Processor Speed Standardization

Hello. This is my first post on this forum. I have a question that some of you who are good with numbers and theory may be able to help me with, I hope. I have done much research and some experimentation, but am anxious to see if anyone here can help, because I don't know the answer.

There are many processors out there- some fast, some slower. This is based on number of pipelines, cache, and Mhz, generally, and some other factors. Those 3 being most important for overall speed. Today's fastest processor is the Core 2 Duo. I have 2 questions. First, how do we tell what is faster without setting up tests and controlling for all other factors minus processor? For example, we know that the Core 2 Duo is faster than, lets say, the Pentium 4 HT. However, at what point does that hold true? What if an overclocker has a 4Ghz P4 HT and compares it to a 1Ghz Core 2? Or, theorhetically, in an extreme example, a 5Ghz p4 HT vs a 500 Mhz Core 2 (in theory only). Certainly there will be a point when the other chip is faster than the one generally known as the fastest (or shall I say there WILL be a point that the slower one will be faster if you keep upping the clock speed, even if impossible in a practical application- think 8Ghz P3 vs 300Mhz Core 2- yes, silly, I know, but you get my point). How do we know at what point one processor is faster than another? This applies to anyone who may be ready to build a new system. you may ask yourself, "Do I go with that Core 2 of 1.xxGhz or that Pentium D at 3.8?"

That brings me to my second question. How do we standardize processor speed? Is there, or has anyone developed, a formula for standardizing cpu speed. For example, (Mhz) / (number of pipelines) = Speed rating. This is just a very crude example, but I was thinking along these lines.

Here is the application of this and why it may be very important. What if there is a system builder who wants to build a new system, but has a budget. Say he can afford the lowest clocked Core 2 or the Highest P4 HT (or Pentium D). How does he figure out what is faster??? This is the core of my question. I am a new system builder and I have not been able to find a way to standardize this process. Once again, bear in mind that my formula was not a real consideration, but just an example. I would be grateful if anyone can offer any input on the subject even if it is not very specific to what I asked. How do you determine this, Johnny Systembuilder??? Sorry for such a long post Mods. Much appreciated!

 

Remember that two heads are always better than one (processor core wise)!
Sometimes its not just a matter of how fast, but how much! ie...How many programs or tasks can it handle simultaneously.
An O/C single core may be higher in specs, but when this runs out there is nowhere else for processing tasks to go (hyperthreading may help to a certain extent). A dual core is effectively double the clock specs!

You are right though...standardization for testing purposes would mean running in conjunction with exactly the same hardware specs for other devices.

Regards....

 

Thanks, nitecrawler. Good points. I wonder, though, at what points does any given processor beat another. Lets say Pentium II vs Pentium III. Clock speeds being equal, the III is the pick. How do you know when the II is the pick? It is very complicated! What do you experienced system builders do??? Just say, "well the Core 2 is the fastest- I'll just go with that in highest clock I can afford?" How do you know your Pentium D clocked higher doesn't perform better than a lower clocked core 2? It seems like we are in the stone age with comparing processors. You have AMD guys saying this is the best, but Intel guys say that is the best. In this day and age, how is this even debatable? At what point can we say.....this processor has these specs, this processor has those- therefore this is faster, all else being equal. Question is, what specs do we look at, how heavily does each weigh in performance, and then what is the mathematical formula that tells us the answer (I mean taking in EVERY physical aspect- cores, clock, pipelines, cache, anything else). I know we use benchmarks, but those can be very problematic to say the least, they don't isolate well, and they are after the fact. By after the fact I mean you must have your system built to run it and then it is too late to compare potential chips you could have bought.

 

I understand your point, however there are many factors that weigh into the equation of 'system performance', other than CPU capabilities!

These would include RAM capacity and clock speed, motherboard FSB, or even hard drive interface such as sata v's pata (with regards to data tranfer speed).
You may have a super equipped CPU, but if your FSB aint up to it, it will create a bottleneck that will slow performance down.
Sort of like putting a small 4cyl carbuerator on a high performance 350 chev (if your into cars). No good havin a high grade engine if you cant get enough fuel to it......know what i mean

All these factors are important, cause after all, its complete system performance that we all strive for!

Hope i make sense.......Regards!

 

Yeah, I know that the multiplier and FSB on your processor makes your GHz. The FSB*multiplier is your MHz. For example a processor with a FSB of 166 MHz and the multiplier is 12. 166*12=1992 MHz = 1.992 GHz. THat's all I know, Simple amth.

 

Again good points, Nitecrawler. I am trying to isolate processor performance only. All else being equal, how do we determine processor performance. You make a hell of a good point. A motherboard for a P3 superoverclocked compared to a board for a P4 underclocked would still yield an advantage for the P4 due to the advancements on the board alone. Here is a better example of what I am getting at then. Say we have an LGA775 board that supports the P4 HT, Core 2, and Pentium D. Now, what chip is the fastest here if all are going in this board? Do you take a 1.xx Ghz Core 2 or a 2.8 Pentium D? Or how 'bout the 3.4 Ghz P4 HT? At what threshold does one chip have an advantage over the other in this instance? This is a case where all other hardware is controlled for. It is also a real-world question considering I have a board that fits this scenario. When do we say, "Ok, at this clock the P4 gets the Core 2 at this clock and at this speed the Pentium D gets the Core 2, and so on? I think that it would seem obvious that a Pentium D at the highest factory offered clock speed would be the faster chip vs the slowest Core 2 offered. Now we can raise the clock of the core 2 by buying new factory chips or OCing, but when does the D catch the Core2 or vice-versa and when? This is just an example, but therorhetically how do we determine at what point one cpu has an advantage, all else equal??? A formula is possible, just that no one yet knows it. It may be ungodly complex, but there is an answer. I wonder how you seasoned system builders would tackle that question?

 

I guess all things being equal, a benchmark is the only true test of CPU superiority if that is what you are trying to isolate (when complete system performance is not the issue).

A seasoned system builder would still look at overall system performance in preference to simple raw CPU clock...
I mean even if the overclocked P4 outperformed any of the other processors, (all hardware equal, including cooling and PSU) they would still think twice about using it, if it created excessive heat and caused system instability.

Sorry Walter, i will butt out now and let you get other opinions!
<nitecrawler, crawls back into his box>..hehe

Regards mate, and good thread!

 

BAMF!?

*couldn't resist.*