Difference '030@50MHZ and '040@50MHZ

[shoggoth]

LaserBack wrote:
if anyone here believe that the 060 is 2 times faster than a 040 is because never had both cpus to compare

Not true at all. The 060 is twice a fast at the same clock speed. And yes, I've had both to compare.

for example...quake on a 040 about 14 fps...on a 060  maybe 15 fps
alien breed 3d 2 ...on a 040 18fps...on a 060 19fps

What you're referring to is the main bottleneck of the system; chipram bandwidth. The CPU is a lot faster, but the RAM is the limit there. Doesn't matter if the CPU is twice as fast - the data still can't get stuffed into chipram faster. You're simply using the wrong kind of benchmarks.

The 2x times faster comparison was done years ago on magazines..it was a lie for marketing purpose
[/quote]

Not.

[shoggoth]

Effy wrote:
Now a question that some of you will definitely not like,
what about an MC68LC060/75 without fpu Does the lack of fpu justify the higher speed ??

IMO - a 060 w.o. FPU is kind of pointless. The FPU in the 060 is really fast, which means it's used quite a lot in 060 demos/apps. Later incarnations of the chip can run at 75Mhz (or even 100Mhz), and includes an FPU.

[shoggoth]

Effy wrote:
The reason for using an 68060/75 without fpu is not to use 68060 software but 68020 optimized ... does that make sense  :-?

Nah, you'll just miss the whole point and want to upgrade to a "full" CPU later on. Take my word for it - I did just like you said (albeit at another platform - Atari).

[shoggoth]

stefcep2 wrote:
At what point do they become "two totally different processors"?  Is an 030 "totally different" to 68060?  Well other than running nearly the same 68k code as opposed to x86 code or alpha code, they share nothing else in common in terms of transistor number, pipelines, cache, fpu instruction set, design, number of executed instructions per cycle. Turn the caches off on an 040 or 060 and they execute the same instructions slower than 50 mhz 030.  Turn the caches on and they may not execute at all!! AIBB is good (but not for 060) because it does things like rendering a beachball and timing it and yeah mips gives very misleading results.

I think we need to make a distinction between CPU(+memory) benchmarks and system benchmarks. The former is of interest to those using CPU intensive stuff like rendering, while the latter takes system bottle necks into the equation.

Personally, the mips figure makes sense to me since I get a rough idea of what kind of performance my code can get when only accessing RAM.