I have an interesting question, if you dont mind. Its also may seem quite strange.

From the asm coder perspective, what equation satisfies best a proportion of effectivness between a
- year 1997 64 bit DEC Alpha workstation , clock speed 500mhz (A) , and a
- year 2000-ish Intel 32 bit P4 workstation clock 1ghz (B)

now, xA = yB, whats (x,y)?

Posted on 2011-01-30 12:08:47 by Turnip
I am not sure what is A, B, x and y.

However, looking at patterns; i am assuming that:
A = 500mhz, x = 64-bit
B = 1ghz, y = 32-bit
Then if: xA = yB -> (500mhz)(64-bit) = (1ghz)(32-bit)
Therefore: (x ,y) = (500mhz, 1ghz)

Or the alternative combination:
A = 64-bit, x = 500mhz
B = 32-bit, y = 1ghz
Then if: xA = yB -> (64-bit)(500mhz) = (32-bit)(1ghz)
Therefore: (x ,y) = (64-bit, 32-bit)
Posted on 2011-01-31 12:51:17 by banzemanga
You did word your problem very vague.
In my opinion you have to study both time coding definitions for each system.
After that you might be able to build your own algorithm/formula for a conversion routine for it.
Posted on 2011-01-31 16:31:24 by TasmDev
It all depends on what you want to measure.
Different architectures have different strong and weak points.
People have tried with Dhrystone (MIPS), Whetstone (FLOPS), SpecINT/SpecFP and various other benchmarks... but over time these benchmarks lose their meaning, since CPUs evolve, and these benchmarks can not predict performance of new features accurately.
For example, in recent years, Intel has been adding SSE instructionsets, which mainly focus on things like video decoding or encryption. These things aren't benched... so while a certain CPU may lose to another CPU in the benchmarks mentioned above, it may still be faster at tasks such as video decoding, because it can use special instructions that the other doesn't have.

Having said that, the Alpha is a legendary CPU though, which coupled a relatively high IPC to very good clock scaling (it reached speeds of 500+ MHz long before Intel and AMD did), which most people would consider efficient.
The Pentium 4 is notorious for its relatively low IPC, which, although it had impressive clock scaling, it could not entirely compensate.
Posted on 2011-01-31 17:10:50 by Scali