Early in this century, I went to pick an Apollo DN 3000 in Apt, in the southeast of France. It turned out the machine was a (much older) DN 300, and the vendor had made a typo.
Although quite disappointed, I decided to pick the machine anyway; and the vendor, truly sorry, added some sparc gear he had lying around, so the trip was worthwhile overall.

This machine had been bought new in december 1983 or january 1984, and has been used as a CAD system for a few years. Then it was replaced by a more modern system, and kept in storage until its owner sold it to me for a ridiculously small price.
Unfortunately, years of storage have taken their toll on the machine's health; the built-in display is exhausted and very dim, even with the luminosity knob in the maximum position; and the CPU board doesn't lit its diagnostic LEDs, which makes me think some components are dead (the fans still blow, so at least the power supply is still working).
Since this machine is the eldest of my collection, I could not resolve to get rid of it, although it was obvious I wouldn't ever find time to work on tracking down the electronics failures. I eventually donated it to the french Silicium museum.

According to the sales brochure, the DN 300 is a DN300 Desktop Computational Node, and is described as:

Apollo Computer's DN300 Desktop Computational Node places mid-range 32-bit supermini performance on every user's desk. In a single package, it delivers virtual memory capability, up to 1.5 Mbytes of main memory, high-performance graphics (1024 x 800 pixels), and access to the 12 Mbit per second DOMAIN local-area network.

The DOMAIN system is a high-performance local-area network of dedicated computers operating in a distributed environment. Computational Nodes are connected together using commonly available, high-bandwidth coaxial cable organized in a ring topology. This network is the mechanism that allows system-level hardware and software resources to be shared throughout the DOMAIN system.

By itself, the DN300 is a powerful workstation designed to expand the productivity and creativity of an individual technical professional. The DN300 can also become part of a larger, shared-resource computing environment as one of many Computational Nodes connected together by the DOMAIN network.

And now, for a tour of the machine, with pictures! Click on the thumbnails for high resolution pictures.
Here is what the machine looks 25 years after being manufactured. Yes, that's me in the screen reflection.

The machine badge has had better days...

And now for the back of the machine, where I/O connectors lie.

Another way to figure out this machine is a DN300. And it has a 512KB memory option, for a total of 1MB!

From the bottom up, the I/O connectors are:
Two DB25 serial ports.

Then a Token-Ring network transceiver.

Look at that connector, it looks like an old motherboard power supply connector with no less than 18 pins!
With the transceiver removed, one can better see the reset button (on top) and the normal/service switch (bottom).

And the plate at the top hides two 50 pin connectors, and acts as a cable guide to prevent them from getting loose.

Time to open the machine! There are only two screws to remove to be able to remove the left cover (see the screw guides at its bottom).

Apollo engineers have been nice enough to tell us what goes in what slot.

Here is what the fully populated cardcage looks like on my DN 300. Note that the two rightmost cards are connected by a thin cable, and the rightmost card also is connected to a cable coming from somewhere in the chassis...

From left to right (thus in decreasing slot number, according to the configurator), the cards are:

• The floppy and hard disk drive controller. Unlike the other cards, it has two connectors on the backplane side; the second one connects to the 50 pin I/O connectors on the back of the chassis. The light blue covered area between the connectors looks like a set of batteries to me.
  
• Then comes the Token-Ring controller. Apparently the single backplane connector is enough to carry the necessary signals to the external transceiver.
  
• The frame buffer.
  
• The CPU board. The two large purple chips are the CPU at the top, and the DMA controller in the middle. Both run at 8 MHz. I was expecting the CPU to be a 68010, yet this is a SC87842, which doesn't ring a bell. According to the sales brochure, this is a dedicated 32-bit VLSI processor with integral page fault hardware (as opposed to high-end models, back then, such as DN460 and DN660, featuring a 32-bit bit-slice processor and cache memory).
  I first thought the MMU was the Motorola 68450, but checking against old databooks, the 68450 is actually a DMA controller. This would imply the MMU is built-in the SC87842 dedicated VLSI processor.
  Quoting to the sales brochure again, the integral page fault hardware provides 16 Mbytes virtual address space per process, up to 15 concurrent processes per node ; MMU dynamically maps 24-bit virtual addresses into a 22-bit physical memory address space ; MMU maintains protection and usage statistics for each page of memory ; page size is 1024 bytes.
  Note the unusual page size. While the Sun-1 had choosen a 256 byte page size (even smaller than the VAX 512 bytes!), Apollo was favoring larger pages, which was the right thing to do, even with the scarce physical memory sizes back then.
  The bottom-left quarter contains 512KB of memory. The bottom-right corner sports the boards EEPROM, as well as the SCN2681 dual serial port chip.
  
• And the last board is the 512KB memory expansion. It is only half the size of the other boards, and thus can not connect to the backplane. Thus the front connector must be a power source.
  

And here is what the card cage looks with all the boards removed.

Looking sideways, one can notice there are only four connectors on the backplane; even if the memory expansion was full size (which might be the case of the 1MB expansion board), it would still require an external power supply connector.

Now let's have a look at the external storage enclosure. At first glance it looks like a boring 8 inch floppy disk drive.

However, it has two cables to connect to the DN300. This because one cable is connected to the floppy drive, and the other to the hard disk drive inside. See for yourselves at the back of the enclosure:

Let's look closer at the label... Indeed, it has a 1.2MB floppy drive and a 34MB disk drive.

Here is what the storage enclosure looks like without its cover.

Almost nothing fancy in the floppy drive. Note the metallic cylinder at the top right edge, it's the main motor...

...which drives the rubber belt, if you look underneath.

While there, have a look at the floppy drive manufacturer's name, a name long dead (just like Apollo...), and the odd manual adjustment, likely needed to make sure the data tracks furthest from the axis fit within the magnetic floppy surface.

And now, the hard drive. Being 34MB, it weights 34 lbs, of course (actually, a bit less).

What's special about this one is the external lock, just like on a washing machine, which emerges from the enclosure when the disk is properly seated.

And now for the disk identification. Micropolis, yet another vendor which got bought more than 15 years ago. I could not find out what technology the disk uses; at first I thought this was an SMD disk, but it only has a single 50 pin connector, so SMD, ESDI and ST-506 are out of question. Maybe a SASI disk?

And finally, with both the floppy and hard drives removed, the bottom of the enclosure contains the power supply.