well, it has been a while since I have posted to this blog. Since the last post, I have mostly been working on my radiology related computer vision program which I presented at RSNA about a month ago. It went pretty well.
Since that time, I have decided that a single workstation setup is insufficient for large scale testing of my program on multiple CT angiograms. So, I have started building a computing cluster in my apartment. Since I have a bunch of graphics hardware sitting around (2x 280 GTX and 2x Tesla), I am setting up a single workstation for each card. I have the first two nodes setup already and have successfully launched a test program across the the network.
The ultimate goal is to have all four nodes up and running by the end of the month. That way, I can launch 5 instances of the program simultaneously and cut down on the run time for testing new builds on 12-24 scans.
For the processors, I went with the 45W AMD dual core chips running at 2.6GHz. I have been surprised at how well they perform despite being fairly low on the L2 cache side. It seems that the only thing that really changes my run time is increasing RAM (beyond 4GB) and increasing the GPU power. The newer NVIDIA cards seem to have relxed the memory coalescing requirements compared to the G80 series which has benefited me greatly.
Well, time to go... let's see if i can keep up with posting more frequently these days.