ASUS T-Topology Explained - Higher Memory Overclocking
Category : Tutorials
Published by Marc Büchel on 02.08.12
With the Crosshair V Formula-Z and the Sabertooth 990FX R2.0, ASUS is going to release two new premium AMD motherboards, one for the TUF and the other for the ROG series. You can expect some changes and also some new features for AMDs latest FX Bulldozer CPUs as well as AMDs upcoming Piledriver processors. The biggest improvement concerns the implementation of ASUS' T-Topology on AMD motherboards. The manufacturers latest Z77 motherboards already incorporate this new feature and ASUS promises that overclockers can reach higher memory frequencies, when placing four DIMMs. ASUS was also working on the power delivery to the DIMM slots which should also help reaching higher clock frequencies. In this article we're going to explain what is special about ASUS' T-Topology technology and why it helps with overclocking. 
What is T-Topology?
Imprving a motherboard and adding new innovations to it isn't easy at all. It takes a substatial amount of time for developers to implement new designs. In the case of T-Topology ASUS engineers have been working for more than a year until they were finally able to release products which incorporate this new feature. According to the manufactuere the big advancement should be higher overclocking capabilities as well as improved compatibility.Since summer last year ASUS engineers have been working on T-Topology for DDR3 on AMD motherboards. The basic idea behind was to create shorter communication ways between CPU and memory. Reaching the last of four DIMMs in a setup has the consequence that a signal has to be routed through all three DIMM which are in front of the fourth and afterwards it has to be routed back to the CPU the same way. Therefore it takes more time to receive data from that last DIMM than from the first DIMM.
In this case you can keep a rather simple equation in mind: Time = Distance / Speed
What's the benefit from T-Topology?
Electrons travel with the speed of light and the speed is also constant, so the only thing that adds latency is the way electrons have to travel. Is the distance bigger, then the latency is higher and is it shorter, then the latency is lower. Therefore it sounds quite logical that ASUS wanted to make this travel distane shorter. Of course this is a highly theoretical scenario and in reality engineers also have to deal with clock skew, the quality of the integrated memory controller, the quality of the DIMMs, consistency of voltages and quite a few other things.Instead of the serial solution which has been used until today ASUS went for a parallel one. Placing DIMMs in all available slots doesn't have such a big impact on latency as well as compatibility like it used to have before. Usually serial data processing is much more vulnerable to synchronisation issues. Furthermore ASUS is promising that they increased memory overclocking capabilities by up to 15 percent with T-Topology. Therefore it is well possible that our well knonwn and excellent memory overclocker Roger "splmann" Tanner, Christian Ney and Hiwa Pouri could break some of their own world recors again.
In order to get an idea how complex the implementation of such a new fundamental technology is we can present to you an image from an ASUS engineer. On this picture you can see different layer tracings. Each and every route has to be checked and if lengthened and shortened if needed.
What about overclocking?
ASUS also provided us with a benchmark results from a Crosshair V Formula-Z at CL9 and a Sabertooth 990FX R2.0 at CL10:
How to best configure your setup?
If you choose to install two DIMMs, then you can use either the brown or the beige slots. Never the less you should prefer channel A (brown). The reason is really simple. Because of shorter signal travel distance you get lower latencies and therefore higher performance.If you have four DIMMs you want to install, it is quite obvious that you'll have to populate Channel A (brown) and Channel B (beige). You should make sure that you're using two dual channel memory kits and that you place those kits in the sockets with the same color.
Another thing we have to mention when it comes to memory overclocking is, that is can well be that the processor is the bottleneck in the system. Therefore it can be that you wont be able to reach the frequencies we mentioned here. On the other hand we can tell you that todays world record for memory frequencies, held by our overclocker Chrisitan Ney, is at almost 4 GHz.