Dual Channel

Its best to have matched pairs or atleast 2 that have close spec. But I have run Dual Channel on a 939 board with one DDR 333 and a DDR 400, it just back clocked to 333 but it ran in dual channel. But if you dont have matched pairs or atleast close theres no guarantee it will work
 
The most commonly seen advice is for using identical sized dimms of the same type of memory. You wouldn't want to mix two different brands with different CAS Latencies. If you are running a 1gb dimm along with a 512 you would install the 1gb in the first slot as well as a faster dimm if mixing two dimms seeing different speeds.
 
PC eye said:
You both missed one thing there when I mentioned that this was one way the dual channel mode was explained back 3 or 4yrs. ago when the dual channel boards were first coming out. The online information seen at the dictionary is far more detailed and updated there.
Click to expand...

What was the missing thing? That the information back then was wrong or...?
Anyways, he asked how it worked, so why give some out-dated and wrong information?
 
When researching the avaiable information a few years back that was how things were explained when comparing the differences seen between single channel boards and those with the dual channel mode then becoming available. The initial comparasons pointed out the problems with bottlenecks when trying to multitask on the old "single core" cpued single channel boards. The secondary channel being opened up allowed for background apps and basic services to run there when some large app was tying up the primary. That would be similar to articles seen on Vista prior to it's release then.

For one actual older comparison made with an nForce 2 chipset with comparison charts the article seen at the link here goes into describing the overall effect when comparing the single channel to the dual channel mode. http://www.digit-life.com/articles2/nforce2-1vs2channels/

Another old article that hots the spot far better has this to say about "dual channel memory"? as well as the dual channel mode seen from the nForce 3 chipsets up.

"Take note that the memory isn't dual channel, the platform is. In fact there is no such thing as dual channel memory. Rather, it is a memory interface composed of two (or more) normal memory modules coordinated by the chipset on the motherboard, or in the case of the Athlon64 FX and Opteron processors, coordinated by the integrated memory controller. But for the sake of simplicity, we refer to DDR dual channel architecture as dual channel memory." http://www.dslreports.com/faq/9216
 
The secondary channel being opened up allowed for background apps and basic services to run there when some large app was tying up the primary.
Click to expand...
It didn't work that way. Dual channel memory has always been about the memory controller using 2 64bit paths to talk to the RAM. The OS addresses a range of memory and the controller takes care of where it physically is.

The article that says the memory isn't dual channel the platform is is correct, RAM isn't dual channel, it's interface is 64bits wide the memory controller has 2 64bit interfaces. That allows it to talk to 2 sticks at the same time.
 
Where in all that does it say what you said about the Primary data going to one channel and secondary data going to the other. And nobody said there was Dual Channel memory, of course its a memory controller capability. Stay with us now dont wonder off.
 
The old article seen at the time made that statement about "forcing background apps and services onto the secondary channel" in that type of statement. That likened the dual channel mode to what happens with two memory modules installed with one filling forst and then going to the second on the old single channel boards in that sense. That was seen in a 2004 article or blog at the time.

A far better reference on memory recently published leads off with:
What every programmer should know about memory, Part 1

http://lwn.net/Articles/250967/
 
"forcing background apps and services onto the secondary channel"
Click to expand...
I can't seem to find anything similar to that in any of the articles you've linked so far. Can you find the old site or whatever it was and link it? To my knowledge it has never worked that way.
 
I've been trying to find that one or a similar article from 2003-4 with no results. The problem is these blogs or articles are seen for so long and then even if you have a link saved the domain changes hands or the article is eventually moved or removed or after this amount lost period.

I came across that one day while searching for something else then. You won't find anything that in the links here since those far more udated and specific. The search tern at the time may have been dual channel architecture or similar leading to one odd rare one time article.
 
It's never a good idea to base all your research on some random article found on some random website, that way you will just end up getting wrong information
 
The one thing people forget often is that boards prior to Socket 478, 939, 775, and AM2/AM2+ only saw a simulated not true dual channel mode. The AMD 754s were single channel only. The article run into then was most likely based on the older simulated dual channel and didn't use the terms memory controllers or 64bit/128bit then.

