CAS/Frequency simple question

[insolito]

Hello,

Is the CAS/latency division really a number to compare DDR4 memory performances?

Example, with both DDR4 memory:

MEM1 = @2400MHz CL16 memory
MEM2 = @3600MHz CL14 memory

So,

MEM1 CAS/Frequency = 0,00666
MEM2 CAS/Frequency = 0,00388

My question is: Since 0,00666 / 0,00388 = 1,71, It means that MEM2 is 1,71x faster then MEM1?

Thanks!

 

[Eximo]

That math doesn't really add up.

In terms of bandwidth, it would be 50% faster.

2400Mhz at CAS 16 would be 13.33ns latency (I think you forgot the Double Date Rate part in your calculation, the real frequency is 1200Mhz)
3600Mhz at CAS 14 (which is very optimistic) would be somewhere around 7.75ns latency.

So you don't really divide those into each other to get a result. 3600 at 14 is way better. Higher frequency lower latency.

 

[insolito]

But, if i divide 3600MHz per 2400Mhz i got 1.5 with equals with 50% as you said (In terms if bandwidth)

So, if i divide 13.33ns per 7.75ns i got 1.72, then it is not 72% slower? (In terms of latency)

 

[Eximo]

That should be 8.75 a slight typo.

If you are comparing single statistics as raw numbers, yes. CAS latency is literally a measure of clock cycles, a count. (1 / frequency)* CAS Latency 1 / 1,600,000,000 = .625 ns x 14 = 8.75ns That is how long it has to take to read one bit. But overall you aren't going to see a 72% increase in performance if that is what you mean. It really depends on what the memory is doing, if it is on target to read the right bank, if what is in the memory is the right thing, etc...

The short answer is it doesn't matter all that much. If you want fast memory, lower latency at a higher frequency is good.