Now running Matlab (latest release) analysis on huge data sets. I have been using the 3GB switch to get everything to work under winXP, and now have a dual boot setup with openSuse 10 64 bit to take advantage of the 8GB of ram in 64bit OSs. (Large neural data sets, looking at informational theory analysis as well as signal processing applications to analyze low frequency rhythmic activity.) Small differences can make a big deal - right now I have a AMD 64 X2 dual core 4600+ with 8GB of DDR2-800 memory, running through one subject's data is a 3-5 hour run. (And not really willing to risk overclocking... sorry)
Looking at making a dedicated analysis machine, and was wondering about the CPU comparison charts. Which would be the best comparison of the available options to judge performance/cost for this type of application?
If there isn't a good direct comparison, any one know how to get a "scientific computing" comparison rating added? Many of us make our own machines, or at least look at these issues and resources like Tom's before choosing a premade system.
Basically you should go for the highest clock CPU you can afford on a 45nm wafer.If your thinking Oc'ing will render your system unstable, don't. If you can afford it buy some ECC ram, people mistake oc instability for memory address errors most of the time. If it can handle prime for 6 hours, not much is left that can crash it. I'm also betting the highend D5400xs would be perfect for a Matlab application.