What determines how much a processor can handle?

nightly1029

Distinguished
Aug 27, 2009
50
0
18,630
I just assumed the more gigahertz the processor has, The more programs it can handle before going under 100% load, But I've been told that is wrong. So, Since some Intel I7's and high performing CPUs have the same gigahertz, And even the same L2/L3 cache sizes as lower performing CPUs, What is it that makes them better?
 
Solution
The architecture.

Back in the days of netburst P4's, Intel was adding more GHZ in order to compete with the superior AMD Athlon processors of those days. Then Intel hit the ceiling of around 4.0ghz and really got no farther than that without running into heat issues (which is already abundant from processors of that era).

Meanwhile AMD released their dual core athlons etc and continously slaughtered Intel's processor in performance but due to Intels superior advertising skills, AMD was still behind them.

Once Intel Core Duo and Core 2 Duo lineup was released, AMD has been left behind. Superior architecture and superior thermal performance made them ~30%+ better performing than any ealier petnium CPU. Performance continued to be gain...

lilotimz

Distinguished
Aug 31, 2009
875
0
19,160
The architecture.

Back in the days of netburst P4's, Intel was adding more GHZ in order to compete with the superior AMD Athlon processors of those days. Then Intel hit the ceiling of around 4.0ghz and really got no farther than that without running into heat issues (which is already abundant from processors of that era).

Meanwhile AMD released their dual core athlons etc and continously slaughtered Intel's processor in performance but due to Intels superior advertising skills, AMD was still behind them.

Once Intel Core Duo and Core 2 Duo lineup was released, AMD has been left behind. Superior architecture and superior thermal performance made them ~30%+ better performing than any ealier petnium CPU. Performance continued to be gain after every die shrink (90mm->65mm ->45mm).

Then the first generation Core i7s/i5/i3's were released. They brought in a much more optimized architecture that further increased performance compared to the latest Core processors with the I7's routinely 15-30% better in many tasks compared to the latest core 2 quads.

Finally the second gen core processors came out with a much more mature architecture and increased performance compared to the first gen.

Sum it up, better architecture + programs being able to take advantage of multiple core's made higher ghz not as important as architecture.
 
Solution

lilotimz

Distinguished
Aug 31, 2009
875
0
19,160
It's really simple. Straight out just pure performance it goes something like this.

Core 2 Duo-> AMD Athlon II X4 -> Petnium G (core architecture) ->Core 2 Quad -> Phenom II X4 -> Phenom II X6 -> Core i7-XXX/i5-XXX/i3-XXX -> Core i7-XXXX/i5-XXXX/i3-XXXX/

Usually an AMD X3 or Core i3 is more than enough for many normal users. Phenom II X4 are good lower cost gaming CPU's while Core i5-2500k is for the enthusiast gamers (best perf for price) while the 2600k is for people who do a lot of professional work.
 
Not really, no.

For instance, a Core 2 Duo running at 1.2GHz is about as fast as a 3.2GHz Pentium 4. This is despite the fact that even assuming 100% scaling, the C2D can only achieve a combined 2.4 GHz worth of processing. Thats because the architecture is superior.

IPC = Instructions per clock cycle, or how many operations a CPU can perform over a period of time. This has just as large an effect on CPU performance as GHz does, but there is no way to directly measure it. Because of that, you can't just look at cores + speed and conclude one processor is faster because it has more cores and runs faster.
 

pepe2907

Distinguished
Aug 24, 2010
643
0
19,010
The number of cores even more as the architecture. Two cores usualy outperform one of them :). And i7 has 4 cores with multythreading - that means it can perform operations on 8 threads simultaneously /although it still has only 4 ALUs, 4 IO blocks and most importantly - 4 FPU blocks, so it gains only from faster switching with less stacking in the cache/.
 
^+1. Furthermore, multiple cores don't help single-threaded applications run faster.

So you need to know the processor AND the applications you'll be running to determine which will be faster doing what you want to do. A 6-core processor might run transcoding faster, but might well be slower than a dual core processor when gaming.

For gaming, the cpu chart Tom's publishes each month is your best bet for an "on-average" answer. For other applications, look at specific benchmarks. Or post your intended usage here, and ask for opinions.