Final-Fan said:
Should we not mention the clock speed of CPUs, for fear that people would compare equivalent AMD and Intel processors and think one is way faster when it's not true? |
Good point... clock speed, mips, fips, etc. have all been used by companies to mislead consumers. Benchmarks have even been terribly twisted by hardware vendors who have intentionally tweaked CPUs to perform well when running the benchmark software.
But you wouldn't compare the clock speed of a GPU and a CPU to try and determine which was faster. CPUs can do GPU-like tasks, and GPUs can do CPU-like tasks, but it doesn't mean that a 3-GHz CPU would make a better GPU than a 1 GHz GPU card.
Here's an example... I run a company with 50 employees, and we use Linux servers. Someone says, "replace those Intel 2-core machines with AMD 4-core machines, both running at approximately the same clock speed, and your Linux users will experience improved performance." This is very likely a true statement.
Then someone says to me, "replace your Intel 4-core machines with Cell 8-core machines and your Linux users will experience improved performance." This isn't necessary true.
So, what can generally be said about adding more cores on an SMP-aware machine (Linux, Windows, OS X), can't necessarily be equivalent when talking about the Cell as an "8-core" processor. It would depend largely on the OS, hardware support, etc.
I admit that for those who understand processors, saying the Cell is an 8-core CPU makes sense. But for end-users of the machines, I still believe it is misleading. I don't believe a 3 GHz "8-core" Cell general purpose machine would be necessarily faster than a 3 GHz 4-core Intel machine for general purpose computing tasks. But calling the Cell an 8-core CPU would likely lead a lot of end-users to that conclusion.
