Comparing Processor Speeds

[artbuc]

Just read on another forum that GHz is no longer a valid way to judge processor speed. They said it was like trying to determine car speed by engine rpms. They said the only way to compare processor speeds was to use benchmark data. Is this true? Where dos one find this benchmark data? Thanks.

[DaveLembke]

Benchmarks are best for overall.

But depending on what your running the overall core design, L1, L2, L3 cache supported instruction sets SSE2, SSE3 etc, FSB, 32 or 64 bit execution, TDP...

Additionally the CPU is also affected by the motherboard performance such as a lower end motherboard might support the CPU, but might not be the best of performance as well as features might not be available. *I ran into this when going cheap on a AMD FX-8350 4Ghz build. Found out after the fact that the chipset doesnt support the 4.2Ghz Turbo that the CPU was advertised as having.

It really comes down to what your going to run with it. If your running a single threaded application and ran it on a Dual Core 2.8Ghz and a Quad Core 2.8Ghz with same core design and cache layout/allocation per core, you would notice very little difference.

Clock speed alone should never be used as for I have a 1.3 Ghz Sempron Kabini Quadcore 25 watt TDP that blows my 3.0 Ghz Pentium 4 HT 95Watt out of the water performance wise. However with some (older) single-threaded applications the Pentium 4 3.0 Ghz HT is better. However most modern applications that are multi-threaded the 1.3 Ghz runs way better.

[BC_Programmer]

Clock Speed is only an indicator when all other things between the processors being compared is effectively Equal.

For example- an 8088 at a higher clock speed than another 8088 processor is going to be faster. Once you start comparing different processors- even similar ones such as the 8088 versus an 8086; the differences in the processor design have larger effects, an eventually- once we start comparing CPUs with vastly different designs- the architecture difference has a much greater effect than the clock rate on how fast things are processed.

If it was possible to do so, for example, a older 286 processor overclocked to 10Ghz would still be vastly outperformed by a Pentium 4 chip, solely due to the architecture improvements in the chip design that occurred in the meantime.

If you want to compare two systems/processors in terms of raw speed a tool like GeekBench usually fits the bill. I mention that since I used it last week to get a reading on all the systems i have that can run it. It also runs on a number of platforms/systems so if you were inclined you could compare the raw speed of a Smartphone or tablet's processor with a desktop or laptop system.