|
||
|
||
|
The 512mb Tests The first tests I ran were designed to see if the number of installed DIMMs made any difference when it came to 4-way interleaving. I also wanted to be sure to isolate the amount of physical memory to make sure that it was consistent across all tests. To accomplish this, I settled on a physical memory size of 512mb. This made it easier to test because I could use four 128mb DIMMs, two 256mb DIMMs, and a single 512mb DIMM. Therefore the only thing actually changing from run to run was the number of DIMMs installed. I also wanted to verify that the interleaving was indeed running as it was supposed to be, so before running any tests I checked Sandra's "Motherboard Info". If you scroll down a bit in the results, you will see something like this:
The above image shows what you would see with four 256mb DIMMs installed (two 128mb banks per DIMM). We can indeed see that the memory speed is 134mhz and that 4-way interleaving is enabled. Also note that our memory timings are set pretty aggressively (2-2-2) which proves that George's Memory Interleave Enabler is doing its job. Cool... Everything looks in order, let's move on. SiSoft Sandra - Memory Bench
WOW! How depressing! Sandra is showing absolutely no improvements no matter the number or size of the DIMMs installed. Man... I was hoping to see something change. Maybe CliBench will tell another tale. Let's take a look:
Newp... Same story here. I guess we can make some pretty good conclusions about the affect of DIMM size and quantity when using 4-way interleaving. Well, ummm, there are no benefits to having multiple DIMMs installed. I guess the old overclocker's motto "One big stick is better than multiple smaller ones" could apply here. On the other hand, if you already have a few smaller DIMMs installed, you shouldn't worry about the addition of another DIMM or two affecting your memory performance. Looks like a total wash to me... Let's try something else. |
||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
|