🚀 go-pugleaf

In article <720553537snx@dartac.demon.co.uk> sjr@dartac.demon.co.uk (Simon Roberts) writes:
>Whilst I can see that there might be some advantage for DOS applications,
>doesn't the fact that a UNIX system is unlikely to be in the 8K internal
>cache for any length of time suggest that the DX-50 could come out
>ahead?

Well, the effect of the 8K internal cache is greater than you might suspect.
I ran some tests (under unix) which i derived from old Byte benchmarks
(but adapted for the faster machines available today).

I compared a 486DX-50 without cache, with internal cache and with
256K external cache.
The result is that:
- for very small programs the effect of the external cache is zero
  but the internal cache increases speed by almost a factor of 2 over
  no cache.
- for somewhat bigger programs the effect of the external cache gives
  an improvement of almost a factor of 2 over the internal cache only.
- for a test of 15 shell-scripts using several pipes, the effect of the
  external cache drops to about 10%, but even only the internal cache
  still is a major improvement over no cache.

This suggests that even for heavy multitasking the 8K internal cache
should make a DX2-66 come out ahead over a DX-50.

Given 500 interrupts per second (that's an awful lot for a Unix system)
the time in between interrupts is still enough for a 486 to execute
20000 to 40000 instructions, enough to take advantage of the 8K internal
cache.

Paul.