• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Core Duo and my confused brain

Associate
Joined
5 Feb 2003
Posts
365
I am a little bit confused over the whole dual and quad processor thing. My workstation has a Pentium 4 2.66ghz or something like that. When I read of people using a core duo 1.5ghz, does this mean 1.5 x 2 processors which means effectively a Pentium 4 3.0 ghz? Or what?

Yours confusedly,

Jon
 
Its not quite the same no. Its like having one fast bloke doing a job or two slower blokes doing a job. Just because there are two people doing the same job doesnt mean that its quicker, as the job might be something only one person can work on at a time, like hammering a nail into a peice of wood. But if it is something they can both work on at the same time, like painting a fence, then it will be quicker.

I hope that makes sense.

PK!
 
I understand precisely what you mean. But I still don't see the difference, since the Pentium 4 3.0ghz share's its cycle time between the different cpu intensive tasks. How is that different to dual core where you get 2 cpu's sharing their cycle time?
 
There are many factors towards performance, number of cores and clockspeed barely scratch the surface. Otherwise the Pentium D (dual Pentium 4) would've been a bloomin' rocket on speed, but we all know otherwise.

For example, a single core Pentium M CPU running at 2Ghz often bested a 3Ghz Pentium 4 when running on similar platforms. The reason isn't just down to clockspeed or number of cores, but the way the processor is optimised and designed. The Pentium 4 (Netburst architecture) for example was one of the worst performance per clockspeed processor designs ever released, as Intel designed the processor to reach high frequencies rather than making it an efficient design.

Netburst wasn't all bad, though. Without it, we probably wouldn't be seeing Core 2 reaching the amazing overclocks it is today.

Best way to judge performance isn't to look at the numbers shown to you when someone's trying to flog you the processor, look at some benchmarks and such as buying a piece of hardware is very rarely as simple as two numbers.

Multi-core processing only really comes into its own in two situations:

Where you're running many intensive programs (say, unraring a large file and watching a HD movie at the same time)

Or where a program has been programmed to take advantage of multiple CPU cores.
 
Whilst uncompressing Winrar will be mostly waiting for the HD to write the data hence the low CPU utilisation.
Try compressing a large archive that should be more CPU intensive.
 
There are two issues here, firstly dual core concept, and then pentium V core duo.

If you dont mind.. lets split this into two questions..

Single core V dual core.
As your familiar with a 2.66Ghz Pentium IV, lets compare with a Dual Core Pentium D running at 2.66Ghz. Program design can either be threaded (which means the program is divided up into several 'micro programs', or a single process. A single process application cannot make use of multiple processors, so it will perform almost exactly the same on a dual core, as it would on a single core processor. I say almost, because windows itself has a small overhead, and any background tasks you run at the same time will benifit from having a spare processor.

Of course if you run several single process applications (multi tasking), then the second core will be used considerably more. It still wont be like having a 5.32Ghz processor.. The applications wont complete in half the time, but it will mean the system can run more applications than a single core computer before it slows down.

Finaly there is the holy grail of multi processor goodness, the multithreaded application. Programs written like this break down into several mini programs, each of these mini programs can be run on separate cores, only pausing when data from one thread is required in a separate thread. Typically performance benifits in a well written multithreaded application can be anywhere between 50% and 95%. ( A web server application for example could use 1 thread for each user request, so if 100 people are browsing, 100 theads could be spread across 100 cores and be extremely efficient). This is the only time that a multicore processor will actually be able to complete a single program faster than a single core chip. (Edit.. I dont mean webserver is the only example.. I mean multithreaded applications are the only time that dual cores can complete single applications considerably faster)

Ok, So thats Single V Dual (Feel free to ask more questions :P) so now Pentium IV compared to Core 2 Duo (Core Duo was replaced very fast so I will focus on the superior Core 2 Duo).

