• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

What so good about dual channel?

holycpu

Member
DDR400 Runs at 200 Mhz with x2 multiplier. Hyperthreading or Dual channel gives it 800 Mhz FSB.


So I am assuming that dual channel will give better benefits to Pentium system.

Then, will dual channel DDR400 gives 800 Mhz FSB to AMD64 system as well? OR dual channel has no effect on AMD64 system; therefore, you only get DDR400 speed, which is Runs at 200 Mhz with x2 multiplier?

Thank You !
 
no no my friend you have it all wrong

Check this Link

i decided to give a massive explaination to another user about all things bright and beutiful (well everyting you need to know anyway), this should clear some things up for you.. dont worry everyone has to learn

mine is the 13th post down
 
Originally posted by: RichUK
no no my friend you have it all wrong

Check this Link

i decided to give a massive explaination to another user about all things bright and beutiful (well everyting you need to know anyway), this should clear some things up for you.. dont worry everyone has to learn

mine is the 13th post down

wow.........so hard to understand (not that your explaination is not good but I am just stupid ...... )

I am willing to learn if you are willing to teach 🙂 !!!

I will read your thread again to try to understand it more.

 
Dual channel doubles the bandwidth. P4's are more bandwidth hungry so this benefits them more than AMD chips.

Dual channel has nothing to do with hyperthreading or the P4's 800mhz FSB.



 
ArsTechnica has some GREAT articles that explain how processors work which will help you understand the benefits of Hyper-Threading and dual core processors.

Here are some links, in the order I suggest you read them in...

Pipelining: Part 1
Pipelining: Part 2
Multithreading, Superthreading, and Hyperthreading
Hyperthreading and how/why Intel uses it (Prescott specifically)
AMD's Hammer (K8, Athlon-64, Opteron) Architecture (with this one, keep in mind that the Athlon 64/Opteron architecture is closer to the Pentium M thant he Pentium 4, so anything stated about the Pentium M will generally be true for the Athlon 64 in comparison to the Prescott)

*EDIT* x86-64/64-bit processing This explains what 64-bit vs. 32-bit means, and also talks about AMD's implimentation, x86-64/AMD64.
 
Originally posted by: Megatomic
LoL, you guys flooded him.

See you tomorrow holycpu. 😀

I love those Arstechnica articles. 🙂 The only bad thing about them is some of them are so long that people are put off by that and don't read them all, just sorta skim them. They are long, but they're packed with good info. So take the time to read them!
 
Originally posted by: holycpu
Originally posted by: RichUK
no no my friend you have it all wrong

Check this Link

i decided to give a massive explaination to another user about all things bright and beutiful (well everyting you need to know anyway), this should clear some things up for you.. dont worry everyone has to learn

mine is the 13th post down

wow.........so hard to understand (not that your explaination is not good but I am just stupid ...... )

I am willing to learn if you are willing to teach 🙂 !!!

I will read your thread again to try to understand it more.

In basic ...

Forget about hyperthreading and 800fsb this is nothing to do with what is effective (real)

If you look on the link that i have provided which i wrote for another user, and look under what i wrote for "Quad pumping" this is where you get 800fsb from an intel chip(theoretically)..

Dual channel - all this means is that the motherboard supports two DIMM slots/channels for memory, each channel is governed by the MEM controller and the bandwidth from the RAM, which is measured in MB/s, so with the standard PC3200 RAM you are getting 3,200MB's, if you were to bump up the FSB/HTT from the standard 200Mhz to 250Mhz you will get RAM bandwidth of 4,000MB's, which you might be farmiliar with as PC4000 RAM. Now each channel in AMD's case is controlled by the mem controllers on the CPU die itself, as apposed to the northbridge for the intels. Each mem controller is 64bit (this is broken down into 2x32bit, which means 32bit addressing upstream and 32bit addressing downstream (Communication from the ram to CPU and vise versa) 32bit is the addressing in which we run our OS's on aka windows xp which is 32bit, 64bit OS's is a whole different ball game, i can't be bothered at the mo to talk about that). Now basically if you add two mem controllers aka AMD's 128bit mem controller, which is 2 x 64bit mem controllers, you then get two individual channels for mem to work on, so if you then put in 2 x512Mb sticks (at PC3200 spec) into single channel like on the socket 754 athlons you get bandwidth of 3,200MB's no matter how many sticks you stick in, but with the new Athlons/FX's if you put 1 x512Mb PC3200 in the first channel, and then 1 x 512Mb pc3200 stick in the second channel, you get a total bandwidth of 6,400MB/s, which AMD does not fully use down to the high effciency of the onboard mem controllers, therfore AMD chips are more inclined to have lower latancy preferabley CAS 2 .. but depending on what RAM you have and what HTT you run this is not always possible....