Besides probably pming Cromewell with the link if that one is ever run across again there's one 5 page article that sums up a few things I try to explain to people when going to buy more memory titled
Dual Channel - Why dual channeling makes a difference

Another factor to consider is the type and quality of memory modules you get. It is highly recommended that you buy two of the same modules, and even better, at the same time. Manufacturers over time can switch what chips are in the memory. Make sure the RAM are the same speed and memory latency. You don’t want to buy a high speed or low latency memory only to have it washed out by the other stick, which isn’t as fast or possesses a higher latency. When running in dual channel each stick must perform exactly the same, so they must have the same speed and same latency.

That was an extract from the article found at http://www.devhardware.com/c/a/Memory/Dual-Channel/
 
The article doesn't cover it all. Depending on the memory controller, several configurations are possible. For example 1024+512+512, which gives an equal amount in each channel
 
PC eye said:
The one thing people forget often is that boards prior to Socket 478, 939, 775, and AM2/AM2+ only saw a simulated not true dual channel mode. The AMD 754s were single channel only. The article run into then was most likely based on the older simulated dual channel and didn't use the terms memory controllers or 64bit/128bit then.
Click to expand...

I used to build all kinds of Socket A boards with the nForce 2 chipset that was Dual Channel. What are these chipsets your talking about that has simulated Dual Channel?
 
StrangleHold said:
I used to build all kinds of Socket A boards with the nForce 2 chipset that was Dual Channel. What are these chipsets your talking about that has simulated Dual Channel?
Click to expand...

The older Asus boards I put into builds here saw nForce 2 chipsets. so? The older boards were still never to see full support for the dual channel mode. The 754s skipped that stage altogether for the 939s to see actual dual channel put to use.
 
PC eye said:
The older Asus boards I put into builds here saw nForce 2 chipsets. so? The older boards were still never to see full support for the dual channel mode. The 754s skipped that stage altogether for the 939s to see actual dual channel put to use.
Click to expand...

Tell me how the nForce 2 was not true dual Channel or show me a link that says how its not

The nforce2 platform has two 64bit memory controllers (which are independent of each other) instead of just a single controller like other chipsets. These two controllers are able to access "two channels" of memory simultaneously. The two channels, together, handle memory operations more efficiently than one module by utilizing the bandwidth of two modules (or more) combined. By combining DDR400 (PC3200) with dual memory controllers, the nForce2 could offer up to 6.4GB/sec of bandwidth in theory. It is also possible for DDR Dual Channel architecture to reduce system latencies and timing delays that inherently occur with one memory module

Where talking Dual Channel why do you keep bringing up 754 when you know there not Dual Channel. What are these chipsets that are simulated dual channel, since you say they exist you have to know what they are.
 
If you actually looked you will find plenty of people stating that. That was one main critism about the Socket A nForce 2 boards when the 754s were first out. Some were going to wait for the next line of board since according to them it wasn't true dual channel. The reference to the 754s was just that since those were strictly single channel. :rolleyes:
 
In other words you cant find one, lets just drop this because your not showing anything to back up what you said.
 
StrangleHold said:
In other words you cant find one, lets just drop this because your not showing anything to back up what you said.
Click to expand...

I could easily say the same when you state that HyperThreading is strictly AMD not Intel. :rolleyes:

"The appetite gap even affects relatively high-end chipsets like Intel's 845PE, which supports the latest Pentium 4 chips and Hyper-Threading technology, but is limited to the 333MHz memory speed of DDR333 (a.k.a. PC2700). The latter type of memory is a great match for a 333MHz-bus Athlon XP, but pairing it with a 533MHz-bus Pentium 4 is like using a Porsche to drive in city traffic." http://www.cpuplanet.com/features/article.php/1587771

Note the date on the article seen there was 2/19/2003.
 
You must log in or register to reply here.
Back
Top