Performance of the 8088 on PC, PCjr and Tandy 1000

It’s well-known that you should measure the performance of your code, and not rely only on the opcode’s “cycle counts”.

But how fast is an IBM PC 5150 compared to a PCjr ? or to a Tandy 1000? or how fast is the Tandy 1000 HX in fast mode (7.16Mhz) compared to the slow mode (4.77Mhz) ? Or how fast is a nop compared to a cwd ?

I created a test (perf.asm) that measures the performance of different opcodes and run it on different Intel 8088 machines. I run the test multiple times just to make sure the results were stable enough. All interrupts were disabled, except the Timer (of course). And on the PCjr the NMI is disabled as well.

Without further ado, here are the results:

And these are my conclusions:

  • As expected, 1-byte size + 3-cycles opcodes are the fastest: nop, xchg ax,xx, inc ax (note that nop is literally a xchg ax,ax)
  • 2-byte + 2-cycle opcodes (like mov al,al) are twice as slow as the 1-byte + 3-cycle opcodes.
  • CPU intensive instructions like mul and div perform exactly the same in all machines. Cycle-eaters don’t affect them that much (kind of expected).
  • In theory, the 7.16Mhz Tandy 1000 HX should be 50% faster than the 4.77Mhz mode. And that’s true for CPU-intensive opcodes like mul and div. But opcodes that are “cycle-eaters”-bound the performance gain is 0% in most of them.
  • IBM PCjr vs. IBM PCjr: The PCjr could be by far the slowest of the 4.77Mhz machines, or the fastest:
    • When running code in the first 128 Kb RAM the performance is terrible! A simple mov al,al is 2.6 times slower! This is due to additional wait-states.
    • On the other hand, when running code above the 128 Kb RAM, it is the fastest (albeit by a tiny bit).
  • One of my IBM PCjr is a tiny bit faster than the other one. Couldn’t find out why. The CPU is the same, same clock speed.  Something in the bus (?) perhaps ? It is faster enough to break some cycle-dependent code.
  • From fastest to slowest:
    • IBM PCjr (code running above 128 Kb RAM)
    • IBM PC 5150 (tiny bit slower)
    • Tandy 1000 (tiny bit slower)
    • Tandy 1000 HX 4.77Mhz-mode (tiny bit slower)
    • IBM PCjr (code running below 128 Kb RAM) (slower by a huge margin)

One opcode that is worth-noting is cwd:

  • It is a 1-byte size, 5 cycles opcode
  • It is a tiny bit slower than a nop
  • The negative side is that it destroys ax and dx

What I like about it, is that it is between 7%~17% slower than a nop (depending on the machine). And when doing timing dependent code, sometimes you need an opcode that is just a tiny bit faster or slower than another one. In my case, I was able to have a stable raster bar on the Tandy 1000 thanks to it (more on this on a future post).


Published by ricardoquesada

cocos2d, unicyclist, commodore 8-bit

8 thoughts on “Performance of the 8088 on PC, PCjr and Tandy 1000

  1. What RAM expansions are on each PCjr? This could explain the performance differences both between A and B, and between the PCjrs and the PC.

    1. Do you mean the internal 64k RAM module? or the ones in the sidecars? If the internal one, I’m using the
      “default” ones, I guess. Probably I should try swapping them and see if the results changes.

      If you meant the sidecars, I tried with and without, with different sidecars and the results were the ame.

      1. I meant the sidecars, but that is interesting. Yeah, wouldn’t hurt to try swapping the internal modules, but now I’m wondering if there’s a V20 or something in the faster PCjr…

  2. To the core, basics. Interesting.

    Noting that x86 haven´t changed much.. If one could make the instructions a bit more like highlevel lang, say a lot of DSP, are clear candidates for combining several clocks into one, a 4pole filter (one input, one output), can be combined into a one clock operation aswell. JSR could take arguments. And ofcourse commonly used functions. Even larger things that has just one input, and one output. And this could even be done in parallel. would not need C much at all.
    C indeed seems to be a “mr. fix it”, for masochistic assembly language.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: