Need Help Understanding CPU's

[ajsellaroli]

Hey, I am trying to understand CPU's, in general. I have a pentium 4 processor now, at 2.8 ghz. When looking at the benchmarks provided by tom's here, I saw that a 1.8 ghz dual core processor performs almost 3 times better than mine in that particular benchmark test. How can this be? it has far less ghz...right? I don't really know how these things are rated, but am very curious, I would appreciate all the help i can get, thanks.

 

[hughyhunter]

Yeah... I dont get it either!!!

 

[Grimmy]

....

What dual core?

If its an AMD, its a different architecture... it uses 3 instruction sets vs 2 instruction sets, as well has a memory controller on the chip.

If its the C2D by Intel, again different architecture.. it executes 4 instruction sets vs 2, and has a share L2 cach.

So an analogy of that would be..

If you were to scoop sand with what is used in a sand box, and someone else had a regular dirt shovel, hes just going to dig more sand then you.

By that, I'm just basically saying per cycle (even at a lower speed), a particular CPU is going to do more then the other, and its more like comparing apples to oranges. They are not the same.

 

[ktheripper]

Grimmy,

I think what's confusing for them here is these ratings through out the years to my understanding are based on how many instructions a CPU can process per second. (Think it's second)

With that said if it's meant that 1 ghz can process 1 billion instructions per second. 1Ghz should be 1 ghz no matter what architecture it is.

Just like 55 mph is 55mph whether you're in a pinto or lambo.

aj the way I try to make sense of it is now with say dual cores, the proc is (it's not as simple as what i'm saying here but to try and help you out) so just processor alone 1.8 x 2 is cable of processing 3.4 billion instructions per second provided software, os and everything being used is optimized for mulitiple cores.

However, that is just part of the puzzle in actual performance in most of these tests. The bus architecture on the core 2 duo's is going to be faster, if they are doing encoding tests disks and IO could factor in. I don't have time to go into a lot of detail but hope that helps a little bit.

 

[hughyhunter]

http://www.hardwaresecrets.com/printpage/209/9

This is the very basic way a cpu works... but like our age old friend Grimmy (how old are you anyway? 3000+ posts! Jeeze) said... different architecture on different chips. Dual cores are different too... Basically a dual core can do two sets of instructions... very easy to understand when you understand a single core... but intel has shared cache so when one core is not doing much it and the other is the loaded core will "share" cache from the idle core... AMD doesnt have that feature yet (pooh) and therefore wont share the workload between the two cores.

For a single series of chips though (like the conroe) you can rest assured that the higher clocked chip (if they have the same amount of L1 and L2 cache) will win in performance. But as soon as you start adding cores (only if the app that you are running will utilize those cores) the one chip with the most cores wins (as long as the program uses all four cores).

 

[Grimmy]

but like our age old friend Grimmy (how old are you anyway? 3000+ posts! Jeeze)

. o O(Are you trying to say I POST too much???)

I'm old enough to know better... :lol:. o O (I think)

 

[Raviolissimo]

when i was in college i built an 8 bit adder using mechanical
components.

i also took classes in machine language.

i still don't feel like i really understand computers.

a comparable phenomenon in nature is the rate
at which birds flap their wings.

big birds like a hawk might flap it's wings once
a second. with a hummingbird, it's a lot more
than once a second.

yet the hawk, with the slower flapping wings,
moves a lot more air.

if i continue i'm going to sound like Cheech &
Chong trying to explain computers.

 

[BigJ3384]

I have a pentium 4 processor now, at 2.8 ghz. When looking at the benchmarks provided by tom's here, I saw that a 1.8 ghz dual core processor performs almost 3 times better than mine in that particular benchmark test. How can this be? it has far less ghz...right?

