How this dual channel works + CPU overclocking ?

[leo888]

Hi,
I wanna to know how the dual channel ram works, I mean I have 2 X 256MB PC2700 so that means the total speed should equal to 666 (2 x 333), if so I would like to overclock my cpu to 2.4GHz from 1.8Ghz and I now trying make sure that I aren't pushing my ram too hard. If anyone who have overclocked their 1.8A , Please leave some suggestion on how much did you guys push it to that is without problem.

I have a P4 1.8A CPU w/ Alpha 8942T heatsink and fan, Albatron PX865PE Pro v2 Mobo, 2 Kingston 256MB PC 2700 RAM.

[demon_wolf]

Hey leo888,

First off, the P4 1.8A runs standard at a 400 FSB (100 x 4)

By changing it to (133 x 4) or 533 FSB that will run the Processor at 2.4 MHz.

You can pretty much run that with standard Heatsink/Fan no problems. With your Heatsink/Fan it will be no troubles. The memory will run fine.

My question is: with a board like that, why run a P4 1.8A?

If you really wanted to get into overclocking, trade up the processor to a 533 or 800 FSB and go from there.

With the Memory, Dual Channel is like a Raid 0 Array. It doesnt really double the performance but basically uses more bandwidth.

Basically Dual Channel processes faster.

Hope this helps.

[leo888]

thx for the reply

so that means the ram total speed = 666 or is it still at 333 with 2 x 256MB PC 2700 in dual channel,

One thing that i stick to the 1.8A, it is that I still couldn't decide on which CPU to get, Prescott or northwood ? any suggestions....

[demon_wolf]

The Ram total speed is still at 333 MHz, it basically uses all of it instead of some compaired to Single Channel Memory.

For the CPU, theres a few combinations:

1) Any 533 FSB, the more Cache the better.

or

2) Any 800 FSB, again the more Cache the better.

It really depends on how much you are willing to spend.

[Christian-Comp]

I'd say something, but I'd probably just get flamed again...

[johnnyis42]

i'll just add a bit to clarify why dual channel runs faster without increasing the speed of the memory in megahertz....

bandwidth is measured as a function of the width of the bus measured in bits, and the frequency at which data can be communicated measured in Hz, usually MHz or GHz.

the intel processors since the original pentium (and earlier processors maybe as well, but i am certain everything since the pentium apply to this) have a 64 bit wide memory bus. this means that as long as the intel archetecture has the 64 bit memory address, memory badwidth is determined with an increase int he bus frequency. since the frequency increases require improved hardware and power requirements, there is a big limit placed on memory bandwidth by current technology.

however, the dual channel concept fixes that my inserting a 128 bit bus, 2 seperate 64 bit busses running in parallel between the north bridge and system memory. granted the processor still uses a 64 bit wide fus to communicate with the memory, the difference on the newer pentium 4 chips with the 800 mhz front side bus is that the 800 mhz is twice that of the 400 mhz (200mhz ddr) RAM used in the system. since the front side bus is twice as fast, it can switch between the two seperate 64 bit memory channels without missing a beat, effectively being able to read from or write to both channels once before they send their next bit down the line.

sooo, the end result is getting an effective 800 mhz over 64 bit wide bus by using 2 64 bit busses running at 400 mhz. in order to get the full potential out of this system your processor's front side bus speed has to be at least double that of the memory bus.

on the issue of the pentium 4's quad pumped bus speeds.... some people say it's a scam or whatever... the truth is that there is a system bus speed that goes through a logic multiplier to increase data speed by a factor of 4. the catch is that all data transfers at this speed... however address bits (the actual location of data in paged memory, which DOES NOT mean pagefile.sys, but the location in the memory pages in system RAM) transfer at the speed of the system bus. seeing as data is going to be the biggest load on your memory bandwidth, you will likely never notice the fact that the address bits are so slow. in certain cases of things like encryption/decryption, and other cases where you go through alot of little pieces of info that fill up memory locations all over the place, you may notice a difference in performance compared to other hardware platforms. but in real world scenerios most people never notice. most people in situations where this is a huge performance hit in mission critical systems are using Sun Alpha or AMD Opteron enterprise platforms.

anywho, hope this was informative

[leo888]

thx a lot guys, the info was helpful enough for me to determine how far I can over my CPU without OC' my ram too much,

As for the cpu..
I would say it doesn't really matter is a 800 or 533 but higher is always better but I have find that there is two version to it, Northwood and Prescott.. I did some reading around the web and many have different thoughts of the two so, some say it northwood is better in turns of heat but prescott give off a higher speed and some even say there both are the same.. Any ideal how these two compare to each other ? One more thing I find the prescott is lower in price compare to the northwood, Why is that?

[Christian-Comp]

From what I've heard, Prescott has been discontinued and they are trying to sell them until they run out. I could be wrong about taht.

[demon_wolf]

The price difference is because as technology gets better,prices fall. A good example is Ram:

512 SD Ram PC100/133

512 PC2700 DDR

If you purchase both from the same company and similar performance (Not Expensive vs Budget) , they are nearly the same price, but the DDR far outperforms the SD Ram. The same is true for processors.

Basically, older technology can sometimes cost more to produce than newer technology.

As far as the Northwood/Prescott argument, it doesnt really matter. Just go for the highest FSB/ most Cache you can afford.

[johnnyis42]

actually, there is a huge difference between prescott and northwood.

northwood l2 cache= 512kb
prescott l2 cache= 1Mb

with the long pipeline and register shifting techniques the pentium 4 archetecture uses, there is a significant performace increase with adding more on die cache. take for example the original pentium 4 with only 256k ls cache, when AMD was dancing all around it with it's athlon xp, brand new at the time. it wasn't until the pentium 4 added the 512 cache with the faster bus speeds that it all of a sudden became a real gamming contender. the rest is history in any case.

also, if you read the reviews, the LGA packaged prescotts are just as good of overclockers, if not better than the northwoods. its just the availability of a motherboard with an AGP slot that can give you trouble.

intel may discontinue the socket 478 prescotts, but that is business as usual. the LGA 755 prescott chips will be around at least another year.

as for the memory pushing too hard, you can run the memory asynchronously from the system bus, but only at certain ratios. clock your memory down and raise the front side bus, see what you get. it may end up you run your memory less than 333 for a stable overclock, and you may also run it a little faster when you hit that sweet spot.