Permalink - posted 2014-02-24

Nikon D7100 file sizes vs image quality

▼ My new Nikon D7100 has a 24 megapixel sensor. I shot some test photos at 24, 13.5 and 6 megapixels with low, medium and high compression to see what the tradeoff between image quality and file size is. Have a look at the results.

These are the options for saving your photos on the D7100:

• raw:
  • 12- or 14-bit
  • lossless or lossy compression
• JPEG:
  • full 6000x4000 resolution, 4496x3000 or 2992x2000
  • fine, normal or basic compression
  • size priority or optimal quality

So how big do the JPEGs get? According to the manual:

	6000x4000	4496x3000	2992x2000
Fine	8.2 MB	5.0 MB	2.7 MB
Normal	4.1 MB	2.5 MB	1.4 MB
Basic	2.0 MB	1.3 MB	0.7 MB

Don't forget that 24 megapixels means 12 million green pixels, 6 million red pixels and 6 million blue pixels, while in a JPEG every pixel consists of three subpixels for the three primary colors. So a 24 megapixel JPEG actually stores 72 megapixels the way the camera makers count them. (Nikon even says the 640x480 LCD display on the back has 1.23 megapixels even though 640x480 = 0.3 MPX. They do have four subpixels per pixel, though.)

I wanted to see what the difference between 4496x3000 (13.5 megapixels) and 2992x2000 (6 megapixels) is at different JPEG compression levels. For this, I used the following scene, which should be fairly challenging to the JPEG compression scheme with lots of fine and contrasty detail. (35 mm f/1.8G DX lens at f/5.6.)

The red square indicates the part of the image shown below. When the page first loads, you get the 6000x4000 version. I'm not sure if this is the sharpest possible image, though, and I'm not sure what the compression setting was. Probably fine, as the file was 19 MB. The different versions of the file are all zoomed to different magnifications so the image remains the same size on the screen. Click the links below the image to load the different versions.

Comparing the 24 megapixel version with the two smaller ones using fine compression, it looks like the 24 MPX version encodes more data than the sensor and lens can deliver, the 6 MPX version less and at 13.5 MPX the file resolution and lens/sensor resolution seem reasonably well matched.

At 13.5 MPX, if you look hard you can already see a hint of JPEG compression artifacts in the fine version. They're easy to see in the normal version, and pretty bad in the basic version. At 6 MPX, they're bad in the normal version and hideous in the basic version.

Note that with "optimal quality" (i.e., JPEGs are allowed to be larger if needed to reach the selected quality level) and this chosen subject, the file sizes (as reported by the MacOS Finder) are a good deal bigger than suggested in the manuall:

	6000x4000	4496x3000	2992x2000
Fine	19 MB	12.0 MB	6.1 MB
Normal		5.9 MB	2.9 MB
Basic		3.0 MB	1.4 MB

So I think I'll shoot most of my photos at 13.5 megapixels normal, but I'll make sure to go to 24 MPX normal or 13.5 MPX fine when I want the best quality. I actually shoot raw+JPEG, where the raw files go to one SD card and the JPEGs to another. So if needed for one reason or another, I can always get the raw file (14 bits, lossy compression). However, I'm not much of a tinkerer after the fact, so with my previous DSLR, for 99% of all photos, I kept the JPEG and threw away the raw version. Photos take up half of my laptop's 250 GB SSD as it is, and 24 MPX raw files are 20 - 28 MB a piece.

My apologies to readers who aren't into photography, by the way. I'm sure there will be a few more posts like this one in the next two weeks or so.

Permalink - posted 2014-02-23

DSLR 101

▼ This is a post about the aspects of DSLRs in general and the Nikon D7100 in particular that those of you who haven't used a modern digital single lens reflex camera may not know about.

Before the days that cameras had LCD screens on the back, seeing what you were about to photograph was somewhat of an issue. The simple solution was to have a viewfinder with its own little lens above the "real" lens. But that doesn't show you exactly what's going to end up on the photo. A single lens reflex (SLR) camera, on the other hand, lets you look through the actual lens you're going to use to shoot the photo through a mirror that redirects the image to a ground glass that you look at through the viewfinder. When you press the shutter button, the mirror flips up and the shutter opens to take the picture.

Of course these days a much simpler way to accomplish much the same thing is to show the image that the sensor sees on the LCD screen. When you press the shutter button, the shutter first closes, then opens to take the picture and finally opens again so you can frame your next photo.

SLRs have been around for more than half a century. There were many innovations over the decades, most notably measuring the light and thus the correct exposure through the lens and autofocus. The two big brands are Canon and Nikon. Canon started from scratch with the Canon EF autofocus SLR system in 1987, so everything since 1986 is compatible, everything before isn't. Nikon came up with its F-mount in 1959, but over the years continuously refined the interface between the camera and lenses. If you really want to, it's possible to mount a lens from 1959 on a modern Nikon DSLR, although for the really old lenses a modification is necessary or you'll damage the camera. Lenses from 1977 and later will fit without trouble, although many compatibility caveats apply, sometimes even with lenses that are still sold as new today!

Because the light doesn't reach the sensor until the shutter opens, (D)SLRs need separate focus sensors to be able to autofocus. These special focus sensors allow DSLRs to focus very quickly, especially considering that parts of relatively big lenses need to be moved. The difference is especially notable when you compare shooting with a DSLR using the optical viewfinder with using "live view", where the LCD screen on the back of the camera is used as an electronic viewfinder and the focus sensors can't be used.

What surprised me when I got a Nikon D90 four years ago is that the mid- and high end Nikons can actually show certain information overlaid over the optical viewfinder image: you can have grid lines turned on and off, and when half-pressing the shutter button so the autofocus system gets triggered, the active focus point or focus points are highlighted. Apparently, there's a black-and-white LCD superimposed over the viewfinder image. (It gets illuminated in red briefly in low light.) In addition, below the image, the selected shutter time and aperture and some other settings are displayed in green or white. So you can change the most important settings without ever taking your eye off of the viewfinder.

Speaking of settings: of course even high end DSLRs offer an "auto" mode, where the camera takes care of all the settings. But my new Nikon D7100 also has no less than 31 buttons, levers, switches and dials that let the experienced photographer tell the camera exactly what he or she expects from it. And suddenly the "all auto all the time" approach of the iPhone camera seems very appealing...

I'm sure some people get a DSLR and use it in auto all the time. And there's nothing wrong with that. But the trouble with letting the camera select all the settings is that you're going to miss some shots because the camera selected the wrong setting. For instance, with the iPhone, it's nigh impossible to get decent backlit photos: if you're trying to take a photo of a person or an object in front of a bright sky, the iPhone sees the bright sky and makes the person in front of it way too dark. You can tap on the person to indicate that you want to have him or her in focus and correctly exposed, but even then the phone still looks at too wide an area and overexposes.

With a DSLR (and also many point-and-shoot cameras that offer more manual control), you can use spot metering, exposure lock and/or exposure compensation to correct or avoid this problem. (And Nikon's matrix metering is very good, it often avoids these problems to begin with.) Although tap-to-focus on an iPhone is pure magic, a DSLR can track moving objects with its autofocus system. But that mode doesn't work so well for some static subjects, so you need to change autofocus modes between shooting sports and landscapes or portraits.

So what it boils down to is that a DSLR gives you many more tools to get the more difficult shots that simpler cameras just won't give you, but the downside is that you have to be forever vigilant that you're in the right exposure and autofocus modes—or at least, not the wrong ones. But there's no rule that says you have to master each and every one of those 31 buttons.

Permalink - posted 2014-02-22