• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Single Channel vs Dual Channel

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Trance4c

Registered
Joined
Dec 3, 2011
Over the past few months I have been attempting to start working on other areas of my overclock now that I have a reliable, stable overclock on my Phenom II chip. I realize many people are now into their icore's and this is a bit dated, but I have never been on the bleeding edge of tech, it just costs too much!

First a run down of the system:
CPU: 965BE C3 overclocked 17.5x246FSB = 4.305Ghz Single Channel, 17.5x240FSB = 4.20Ghz Dual Channel, temperatures hit 60C on water, vcore 1.475 stable.
Motherboard: Gigabyte GA-MA785GM-US2H
Ram: 4GB G.Skill 8500 DDR2 1066 (800 native), with OC runs 984mhz @2.1v Single Channel / OC runs 960mhz @2.1v Dual Channel
PSU: Corsair TX650

What I am testing is to see if there is a benefit to Dual Channel mode ram configuration vs Single Channel mode when the Phenom's Memory Control is working in an unganged mode (i.e. 2 IMC lanes, 64-bit per channel).

In this case we have a motherboard with 4 memory banks, 2 per channel. For our single channel setup we put one 2gig stick in slots 1 & 3. This should produce a situation where each channel in unganged mode is handled independently. Understanding that bandwidth is reduced by the fact each stick is in single channel, the unganged mode allows the CPU to access each stick whenever possible without being addressed to a single core. This effectively creates the equivalent of a dual mode, but does it really scale that way compared to a true Dual Channel setup where a one 2gig stick is in slots 1 & 2?

The only catch - our Dual Channel test is also in an unganged mode, so we have effectively kept the Memory Controller working as two 64-bit units instead of together as a single 128-bit unit.

You will also notice that in order for my overclock to be stable in Dual Channel mode, I also have to reduce the FSB from 246mhz -> 240, giving up 100mhz of CPU speed in the process. This gives the Single Channel an advantage right off the bat, lets see what happens!

Right away we can see that even at the reduced speed the Dual Channel (Green) is out performing the Single Channel configuration (Red) in the FFT size calculations, the Dual Channel mode though is just barely slower in the Bit Trials which is to be expected as we have reduced the CPU clock speed by 100mhz. This reduced Bit Trial speed is great as we should expect to see this happen. I have also taken a few results from the GIMPS site just for comparison of the closest i5 and one with a massively higher clock speed with the i7.

Conclusion - as we all have read and should be aware, Dual Channel is surely faster even when attempting to mimic that ideal with the IMC working in tandem on Single Channel, it just does not keep up. In this case, we see that the best balance for the system is to give up the headroom on the CPU and keep the Dual Channel configuration.
 

Attachments

  • test.jpg
    test.jpg
    80.3 KB · Views: 3,259
Always good to keep testing things out. Especially when you experiment with better configurations for the hardware (IMC). Thanks for posting this up.
 
Back