Interleaving and CAS - worth playing with?

[2ndchance]

I'm new to interleaving. Correct me if I'm wrong, but isn't it like telling your system divide up it's RAM calls between two or more banks? If so, I assume that you could not do it with a single stick?

Anyway, is this worth considering in building my next PC?

Also, if not OCing the CPU, will CL2 Ram make a real-world performance difference over CL3?

Thanks for any opinions!

 

[Renob]

Yes and yes turning on you 4way interleaving and running your ram at cas-2 make a nice difference.

 

[pm]

Depending on the memory configuration of your DIMM you can enable 2-way or 4-way interleaving within the DIMM. Each SDRAM DIMM consists of either 2 banks or 4 banks. 2-bank SDRAM DIMMs use 16Mbit SDRAM chips and are usually 32MB or less in size. 4-bank SDRAM DIMMs, on the other hand, usually use 64Mbit SDRAM chips though the SDRAM density may be up to 256Mbit per chip. All SDRAM DIMMs of at least 64MB in size or greater have 4 banks. Interleaving get's you a little improvement. There's no reason not to use it, and most BIOS's will enable it by default.

CAS2 is 3-4% faster than CAS3 in most real-world apps. Depends on the size of the apps and how predictable it is (if it's small and predictable, the cache will catch most memory accesses and whether or not you improve your memory performance won't matter).

Whether or not you will notice a total 5-6% improvement is debateable. You might if you are doing something like a 2 hour movie Divx encode which is largely uncacheable, takes a huge amount of time (10-20 hours), and a small improvement in performance could result in an hour's less crunching. But if you are just running games and Word, I don't think that you could notice any difference at all.

 

[Technonut]

Here are some benchmarks that I ran. The RAM is at 144MHz CAS2 Turbo, and the benchmarks were ran the same way with a fresh re-boot and background programs shut down.

Interleave Disabled 461/500

2 Way Interleave 512/538

4 Way Interleave 540/637

It makes a noticeable difference to me between Interleave disabled, and 4 way interleave enabled.

 

[pm]

I presume that this benchmark is the SisSoft Sandra memory benchmark? The numbers look about right anyway.

This benchmark is intended to pound the memory subsystem in order to try and maximize the peak throughput of the memory subsystem. It attempts to eliminate the effect of the CPU's cache by moving blocks of uncached data around and measured how fast it can move this data. Seeing a somewhat large difference doesn't mean that your system will perform substantially faster in the real world. In the real world very few applications are moving large blocks of uncached data around.

A more interesting benchmark would be BapCo, or ZD's Content Creation, or WinStone. Something that actually mimics real-world applications. I'm fairly confident that the numbers will drop to somewhere below 5% were you to do this.

Then there's the question as to whether or not even a 10% improvement in performance is "noticeable" in real world applications. If spell check takes 9 seconds rather than 10 sec., is this a noticeable improvement? If you get 66 frames per second in Quake3 rather than 60 frames per second, is this really noticeable?

There are very few applications that will see a substantial improvement in performance by enabling interleaving (SETI@Home is one). In fact, I have heard of a few cases in the real world where interleaving actually results in a performance hit rather than an improvement. But, like I said, there's no reason not to enable interleaving if you can. Usually it's a small, free improvement to memory subsystem performance.