You will find that when running dual channel if in one channel you run poor spec PC2700 and in the other channel low lantency PC3200, the MEM will always default to the Lowest common denominator, therefore the poor spec pc2700. Also you can keep in mind that when running dual channel you do not have to run base pairs, therfore you can run different specs on the different channels with teh ofcourse the sacrifice to the lower spec RAM used, on the same channel you most often then not have to run the same size and spec RAM (you will see nowdays motherbordas with 3 DIMM slots when running single channel, and dual channel has two sets of two DIMM slots.

Now the reasons that the Intel chips are bandwidth hungry is because the FSB has to go through the northbridge before it goes through to the CPU, therfore you can see it as a highway/motorway, where you have four lanes from the RAM to the northbridge then 2 lanes from the Northbridge to the CPU, therefore cutting down RAM performance durastically... Therefore AMD's solution to use on die MEM controllers cuts out the inafficient use of mem controllers on the northbridge and gets controlled directly on the CPU core isself... (thats why AMD are better 😛)


Hope this helps 🙂







 
Sorry i previously stated that hyperthreading is not (real), what i ment by that was just a marketing ploy like 800FSB...


that was a mistake it is real, and it works as a virtual dual core proc, but no help when it comes to gaming (at the mo), only multitasking aka more threads for the CPU to process at one time...

so unless you use CAD and 5 billion other video apps this is not going to help (sorry for the exaggeration)
 
Now the reasons that the Intel chips are bandwidth hungry is because the FSB has to go through the northbridge before it goes through to the CPU, therfore you can see it as a highway/motorway, where you have four lanes from the RAM to the northbridge then 2 lanes from the Northbridge to the CPU, therefore cutting down RAM performance durastically... Therefore AMD's solution to use on die MEM controllers cuts out the inafficient use of mem controllers on the northbridge and gets controlled directly on the CPU core isself... (thats why AMD are better )

Not to mention that the long pipeline of the Pentium 4 suffers more from pipeline stalls. Having more memory bandwidth helps to get data to the CPU to reduce the number of pipeline stalls.

That's also where Hyperthreading comes in. It's not necessarily the fact that the Pentium 4 is executing two threads at the same time, but that it CAN... so when the first thread stalls, the second thread can immediately start using the execution units left open when the first thread stalls.

When you look at all the features of the Pentium 4 and it's chipsets, they all have the same goal of keeping the pipeline full. Quad pumped bus - more data to be processed so the pipeline doesn't stall... dual channel RAM - more data to be processed so the pipeline doesn't stall... Hyperthreading - prevent pipeline stalls by executing two threads at once so when a thread stalls, the other thread can take over so execution units don't go unused as clock cycles tick by.
 
Originally posted by: Jeff7181
Now the reasons that the Intel chips are bandwidth hungry is because the FSB has to go through the northbridge before it goes through to the CPU, therfore you can see it as a highway/motorway, where you have four lanes from the RAM to the northbridge then 2 lanes from the Northbridge to the CPU, therefore cutting down RAM performance durastically... Therefore AMD's solution to use on die MEM controllers cuts out the inafficient use of mem controllers on the northbridge and gets controlled directly on the CPU core isself... (thats why AMD are better )

Not to mention that the long pipeline of the Pentium 4 suffers more from pipeline stalls. Having more memory bandwidth helps to get data to the CPU to reduce the number of pipeline stalls.

That's also where Hyperthreading comes in. It's not necessarily the fact that the Pentium 4 is executing two threads at the same time, but that it CAN... so when the first thread stalls, the second thread can immediately start using the execution units left open when the first thread stalls.

When you look at all the features of the Pentium 4 and it's chipsets, they all have the same goal of keeping the pipeline full. Quad pumped bus - more data to be processed so the pipeline doesn't stall... dual channel RAM - more data to be processed so the pipeline doesn't stall... Hyperthreading - prevent pipeline stalls by executing two threads at once so when a thread stalls, the other thread can take over so execution units don't go unused as clock cycles tick by.

True true ... the analogy goes (with the pipelines) ...AMD wide short conveyer belt?. and Intel thinner VEEEEEEERY long conveyer belt (that is how intel achieve their faster clock cycles on their chips) although i don?t know how much voltage the Prescott?s take to reach 3.8 Ghz and i bet they get quite toasty (3.8 Intel?s highest clocked proc)

what Jeff7181 has stated is basically what hyperthreading is in a nut shell...(yet again another technology to componsate for the design flaws of the prescott .. what a shame 🙁

HAHA ... yeah right there are losers and winners, and i am sorry to say that AMD are clear winners 🙂

 
Originally posted by: Jeff7181
Originally posted by: Megatomic
LoL, you guys flooded him.

See you tomorrow holycpu. 😀

I love those Arstechnica articles. 🙂 The only bad thing about them is some of them are so long that people are put off by that and don't read them all, just sorta skim them. They are long, but they're packed with good info. So take the time to read them!


I have so far only read the Pipelining Articles, but they are packed with information. Thank you for showing me these great articles!
 
Back
Top