Related Articles

The LDT differs from the HT bus as this specifies the speed at which the processor communicates with memory and the speed at which the processor works in relation to its external communication speed. HyperTransport as we covered before is better thought of as the bandwidth between components. Confusion in overclocking the Athlon 64 also stems from the fact that different manufacturers use LDT, HT, HTT, CPU Speed, CPU Clock, FSB as terms that are all somewhat interchangeable. In the following tests, the LDT bus speed was raised to specific values while dropping the CPU multiplier, in effect overclocking the LDT bus speed while keeping the CPU clock constant at its stock ~1800 MHz. The memory divider was kept at 1:1, which also means that the memory is being overclocked.

XviD encoding does not benefit much at all from the increased HT bus and memory speed. The first pass gained 1 frame per second, with the second pass gaining less than half a frame per second.

Interesting enough, even though we are keeping the total CPU clock constant at its stock ~1800 MHz, both the arithmetic and multimedia tests measured a slight drop in performance as we raised the LDT bus speed but dropped the multiplier. What's even more interesting is that even though we raised the LDT bus speed by over 25%, memory bandwidth increased only a measly 3.2%. I didn't even gain 200 MB/s in either test! It appears that memory bandwidth on the K8 platform is highly dependant on the CPU's clock speed, as opposed to just the LDT and memory bus speed.

There was no performance gain here. With CPU scores having gone down ever so slightly, and memory bandwidth scores having gone up ever so slightly, I'm not surprised that there is no overall effect on MP3 encoding.

Half-Life 2 performance sees a very vague gain, in the order of about 4.5 FPS.

Doom 3 exhibits the same ~4.5 FPS performance gain as Half-Life 2.

To summarize this set of results, raising the LDT bus speed and dropping the multiplier so as to maintain a close-to-stock clock speed on the CPU did nearly nothing for overall system performance. In synthetic CPU tests, CPU performance dropped just a tiny bit, and memory bandwidth performance went up a non-impressive 3.2%. LDT and memory bus speeds aren't all that they're cracked up to be if your CPU is not there to eat up the extra headroom given.