What is HyperTransport?
From the HyperTransport Consortium webpage is this following quote:
HyperTransport interconnect technology is a high-performance, high-speed, high-bandwidth, point-to-point link that provides the lowest possible latency for chip-to-chip links.
In Athlon 64 motherboard implementations HyperTransport is set relative to the LDT speed but is not directly associated with processor performance; increased HyperTransport speed means that bandwidth between components increase and conversely, latency should decrease. HyperTransport has been a source of confusion since the introduction of the Athlon 64 and hopefully this helps clear up the distinction between HyperTransport and the LDT bus.
In the following tests, I benchmarked the system at HT bus speed multipliers of 5x, 4x, 3x, 2x, and 1x. The results for the tests performed at 5x, being stock-clocked results, are used in subsequent testings where we compare stock results to other results.
It's apparent that XviD encoding isn't dependant on the HT link at all. There was absolutely no change in the frame rate going from 1000 MHz to 200 MHz. Obviously this task is primarily delegated to the CPU and RAM, which are independent of the HTT link, and hard drive I/O, which doesn't come close to saturating even a 200 MHz HTT link.

The synthetic SiSoft Sandra scores take a slight hit as we drop the HT link speed, with ALU performance taking the largest hit with its measly ~1.5% drop. All in all, the scores remain pretty constant, as the performance hits are quite insignificant (although reproducible).

MP3 encoding, being a similar task to XviD encoding, also exhibits virtually no performance differences as we lower the HT link speed. Only at the very slowest speed of 200 MHz do we see a difference of 1 second.

In total, Half-Life 2 performance drops a total of 1 FPS. Nothing to see here, move along.

Much like Half-Life 2, Doom 3 dropped a total of 1 FPS.
All in all, we can see that a reduced HT link speed has an extremely limited impact on overall system performance, at least when it comes to our specific tests. When running at a speed of 1 GHz, the total theoretical bandwidth of the HT link is 6.4 GB/s. At 200 MHz, that theoretical limit drops to 1.28 GB/s. Imagining a block diagram of the nForce4 platform, it's obvious that this is sufficient for most of the devices that depend on the HT link to bring data to the CPU and memory, namely the Ethernet interface and hard drives. With AGP 8x having a peak bandwidth of about 2.1 GB/s, the 200 MHz 1.28 GB/s HT link is still a tad faster than AGP 4x (which we all know has yet to be saturated in most cases), meaning that the video subsystem ought not be too bottlenecked by 200 MHz either. One of more contentious issues after the debut of the Athlon 64 was the VIA HyperTransport implementation on the K8T800 versus the nForce 3 150 implementation. VIA's HyperTransport was faster and wider (a different number of bits transfered per clock is adjustable) but it looks like a lot of those criticisms were largely unfounded as NVIDIA suggested at the time.