RAM timing comparison benchmark

[turdferguson79]

I know Tom's Hardware is frowned upon over here, but I post this Tom's story as a matter of interest to those, like me, who always wondered what concrete performance advantage there was to buying low latency memory. I couldn't find any good numbers at the time that I was shopping around for memory for my nforce2 board. Here is a comparison that Tom's did that presents some numbers comparing different ram timings on varioius platforms.

 

[Bop]

I'm suprised he didn't do 2-2-2-11 for the Athlon XPs. That is the best timing for RAM on nforce boards due to the anomaly. I got a small performance boost from going to 2-2-2-6 to 2-2-2-11.

 

[turdferguson79]

Anomaly? Do explain. I've heard of people doing that timing, but could never figure out why.

 

[Nazo]

Wasn't it a bug that made it say 11 when it really meant something faster (5 maybe?)

Anyway, I'm surprised to learn that using lower latency memory with lower latency settings doesn't matter so much anymore, especially if one doesn't have a P4 apparently. I always figured such a thing would be quite important. I was kind of annoyed that my memory was made for the active to precharge timing being 6 rather than 5 (so it performs worse on 5) since I figured it would be faster, but now I see there would be a very minimal difference.

 

[Drisler]

Originally posted by turdferguson79
Anomaly? Do explain. I've heard of people doing that timing, but could never figure out why.

11 (generally) does better than say 5 because the nforce2 doesn't make 'internal' adjustments like other chipsets...when you input your tRAS value. That's why this 'anomaly' only occurs on the nForce2s.



eh...maybe Tom's trying to justify why they used horrible timings for A64 3400+ and nice XMS for the P4 in that last review?