With a standard BB2 3.9 system the fast-ram speed it is roughly 30% quicker than the Apollo, it surprising matches the chip-ram speed of the Apollo in most cases but for a couple of the test the Apollo beats the Blizz by about 20% (you need to download the modules to get the exact ratings).
Maybe by boosting the Mhz you have altered things such that Blizzard partly catches up to Apollo Chipram speeds?
In fact, by boosting the Apollo from 50Mhz to 80Mhz it might have even slowed down its chipram access times. One can never know unless one tests. The way accelerator cards access memory is a field of darque blaque majix.
Or maybe you are averaging out all the chipram timing tests which is not something I care about.
The speed of reading from chipram using bytes, words or longwords is interesting from a scientific perspective but not really useful to the consumer in determing which card is best.
The speed of writing to chipram using 8-bit writes or 16-bit writes is interesting from a scientific perspective but not really useful to the consumer in determing which card is best.
The only chipram test that matters is how fast chipram can be written to 32-bits at a time. If a game is writing 16-bits at a time then it is unoptimized slow as molasses anyway and doesn't bear consideration.
When an Apollo has "20% faster chipram access" this says to me, as a game coder, that Apollo cards have a 20% faster framerate, which is fantastically important to gamers and animators. Anyone who uses the Amiga for video purposes, such as SCALA or CLARISSA or Total Chaos AGA or etc. etc.
As a gamecoder, all my games are bottlenecked by the speed at which my code can write to chipram 32-bits at a time. (because chipram is 32-bit memory). All I care about is the speed of move.l Rn,ChipRam
Having faster fastmem is kewl and I love it, but that is not where my bottlenecks are at.
So when I did my timing tests on Apollo vs. Blizzards over the years I was mostly testing stock 50Mhz models (because that is what most ppl have). And I was only paying attention to 32-bit chipram writes and all my tests were performed in 640x512x256 colors mode because that is what my games use.
Would you be willing to post your Busspeed test results from both cards in 640x512x256 colors mode? (fastram and chipram, the whole thing)
Thanx!
Now if you want a surprise, get yourself an Apollo 1240 40Mhz and check out the amazing chipram access speeds! I never had such a card but various playtesters sent in timing test results many years ago and it was really amazing how the 1240 could magically always access chipram at the maximum theoretical value. Even faster than theoretically possible sometimes
