Not all about clock ? (P4>dual/quad)

[digitaldreams]

I currently have the 3.8GHz P4 (670) and am considering upgrading to a dual or quad....

The benchmarks I find appear to show the latest chips outdo the 3.8 P4 even at lower clock speeds (say 2.66). Is this a side effect of having dual/quad cores or are they much more efficient at executing code anyway ?? (even single thread apps).

Regards

 

[PhillyDee]

Both. Clock speed is not a reliable indicator of speed between brands / models. The way AMD number their chips prove it.

 

[digitaldreams]

There is a topic on 'Toms Hardware' that compares several processors @2.4GHz....yet they have only chosen apps that would benefit from multi cores ! (on purpose ?).

I'd like to see a 'single threaded' benchmark comparing my 3.8 670 to say a 2.6 6700 (or Q6700). An older 3dmark or pcmark maybe ?.

<sigh>

 

[nikebee]

Clock speeds have never been the be all and end all in computing.

It's about how many instructions a CPU can execute in a single clock cycle rather than how quickly it can do a single cycle.

The old Athlon 64's proved this, they ran at slower clock speed than the P4's but were much more efficient in dealing with instructions - and thus were actually faster at completing tasks than the P4's!

 

[digitaldreams]

So....

If running an old 'single threaded' ap or game, would moving to dual/quad core from my 3.8 P4 (800MHz bus) slow it down ?. I'm talking at least 2.66GHz & 1GHz bus.....

Regards

 

[Rezident]

So....

If running an old 'single threaded' ap or game, would moving to dual/quad core from my 3.8 P4 (800MHz bus) slow it down ?. I'm talking at least 2.66GHz & 1GHz bus.....

Regards

I can't see how a dual-core would be slower.

 

[digitaldreams]

An old application/game not using the multiple cores. The only core in use though more efficient than the P4's cannot compete with the 3.8GHz of my current 670....

True or false ?

 

[snowdog]

No the core 2 duo's are just as fast as a pentium D @ half the clockspeed, so a core 2 duo @2.6 would be just as fast as a pentium D @ 5.2 GHZ in Dual-Core apps and as a 5.2 ghz Pentium 4 in Single core apps.

Even 1 of the cores of a c2d (or a c2d in single core apps) would destroy your chip in terms of performance.

 

[digitaldreams]

Excellent.....

That, the fact 'Crysis' WILL take advantage of four cores and several other reasons - justifies my upgrade to Quad.....

Thanks

 

[Corasik]

Yes, in 'general' terms, a P4 was designed to do up to 3 instructions per clock, but due to the way it was designed it normally didnt do more than 2.

AMD's Athlon 64 was a more traditional design, also designed for 3 instructions per clock, but it managed to do that virtually all the time.

Core 2 Duo's are designed for 4-5 instructions per clock, and are normally able to achieve 4, so really it is a good design.

The instructions per clock, are based on a single core, IE the Core 2 Duo being a dual core design is actually able to process 8-10 IPC given the right software, or a C2Q 16-20. Of course keeping all the cores busy isnt an easy task.

In addition to the better design of the C2D, it also has a higher FSB, so better memory bandwidth, and its able to process 128bit SSE instructions in a single clock, a P4 needed 2 clocks per SSE instruction. So for applications optimized with SSE, the C2D has even more potential. All in all its a superior design, and will outperform P4 computers with ease.

 

[digitaldreams]

Nice one !....I'm even more sold on the idea. Yes my bus would be going up to 1000 from 800 (or even to 1333 if I dare push my P5W64-WS Pro & 8000ul ram). Yet I hear bus speed has less effect that the other improvements..

Forgive me if wrong, but while the P4 is executing a couple of things at once, it could also be executing a floating point instruction in it's built in maths co-processor....same to be said of all later chips.

Regards

 

[matthew2456]

There is a topic on 'Toms Hardware' that compares several processors @2.4GHz....yet they have only chosen apps that would benefit from multi cores ! (on purpose ?).

I'd like to see a 'single threaded' benchmark comparing my 3.8 670 to say a 2.6 6700 (or Q6700). An older 3dmark or pcmark maybe ?.

<sigh>

didnt look very hard then

http://www.tomshardware.com/2007/03/26/the_gigahertz_battle/page13.html#video