Monday, August 08, 2005

Processor speed's influence on BlueTooth performance

Many of you run overclocking utilities on their boxen. When these programs first came up for the DragonBall CPU's, it was a dangerous adventure to run them. The sound output was altered, debug modes entered et al! Nowadays, the programs are more reliable and don't have too many unwanted side effects.
And this is where TamsPalm comes in. I use the bluetooth radio on my T3 excessively with WebPro. Since I usually play around with the CPU clock(old habit from IIIc days), it was interesting to see if WebPro or the radio implementation is the part that limits the transfer speeds. Thus, the dslreports speed test was executed multiple times at different clock speeds-and here are the results:

The data transfer speed follows the processor almost 1:1. So-faster CPU=more data transfer.

The latency is lowest at 200 MhZ. At 400 Mhz it actually grows a bit. This however-is due to BT beeing a wireless system IMHO.

The total time taken is what really counts though. It decreases greatly at first, but then decresases slower due to the latency. But overall, we dont run into limits of the radio yet...
Get the raw data as an Excel Table here.
What do we learn from this? The transfer rate of the bluetooth radio of the T3 scales up with the processor in a hyperbolic fashion. Thus, WebPro users definitely benefit from a CPU running at full speed or a bit higher-and WebPro seems to be the bottleneck at lower frequencies. The radio hardware is not guilty-for now. The total time is bogged down by the latency at the higher settings-so we start to see problems here already. But how does the CPU affect file sending? We'll cover that next week!