nMP & memory performance observations with various mem configs

[bxs]

STREAM is a "bandwidth virus" benchmark designed to defeat all caches and measure the raw memory bandwidth of the system. It does nothing useful.

IMO, it is interesting for people trying to get into the Top500 Supercomputer list, but mostly irrelevant for anyone considering an Apple, Windows or Linux desktop system.

Most apps benefit from cache, and Intel is currently looking at 2 MiB to 2.5 MiB cache per physical core as the sweet spot. The GeekBench numbers show that is a good decision for almost all of the tests in GeekBench. There are probably some useful desktop apps that need extreme bandwidth, but not many.

One thing that I was happy to learn from this discussion is that AES encryption is one of the bandwidth intensive apps. I'm buying systems for an application gateway prototype which will use 20-core systems to do SSL (AES) encryption. I've learned that populating each system with 8 DIMMs is the way to go. (Some systems only need 32 GiB, so they'll get 8x4GiB.)

'Tis true that STREAM as it 'comes out of the box' is designed to avoid measuring the system's cache read/write performance. However, it can be modified to measure the cache performance if required.

For codes that are very memory intensive and cannot perform most of their calculations using the processor cache, then the processor would be performing many memory fetching activities (loading cache from main memory- RAM) and thus memory performance (bandwidth and latency) become extremely important for these type of codes.

There are many engineering & scientific codes that people run on Apple Mac systems. They aren't in any way interested in competing for being listed in the Top500, but do want the best memory performance for their Macs. The STREAM benchmark is a useful tool for them for establishing the best memory architecture to employ such as best way to deploy DIMMs in the available slots.