Computing power of Apollo systems?

[Belisarius]

I've heard and read all kinds of descriptions about the processing power of the Apollo DSKY computer, usually comparing it to a digital watch or a pocket calculator.

But since all computers do the same job, they can be compared accurately in terms of processing power, memory, etc.

So the question is: What was the processing power and memory capacity of the Apollo on-board DSKY computer? Does anyone have stats for it in terms of the usual criteria for describing computers?

I can't find reliable information on this anywhere

 

[ryan]

Well in it's time it held the most information that an computer that has ever held, it did hold alot of memory and data. If it's compared as a digital watch and it kind of looks like one, that's not true, it's one of the best quidance computers that MIT has ever pulled out.

 

[Tschachim]

Some infos and a couple of links are here: http://en.wikipedia.org/wiki/Apollo_Guidance_Computer

Cheers
Tschachim

 

[Urwumpe]

Also, you can't compare the AGC directly with a modern PC - it's a whole different architecture. If I remember correctly, the AGC allowed executing instructions parallel, something which only the more modern PC CPUs are capable of.

 

[gimp1992]

The AGC is describe as:" The AGC has three principal sections. The first is a memory, the fixed (read only) portion of which has 24,576 words, and the erasable portion of which has 1024 words. The next section may be called the central section; it includes, besides an adder and a parity computing register, an instruction decoder (SQ), a memory address decoder (S), and a number of addressable registers with either special features or special use. The third section is the sequence generator which includes a portion for generating various microprograms and a portion for processing various interrupting requests.
The backbone of the AGC is the set of 16 write busses; these are the means for transferring information between the various registers shown in Fig. 1. The arrowheads to and from the various registers show the possible directions of information flow.".
Also check here, http://astro.uni-tuebingen.de/~wilms/computers/apollo.html

 

[NukeET]

Checkout this recent thread:

http://orbiter-forum.com/showthread.php?t=1119

 

[Belisarius]

According to the Eyles article referenced above, we have 152 Kbytes of onboard computer memory considering both the CSM and LM systems.

The article is so good, I'd like to reference it again:

http://www.doneyles.com/LM/Tales.html

Check it out, Apollo-fans!

 

[Belisarius]

By the way, I love the description of the DSKY verb-noun interface from Dave Scott's article on the same: "It was so simple that even a pilot could learn how to use it"

 

[Urwumpe]

By the way, I love the description of the DSKY verb-noun interface from Dave Scott's article on the same: "It was so simple that even a pilot could learn how to use it"

At least, the design was better as the later Shuttle DPS interface. This requires ten times more key presses as the Apollo DSKY per hour of mission.

 

[Belisarius]

But doesn't it do much more stuff?

 

[Urwumpe]

But doesn't it do much more stuff?

Yes - and no.It not only has more complex subsystems, but these subsystems actually require less attention as they are more robust as the Apollo subsystem equivalents (for example both have three fuel cells). The problem is getting the information. Where the apollo only needed 7 key presses, the shuttle needs you to key in more data as the display format is not practical. The data is displayed by function and subsystem, not by mission phase. So, for checking the state of the vessel, you need to switch over many different displays in the shuttle.

Also, the next problem for the shuttle: Its onboard computer type (GPC) is practically a radiation hardened ancient IBM System/360 (The family of airborne equipment is called System/4Pi among fans and military plane designers) .

So it has to be operated like a mainframe of the 1960s. like for example loading the initial software for all computers in the shuttle from tape - this software even included information which instructions the CPU knows. Your PC CPU actually also has this capability today - but only since the year 2000.

(And you won't ever notice that this feature is active. Such microcode patches are included in BIOS updates and are loaded in a microsecond without ANY message bothering you)

Even just switching the major function from GNC to System Management for a single Shuttle GPC is almost a complete reboot of the GPC - except the lowest level operating system software, all software in memory gets changed.

When a GPC fails, you need to manually distribute all data buses it commanded to other intact GPCs, as the software logic is not able to tell, which of the other GPCs is better. Exception is the switch to the backup flight system - in that case one special GPC assumes command of all and the remaining GPCs have to stay idle, or get even manually disabled.

But that BFS does only work for ascent and reentry - in Orbit, the fifth GPC, which is specially designed to run this kind of software, gets shut down and restarted only for reentry.

Short: If you are computer fan, and like hacking, you will love the Space Shuttle computers. If you are more fan of the good old "Take it out of the box, start the computer and all works properly", the Apollo AGC will be more your kind of flavor.

And if you like colorful screens and using the mouse/Touchpad a lot - you will love the ISS. You don't need to do much, but when you need to do it, there are cool graphical displays.

 

[Belisarius]

Fantastic information - thanks Uruwumpe

 

[movieman]

The LVDC (Launch Vehicle Digital Computer) in the Saturn V was fun too; the CPU had seven pipeline stages, and each was executed independently by three different units, which then voted on the result before passing it on to the next pipeline stage. That way up to seven units could fail out of the twenty-one in the CPU and the Saturn V would still fly to orbit; obviously if you got unlucky and two units in the same pipeline stage failed the whole computer would fail, but that was far less likely than two units in different stages failing. That was the only way they could get the MTBF much over 100 hours in those days.

 

[Belisarius]

That's interesting - but what is MTBF?

What happened to this rig when it got hit by lightning in A12? Did all of these pipelines go offline, or just some of them?

 

[Ghostrider]

MTBF = Mean Time Between Failures. Essentially, the time that passes between events that go further than "aw shucks what's that red light here".

 

[Tschachim]

That's interesting - but what is MTBF?

Mean Time Between Failures

What happened to this rig when it got hit by lightning in A12? Did all of these pipelines go offline, or just some of them?

Fortunately nothing happend to the LVDC, only the CSM systems (including the AGC) were affected.

Cheers
Tschachim