Pentium IV's can theoretically process 2 Integer instructions, and 1 Floating point (or MMX/SSE) instruction at the same time, this is called a 'three issue core' as it can do 3 tasks at the same time. Not only this Pentium IV runs its integer instructions at Clockspeed X 2. So in theory its a very powerfull processor for integer applications. Unfortunatly due to several design issues, its rare that the P4 doesnt just run out of data. It then sits there doing nothing waiting for new instruction to process. It was very fussy about program design, and running programs optimized for older Pentium III processors makes the P4 look even slower. In reality P4's generally seem to average 1.5 to 2 instructions on each clock tick. (Available as single core, or Dual core Pentium D)

Moving briefly to AMD's Athlon 64, its also a 3 issue core, however its a much better design, and is an excellent processor for running applications optimized for Pentium III. Infact even fairly badly optimized code will not slow this AMD chip considerable. A very good design. This chip averages 2-3 instructions per clock tick. (Available as Athlon64, or dual core X2's)

Finaly Core 2 Duo. Each core of a Core 2 Duo is able to process 4 instructions at the same time, and in very rare instances its even able to combine 2 instructions into one before its send to the execution units in the processor, this simulates running 5 instructions at once. That said, the combining is a 'new thing' and Core 2 Duo struggles to do this on 64bit code, so it only really happens with 32bit software. Like the AMD Athlon 64, Core 2 Duo is much better at handling less perfect applications, this chip will average 3-4 instructions every clock cycle.

So in very simplistic terms, at 2Ghz, the Core 2 Duo is processing as many instructions (on a single core), as a Pentium IV (or Pentium D) would process at 4Ghz, because Core 2 duo is twice as efficient. AMD 64 sits somewhere between the two, depending on the application the AMD will be between 10% and 40% slower than the Core 2 Duo.

Its not quite as 'simple' as this.. but in basic terms thats the differences between Pentium and Core 2 Duo... So in the future, when you see people with dual core 2.66Ghz computers claiming they have 5.32Ghz computers you will be able to laugh. Just think about cars... If you have 1 car that can carry 4 people 70 miles in 1 hour, and then you buy a second identical car, you still take 1 hour to travel 70 miles, but now you can carry 8 people. Dual core doesnt mean you can do work faster, but you can do more work at the same time, if and when the programs allow it.
 
Last edited:
That was an extremely informative post and you have clarified so much for me. A big thank you. I suppose if you had an application that could use both cores at the same time, then you could have a computer 4 times the speed if the processor clock speeds are the same, correct?

There is a quad core processor out there for around £170. Does this work on the same principles as the Core 2 Duo?
 
as the job might be something only one person can work on at a time, like hammering a nail into a peice of wood. PK!
P4 = Hammer away, get distracted, remove part hammered nail, start again.
OCed C2D = using left and right hands to hammer at breakneck speed.
OCed Quad = using hands and feet to hammer at the speed of light but generating enough heat to warm the universe.
 
That was an extremely informative post and you have clarified so much for me. A big thank you. I suppose if you had an application that could use both cores at the same time, then you could have a computer 4 times the speed if the processor clock speeds are the same, correct?

Yep, assuming a program that scales across multicores, a C2D running at 2Ghz could be up to 4 times faster than a single core Pentium 4 running at 2Ghz. It would also use less electricity than the P4 computer as well! (Although CPU is only part of a computers power usage, graphics cards, memory, hard disks etc all combine to make the total power usage)
 
P4 = Hammer away, get distracted, remove part hammered nail, start again.

Funny, but so true.. P4 'should' be fast in theory, with high clock speed, and double clock speed ALU's. But in reality, it spends half its life waiting for instructions especially after branch prediction fails and it has to flush the pipeline.

Could be mean.

Get distracted, Hammer nail, realise nail in wrong place, remove nail(flush pipeline), hammer a new nail, start again. :)
 
Funny, but so true.. P4 'should' be fast in theory, with high clock speed, and double clock speed ALU's. But in reality, it spends half its life waiting for instructions especially after branch prediction fails and it has to flush the pipeline.

Could be mean.

Get distracted, Hammer nail, realise nail in wrong place, remove nail(flush pipeline), hammer a new nail, start again. :)

At least one person understood the analogy.:D
 
Back
Top Bottom