High speed encoding - memory - disk - speed issues

I have an application that Im writing that captures video and then converts it, twice. The capture is real-time, the first convert is 3x real-time. The second convert is actually not occurring on the system Im testing right now because it's intel specific and Im testing a dual cpu dual core opteron, so basically I fake that part of the process by just copying the source file to the target files location.

When I run this app on "lesser" systems it seems to run without issue. By lesser I mean dual-core PentiumD 830 (300Mhz). When I run it on a dual xeon 3.4Ghz I have "some" issues. When I run it on the dual cpu dual core opteron, which blows the xeon away in terms of performance, I get alot more undefined errors trying to create the output file.

Im going to throw a 150G raptor and another 2G of memory at it as a first attempt to resolve the issue. The real solution may lie in being able to "see" what the work flow is like as it flows throguh the memory and disk subsystems. What I would love to hear about is a way to monitor the subsystems so I can see where potential bottlenecks are occurring. Like what my streams are doing as they are generated by the application in memory and stream to the 250G Sata2 output drive, or the sata2 Raptor when it gets here tomorrow.

Beyond that I'd love to hear anyones ideas on how best to prevent bottlenecks in bleeding edge systems. Im currently using pc3200 CL3 DDR ECC Reg 400 Mhz memory on an tyan K8N-DL board. Im running XP64 and seeing 4 execution cores that are all running at about 90% during the encoding process.

I have no proof that the problem is realted to speed, just the evidence of running the same app on a range of platforms.

regards,

John