The comparison between a pentium 4 processor which has a single core at 2800 MHz and a dual core processor (Core 2 Duo, etc.) which has two 1800 Mhz cores is something like this. imagine that you have a car that you have to push up a hill. would you rather have one large, powerful person pushing it, or two slightly smaller people pushing it? the relationship to your processors is similar. although the two cores running at 1.8 GHz won't do the same work as a single 3.6 GHz (the increased performance will not be linear), it will do considerably more that one 2.8 core. also there is that fact that both cores will run cooler and not as close to max performance, which will reduce the chances of lockups and burnt silicon. generally speaking, it is better to have two even four cores doing a little work as opposed to one core doing all of it. just note that increasing from one core to two cores, or even to four cores, will not raise your system speed by a factor of 2 or 4. there are still other bottlenecks in the system (system busses, etc.) and the law of diminishing returns to consider. just my two cents...

 

[Smoked Turkey]

Hz Mhz Ghz is a measure of cycles per second based on a sine wave. The number you are seeing is cycles per second. 2 Ghz is 2 billion cycles per second, or on and offs (CPUs work in Binary 1 or 0, yes or no, true or false). Each CPU is made up of regions of transistors that perform certain functions based on what they are designed to do.

Now, as the technology increases the number of transistors on the given space also increases. You have to understand that the Silicon wafers the CPUs are printed on are extremely expensive to produce so real estate is a major issue. With those increases in technology (ie 90nm 65nm 45nm, where nm is the wavelegnth of IR light used to produce the transistors.) they are able to put more transistors on the same amount of material. So your 2.8 Ghz Proc is based on an older, 90 and above, nm technology, therefore on the same sized die there is less transistors performing computing processes, than the newer ones. (The transistors are in the Hundreds of Millions)

Another factor is the chip designers ability to "streamline" how the processor computes a given set of instructions (as in the newer models can perform the same task more efficiently than their older counterparts.)

Also noted is Cache. Cache memory is memory that is built directly into the CPU and runs at full CPU speed with zero latency VS your system Memory or RAM that runs at a speed slower than your CPU (between 8-10 times slower). The CPU can use the cache to perform its calculations at a faster pace than if it was to wait on the system memory. So in most cases a larger Cache will also help the CPU perform faster, and with the shrinking of CPU technology (45nm) they are able to fit more cache on the same amount of space, at zero cost to wafer real estate.

Zang

 

[Thanatos421]

I think one person mentioned it already, but the biggest determining factor is IPC. Instructions per clock. Some will say it is a misnomer, but the general theory behind the name remains the same. Take a P4 2.4 and a Core 2 1.8. Disable one of the cores on the dual and it will still out perform the "higher" clocked P4.

Example (none of the IPC numbers I'm about to mention are fact, unless I get lucky. I'm just pulling them from my hiney.)

A P4 can execute 2 instructions per clock cycle. A Core 2 can execute 4 instructions per clock cycle. So, all things being equal: 1.8x4 vs 2.4x2. Like I said, those probably aren't the true IPC of the chips, but you get the general idea.

 

[BigJ3384]

thanatos is right. combine the fact that the core 2 architecture can perform much more work in the same amount of time as the pentium 4 architecture, combined with the fact that there are two of them doing the work, you easily arrive at your 3x or more performance difference. just my two cents...

 

[JonnyRock]

when i was in college i built an 8 bit adder using mechanical
components.

i also took classes in machine language.

i still don't feel like i really understand computers.

a comparable phenomenon in nature is the rate
at which birds flap their wings.

big birds like a hawk might flap it's wings once
a second. with a hummingbird, it's a lot more
than once a second.

yet the hawk, with the slower flapping wings,
moves a lot more air.

if i continue i'm going to sound like Cheech &
Chong trying to explain computers.

i like this one, birds and cpu's, priceless

 

[hughyhunter]

. o O(Are you trying to say I POST too much???)

I'm old enough to know better... :lol:. o O (I think)

I'm not saying you post too much... i'm just saying your probably the one that invented the internet! Or better yet for sake of this post... you probably are the "one" that invented the microprocessor...

when i was in college i built an 8 bit adder using mechanical
components.

i also took classes in machine language.

i still don't feel like i really understand computers.

a comparable phenomenon in nature is the rate
at which birds flap their wings.

big birds like a hawk might flap it's wings once
a second. with a hummingbird, it's a lot more
than once a second.

yet the hawk, with the slower flapping wings,
moves a lot more air.

if i continue i'm going to sound like Cheech &
Chong trying to explain computers.

Dude... you do sound like Cheech! Had me laughing!