It all depends on what you are doing and where bottle necks are in your system. e.g. When I copy a file over my 1GB network vs. my 100MB network, why don't I see a 10 fold improvement in speed. Ans. My HDD is the bottle neck causing the delay which tops out at about 140 to 170MB so I go spend more $$ and buy the SSD. Point being is if you are say running a big math calculation, then yes the latency will make a difference.
Thanks for the reply. I had a feeling that it makes a different to the people on this forum more than it does to the man on yahoo answers who is using his laptop for writing newspaper articles.
I have a sager with an i7 2.7-3.7GHz with a SATA III SSD (I think about 120 IOPS) - so there is a chance that the memory could be a bottleneck.
I don't do much math on my machine (the most math being done would be a physics engine in a video game, but that's mostly my graphics card's memory which is substantially faster already). But if my understanding of CAS Latency is correct than accessing data in arrays will be increased be the same amount that is the difference (in this case about 20%). That could make for a HUGE increase in speed.
What ram are you looking to get? what speed? 4 sticks? 2 sticks?
I'm looking to get 4 sticks of Kingston Hyper X 8GB @ 1.6GHz for a total of 32GB (I do CAD... just so you know before you tell me 32 is too much.)
I am also considering some G.Skill memory because I have used their stuff since I was in 7th grade and it has never let me down. The G.Skill memory I'm looking at has the same specs but its CAS Latency is 11 and no heatsinks -- I'm not sure if the heatsinks do anything or if they are just for show.
Wikipedia says that at 1GHz latency is equal to the initial time it takes to access the data in ns. So a latency of 11 at 1GHz would mean that it takes 11ns to access the data. This time decreases linearly with increase in frequency.
If my math is right the access time for 1.6GHz ram with latency of 9 is going to be about 3.2ns. If the latency was 11 the initial time would be more like 4.4ns.
100%*(4.4ns-3.2ns)/3.2ns = 38%. I think that would be the gain in access time. That is enough to sell me right there. Can someone else confirm that this is correct?