as example, i need SDRAM memory and have a choice for my machine, DDR3 PC3-10600 CL=9 vs DDR3 PC3-8500 CL=7, which is faster or better? CL # is important, but "10600" is "faster" than "8500"..... so which one is a better choice?
see, with SDRAM the "CL" is # of clock cycles before data can flow.
"The CAS latency is the delay between the time at which the column address and the column address strobe signal are presented to the memory module and the time at which the corresponding data is made available by the memory module."
so if mem "speed" (aka data rate) is the same but CL were different, then lower CL would be better choice. however, in this case the CL is slower on the memory that can do faster data rate.
in terms of raw CAS, CL9 vs CL7, CL9 is "slower" by 9-7/7 = 2/7 = 28.5%.
if there was only a single address made for read/write, i would presume the time lost from CL9 (waiting for data after address request) would be overcome by the faster data rate of the CL9 memory (assuming the data size was large enough to make up the latency diff). but in real world the memory controller is doing a gazillion memory address requests, hence the CL latency has to add up. question is though, will the faster data rate of PC3-10600 make up for the slower addressing latency of CL9 vs the slower data rate PC3-8500 with faster addressing of CL7?
CAS latency use to be an issue a few generations ago, but now with multi-core CPUs and large caches, most of that latency is hidden with pre-fetching or other threads that are able to continue working even if one thread stalled on a memory access.
But yes, latencies would be a huge issue on a single core cpu with little cache.
i my case the memory is for a HP Omni 100z (AMD Athlon II 260u dual-core). i am not sure of which chipsets are used on this mb as i dont yet have the machine, but this cpu has integrated DDR controller and max 3200MT/s bus.
Truth be told, the faster ram may well run with those tight timings if it was set down to 1066(8500) as well.
The faster ram will be more of a benefit for most users to be honest. There are some programs that will like tighter timings, but more will prefer the faster overall speed.
One thing to look at is since it is an HP system, you may not have any control over memory voltages, speeds(hp only sells it with 1066 stuff) and timings. As such some memory needs 1.65 volts to get to 7,7,7,20 timings and the hp system may well only give 1.5 volts.
You should contact hp to confirm this. For instance, I have memory that needs 1.65 to do cas 7 with 1.5 I can only get 8 or 9 with it.
This is something you have to watch for on prebuilt systems like that.
If that is an all in one, it may need sodimms(note book memory) as well
if the hp only allows 1066 memory, the 1333 stuff may just get clocked back down.
check with hp to ensure that the board will use the faster memory.
The reason i say this, is that i had an hp laptop that took ddr 333 and when i gave it ddr 400 it was unstable. It tried to run the 400, but just did not give enough voltage for the 2 sticks, either stick was stable, but together they would fail tests.
I ended up flashing one down to 333 and it worked ever since. I only needed to flash one since they all run the speed of the lowest stick.
In general higher memory speeds are preferred, but some older AMD cpus(mostly those with lower cache memory, did benefit from the lower timings). There are some instances where to lower latency would be better, but more often it will be higher speeds.
nuke, after getting some more info from a HP forum i went and checked AMD site again. their Athlon II X2 line supports DDR2 and DDR3-1066. i am guessing that 1333 will just get clocked down. so the better choice for me is the slower CL7 memory (which will run faster overall due to lower CL, etc).
lessons learned: not always the case to choose the "fastest" memory.