I currently work in the area of computational fluid dynamics (CFD) and am running a numerical computer code. At present I run the code on a 3GHz Pentium 4 'Northwood' (800MHz FSB) CPU.
I am considering changing my 3GHz Northwood machine to a dual Xeon 'Nocona' (800MHz FSB) machine at the same clock speed.
Both CPU's have the same FSB and I plan on using the Asus NCCH-DL motherboard based on the i875 chipset for the Nocona. (I currently use an MSI board based on the i865PE chipset).
All other components in my system will remain the same so I think the main factor is going to be the CPU (the chipsets are similar).
If I run the code as a serial job (since it does not always run in parallel) on a single 3GHz Nocona CPU can I expect similar (or better) performance compared to a run on my existing 3GHz Northwood CPU?
I was just wondering whether there is any major architectural difference that would cause a performance decrease with the Nocona. I think that this is not the case but I can't find benchmarks to back this up.
Major architectural differences from Northwood to Nocona include: most cache sizes doubled more Write Combine buffers increased pipeline depth (number of clocks start to finish)
A majority of single thread applications are expected to run slower on Nocona than Northwood at the same clock speed. The expectation has been that a Northwood would be upgraded to a higher clock speed Nocona, or that the customer wants to use the 64-bit extensions.
If your application is not held up by long latency operations (e.g. divide, sqrt) and needs the additional cache, or needs the additional WCBs, it could run faster on Nocona.
I know nothing of the relative merits of the motherboards.
Sorry, don't know of any answer which could be presented as more than a personal opinion.