RAM Low CAS high frequency vs High CAS Low Frequency

[Parin712]

So I've been wondering for a while now.

What is better?

A ram with Lower frequency but high CAS.

or

A ram with High Frequency but low CAS.

Let me know what you guys think and if there is already a discussion/ video made for this I'd very much appreciate if someone redirects me to it. If now, well Linus, should know what his next video should be on. :D

Thanks.

[Raudi_]

High Frequency low CL is going to be faster every time 2133 CL10 is gonna be slower though than 1866 CL9

[Enderman]

 

 

It depends on the program.

[xg32]

1 hour ago, Parin712 said:

So I've been wondering for a while now.

What is better?

A ram with Lower frequency but high CAS.

or

A ram with High Frequency but low CAS.

Let me know what you guys think and if there is already a discussion/ video made for this I'd very much appreciate if someone redirects me to it. If now, well Linus, should know what his next video should be on.

Thanks.

generally u want frequency up to a point, for current ddr4, prolly 3200-4000, now b-die generally has cl14 3200, cl15 3600, and cl16/17 4000.

 

There's very little difference between the 3 (2-3% in maxed out gaming loads), but when able i choose the combo with the highest speeds. After doing some brief research months back there's little to no benefit going above 4000 atm.

 

As for where you definitely don't wanna fall under, i'd say cl16/3000 atm with hynix ram or cl16 3200 with ryzen.

 

The 2x16gb b-die cl14 3200 ram that went for 260-280usd during black friday on newegg was a steal.

 

It's really hard to see the difference and you'd need multiple kits of ram to test, for the average user there simply enough of a difference to care.

[straight_stewie]

1 hour ago, Parin712 said:

A ram with Lower frequency but high CAS.

or

A ram with High Frequency but low CAS.

Given your question, a higher frequency at a lower CAS will always be better than a lower frequency with a high CAS. But that's because we are comparing a best case scenario to the worst case scenario.

This is simply because the RAM CAS response time in nano seconds is simply: W(f/2)*c, where W() converts a frequency into a period of a wavelength in nanoseconds, f is the frequency, and c is the cas latency. Armed with the knowledge that a higher frequency always has a shorter wavelength and therefore, a shorter period, and multiplying by a lower multiplier greater than or equal to one will always yield a lower result than multiplying by a larger multiplier, we can conclude that RAM with a faster frequency AND a lower CAS will always have a faster response time than ram with a lower frequency and a higher CAS.

Are you sure this is the question you meant to ask? Normally one would consider a lower frequency ram with a lower cas versus a higher frequency ram with a higher cas (generally, as frequency is increased, cas is also increased).

[seon123]

Well, low CL is better, and high frequency is better. Therefore, a low CL, high frequency kit is always better than a high CL, low frequency kit. 

[Parin712]

On 12/3/2018 at 1:28 PM, straight_stewie said:

Given your question, a higher frequency at a lower CAS will always be better than a lower frequency with a high CAS. But that's because we are comparing a best case scenario to the worst case scenario.

This is simply because the RAM CAS response time in nano seconds is simply: W(f/2)*c, where W() converts a frequency into a period of a wavelength in nanoseconds, f is the frequency, and c is the cas latency. Armed with the knowledge that a higher frequency always has a shorter wavelength and therefore, a shorter period, and multiplying by a lower multiplier greater than or equal to one will always yield a lower result than multiplying by a larger multiplier, we can conclude that RAM with a faster frequency AND a lower CAS will always have a faster response time than ram with a lower frequency and a higher CAS.

Are you sure this is the question you meant to ask? Normally one would consider a lower frequency ram with a lower cas versus a higher frequency ram with a higher cas (generally, as frequency is increased, cas is also increased).

I think you are right about what I meant to ask, I didnt realise that.

It was supposed to be Low Frequency with low CAS vs a High frequency RAM with high CAS.

Thanks for pointing that out.

Silly of me

[Enderman]

Yeah it's usually a tradeoff between the two.

What performance you get isn't a simple equation that can be calculated, it will depend on the program.

Typically, as far as I've seen, a higher frequency makes a bigger difference than low latency.

[Parin712]

On 12/3/2018 at 11:35 AM, Enderman said:

 

 

It depends on the program.

That's really informative and right on point. Thanks!!

[Parin712]

13 minutes ago, Enderman said:

Yeah it's usually a tradeoff between the two.

What performance you get isn't a simple equation that can be calculated, it will depend on the program.

Typically, as far as I've seen, a higher frequency makes a bigger difference than low latency.

I am sure someone has figured out the best combination of the 2. We just need to find it.

[Enderman]

5 minutes ago, Parin712 said:

I am sure someone has figured out the best combination of the 2. We just need to find it.

As I said, it depends on the program.

You can't have both low cas and high frequency, as you increase one it decreases the other.

 

[pyrojoe34]

If you want to compare the latency the math is easy to do:

 

CL / (Frequency / 2) = Latency

 

CL is latency in # of cycles

Frequency in Mhz (divided by two since it is DDR) is million cycles per second

This equation gives latency in microseconds, multiply by 1000 to get nanoseconds.

 

So...

It's entirely possible that a ram kit with a higher CL value still has lower absolute latency than a kit with a lower CL value. For example: 4000Mhz kit with CL of 16 has a lower latency (4ns) than a 3000Mhz kit with a CL of 14 (4.7ns). Don't get too sucked into the CL values, they are not absolute metrics but relative metrics and need to be converted if you are comparing kits of different frequencies.

[Enderman]

15 minutes ago, pyrojoe34 said:

If you want to compare the latency the math is easy to do:

 

CL / (Frequency / 2) = Latency

 

CL is latency in # of cycles

Frequency in Mhz (divided by two since it is DDR) is million cycles per second

This equation gives latency in microseconds, multiply by 1000 to get nanoseconds.

 

So...

It's entirely possible that a ram kit with a higher CL value still has lower absolute latency than a kit with a lower CL value. For example: 4000Mhz kit with CL of 16 has a lower latency (4ns) than a 3000Mhz kit with a CL of 14 (4.7ns). Don't get too sucked into the CL values, they are not absolute metrics but relative metrics and need to be converted if you are comparing kits of different frequencies.

This is wrong.

It's straight out of wikipedia.

Even though the math is right, lower absolute latency does not mean that all programs will perform better.

 

People think it is this easy but some programs will perform better with high frequency ram even if the cas latency is high and the absolute latency is higher than some other ram with low ms delay and low CL.

 

So if you actually care about the marginal performance difference between different ram speeds you need to look at real world benchmarks, not divide the two numbers together and think that lower absolute latency is what matters.

[straight_stewie]

1 hour ago, Enderman said:

People think it is this easy but some programs will perform better with high frequency ram even if the cas latency is high and the absolute latency is higher than some other ram with low ms delay and low CL.

 

I'm going to have to disagree. Some hardware may prefer one over the other, but all that software cares about is a quick response time. How that response time is achieved is of no consequence (of course, this only applies to normal desktop tasks.)

But you are correct in that it's not as simple as taking the f/xxxx-t ratings and calculating things: There are many other things that can affect the performance of RAM in a static case, not to mention all the factors that can affect it when the system is running. Looking at real world benchmarks, for the software you plan on running, is necessary if you're that hard up for marginal performance gains.

I've always taken a somewhat simpler approach to RAM: Buy the most aesthetically pleasing (build matching) kit that is both in the capacity and price range that I'm looking for.