What's the difference between Single v.s Dual Channel?

[sikkony]

On the C2D platform, is there any real difference between Single v.s Dual Channel configurations for ram. I want to run one stick of XMS2 1GB on a E6400 on the 975X chipset. But I'm not sure if I should just get the other stick too. Besides the size difference, is there any performance difference that is substantial?

[thelostrican]

i would think that for benching there wont be that much of a difference, but in real gaming i know there is...

[sikkony]

i would think that for benching there wont be that much of a difference, but in real gaming i know there is...

Did you mean the other way around?

[uOpt]

It should be 3% at most, and less in games where you tend to be GPU limited.

Of course it look like more if you use stupid benchmarking such as running an old game at 150+ FPS.

[sikkony]

I thought it would be cool to match dual core with dual channel ram with dual GFX. Right now I have the former and the latter, not the middle. Was wondering if its worth, but 3% is nothing. I would of thought it was much bigger from the replies I saw on other forums on this topic.

[Kango]

I don't think the feedback so far is correct. Dual channel memory provides a significant performance boost. Agreed, in single channel mode the processor will run (along with the cache) at the specified clock rates. However, the memory bandwidth will be half the bandwidth that dual channel provides. So only half as much data will get from memory to the processor (or from the processor to memory).

It's easy to visualize. When in dual channel mode, the memory modules (let's call them A1 and B1, they are the memory slots of the same color) split the load. So while A1 is sending data to the CPU, B1 is in the process of accessing the next address. Then, when B1 is sending data to the CPU, A1 is in the process of accessing the next address. The same methodology exists when the CPU is sending data to the memory modules.

So, while the CPU is running at the same rate, it gets (or stores) half as much memory data in the same period of time. This will be very apparent when loading a game, for instance. I don't know that I could put an overall performance gain percentage to dual channel over single channel because it is very application dependent. However, I would guess that on average you will get about 25% improvement with spikes to 50%.

Example: I have a D805 running SETI@Home 24/7. I had it on a dual channel MB and was getting an average of about 1000 credits per day. The MB went belly up and I only had an old single channel memory MB to replace it with. All other components remained the same. The new average was just over 500 credits per day. But, it depends on the application.

If I've misunderstood the question, I'm sorry for going OT. But I believe you will get a noticable performance improvement running dual channel mode.

Kango

[thelostrican]

in games in general using single channel sucks big time, it feels like a whole other computer a lot of laginess in some taxing parts etc...

[uOpt]

I don't think the feedback so far is correct. Dual channel memory provides a significant performance boost. Agreed, in single channel mode the processor will run (along with the cache) at the specified clock rates. However, the memory bandwidth will be half the bandwidth that dual channel provides. So only half as much data will get from memory to the processor (or from the processor to memory).

Yeah but only when you read enough data in a streaming fashion. You theory is correct but just a theory.

For sure you can show us benchmarks that show a more than 2-3% speed advnatage for real applications? 2-3% is what all the socket 754 versus 939 benchmarks showed us.

So, while the CPU is running at the same rate, it gets (or stores) half as much memory data in the same period of time. This will be very apparent when loading a game, for instance. I don't know that I could put an overall performance gain percentage to dual channel over single channel because it is very application dependent. However, I would guess that on average you will get about 25% improvement with spikes to 50%.

Uh, no.

Example: I have a D805 running SETI@Home 24/7. I had it on a dual channel MB and was getting an average of about 1000 credits per day. The MB went belly up and I only had an old single channel memory MB to replace it with. All other components remained the same. The new average was just over 500 credits per day. But, it depends on the application.

That's really a speculative observation. Can you show us one single thing finishing in better than 2-3% improvement when going dual-channel or point to any benchmark (real applications only, of course useless memory bandwidth benchmarks improve a lot).

[Waus-mod]

i let you got to the market
and lets say you buy for 50 kilo's of products... how would you carry? 1 hand or 2 hands (2x25kilo)

[alexio]

I don't think the feedback so far is correct. Dual channel memory provides a significant performance boost. Agreed, in single channel mode the processor will run (along with the cache) at the specified clock rates. However, the memory bandwidth will be half the bandwidth that dual channel provides. So only half as much data will get from memory to the processor (or from the processor to memory).

The PCI-E slots take so much of the bandwidth between the memory and the chipset away that a doubling in bandwidth can never even closely be achieved when going from single-channel to dual-channel.

It's easy to visualize. When in dual channel mode, the memory modules (let's call them A1 and B1, they are the memory slots of the same color) split the load. So while A1 is sending data to the CPU, B1 is in the process of accessing the next address. Then, when B1 is sending data to the CPU, A1 is in the process of accessing the next address. The same methodology exists when the CPU is sending data to the memory modules.

That's the theory, however in realworld apps the difference is very small. This has to do with the very good prefetch of Conroe and also with the large cache.

So, while the CPU is running at the same rate, it gets (or stores) half as much memory data in the same period of time. This will be very apparent when loading a game, for instance. I don't know that I could put an overall performance gain percentage to dual channel over single channel because it is very application dependent. However, I would guess that on average you will get about 25% improvement with spikes to 50%.

That would be true if the hdd and SATA slot weren't the limitting factor. I'm pretty sure loading times are only effected by memory size. If not everything can be loaded in the memory the rest is put in the pagefile, which means an extra write on the hdd which costs time.

Example: I have a D805 running SETI@Home 24/7. I had it on a dual channel MB and was getting an average of about 1000 credits per day. The MB went belly up and I only had an old single channel memory MB to replace it with. All other components remained the same. The new average was just over 500 credits per day. But, it depends on the application.

That's correct, it depends on the application. In games for example there is no need for dual-channel. Big calculations can be loaded in the Conroe's cache and therefore the fps is not limitted by the memory, but rather by the CPU or GPU(s).

If I've misunderstood the question, I'm sorry for going OT. But I believe you will get a noticable performance improvement running dual channel mode.

Kango

Well, I guess I don't agree

i let you got to the market
and lets say you buy for 50 kilo's of products... how would you carry? 1 hand or 2 hands (2x25kilo)

50kg in one hand, so I can afterwards go to the supermarket and carry some beer with my other hand

[Kango]

I love this forum. The differences in opinion really stimulate thought.

OP (sikkony), Please let us know what you decide and provide us with some data from your experience. Buy two sticks. Install one. Run some tests. Install second stick. Rerun some tests. Differences??

[uOpt]

I love this forum. The differences in opinion really stimulate thought.

OP (sikkony), Please let us know what you decide and provide us with some data from your experience. Buy two sticks. Install one. Run some tests. Install second stick. Rerun some tests. Differences??

That will of course have the situation with two sticks come out better because it has twice as much memory.

It is better to insert the second stick into a memory slot that is not doing dual-channel with whatever slot the first stick is in. Doesn't work on DFI Lanparty but most anywhere else. Or disable dual-channel in the BIOS but that might not work reliably.