johnt@seila.UUCP (john grant) (04/15/90)
Like I'm sure many people have done, I downloaded Speedometer to see just how fast my machine really was - since I have a Prodigy SE I was expecting something to the effect "you thrashed an SE". Well, for a bit of fun I decided to see how much, if any, performance was lost by turning off the 68020 cache and by running without "Virtual". Knowing that 1 MIP (Vax) is about 1750 Dhrystones, I ran the Dhrystone test with and without both features. The results were that the 020 cache improved things by about 11% and the extended address space had no effect. All well and good - I have a machine that is about 1.5 * VAX 11/780 in raw power. I then decided to use the "System performance" test of Speedometer. This time the cache made c. 38% improvement, but now "Virtual" was costing me 5%. In all these tests I just disabled Virtual at startup - I did not remove the '51. To my mind, if the internal cache is to have any effect, then it will come when the processor repeatedly accesses the same instructions, which should imply that it is not running wildly through memory. My results showed that when the cache was at its most useful, the address lookup in the MMU was at its slowest. Can anyone explain why my results seem to be upside-down ? As for the other tests, the MMU doesn't seem to be having much affect. On another note, I was thinking of adding the Orchid MacSprint board to the Prodigy, has anybody tried this - I know I'll have to make some space somewhere.