Solved

Cas, Latency, Timing ??

Tags:
Last response: in Memory
Share
November 4, 2013 10:57:25 AM

I have learned that CPU speed and Motherboard FSB speed are closely related and is what determines the min/max speed rating for the RAM but how do you determine what is the best Cas/Latency/Timing, is it dependent on the intended use of the RAM ?
Is a lower C/L better ?
Where does the RAM Timing come into play ?

More about : cas latency timing

November 4, 2013 11:05:26 AM

A lower CL is faster but not always the best choice for the selected configuration, still a bit clueless on Timing but i will get there with more "time"...
m
0
l
a b } Memory
November 4, 2013 11:06:33 AM

Latencies and timings are practically the same thing. Lower latencies are always better, though you actually have to do some calculations if you want to know the real latencies. The numbers they state aren't in actual time units (nanoseconds), but in clock cycles. And since RAM can run at different clock speeds, that can skew the numbers. To get eg. the CAS latency in nanoseconds, you divide the CAS latency number (eg. 9) by the DDR3 speed (eg. 1600 MT/s) and multiply by 2000: 9 / 1600 * 2000 = 11.25 nanoseconds.
m
0
l
Related resources
a b } Memory
November 4, 2013 11:06:41 AM

The lower the CAS Latency or CL for short, the better the RAM.
The higher the frequency the better the RAM.

BUT having RAM with massively high frequency and high latency is not beneficial, there needs to be a balance somewhere depending on the usage you have.

For example a good balance for your average system is 1600/1866 CAS 9 RAM.
But for an APU based system that uses integrated(on CPU graphics), you can sacrifice SOME latency for frequency. For example 2133MHz / 2400MHz CAS 11 RAM would work well in that situation.
m
0
l
a b } Memory
November 4, 2013 11:11:12 AM

Novuake said:
The lower the CAS Latency or CL for short, the better the RAM.
The higher the frequency the better the RAM.

BUT having RAM with massively high frequency and high latency is not beneficial, there needs to be a balance somewhere depending on the usage you have.

For example a good balance for your average system is 1600/1866 CAS 9 RAM.
But for an APU based system that uses integrated(on CPU graphics), you can sacrifice SOME latency for frequency. For example 2133MHz / 2400MHz CAS 11 RAM would work well in that situation.

DDR3-2133 CL11 has lower latency than DDR3-1600 CL9. DDR3-2400 CL11 has lower latency than DDR3-1866 CL9.

Realistically, actual latencies don't vary much between RAM running at different clock frequencies.
m
0
l
November 4, 2013 11:13:20 AM

The lower the better.

Timings, the lower the better

The cas, the lower the better
m
0
l
a b } Memory
November 4, 2013 11:13:57 AM

Sakkura said:
Novuake said:
The lower the CAS Latency or CL for short, the better the RAM.
The higher the frequency the better the RAM.

BUT having RAM with massively high frequency and high latency is not beneficial, there needs to be a balance somewhere depending on the usage you have.

For example a good balance for your average system is 1600/1866 CAS 9 RAM.
But for an APU based system that uses integrated(on CPU graphics), you can sacrifice SOME latency for frequency. For example 2133MHz / 2400MHz CAS 11 RAM would work well in that situation.

DDR3-2133 CL11 has lower latency than DDR3-1600 CL9. DDR3-2400 CL11 has lower latency than DDR3-1866 CL9.

Realistically, actual latencies don't vary much between RAM running at different clock frequencies.

I don't see how you could say that?
Increasing clock rate does not improve the actual time it takes for the memory controller to read a specific column?

m
0
l
November 4, 2013 11:14:58 AM

Novuake said:
The lower the CAS Latency or CL for short, the better the RAM.
The higher the frequency the better the RAM.

BUT having RAM with massively high frequency and high latency is not beneficial, there needs to be a balance somewhere depending on the usage you have.

For example a good balance for your average system is 1600/1866 CAS 9 RAM.
But for an APU based system that uses integrated(on CPU graphics), you can sacrifice SOME latency for frequency. For example 2133MHz / 2400MHz CAS 11 RAM would work well in that situation.

good stuff, you mention that an APU can sacrifice some CL in favor of more speed but do you actually (need) to make a compromise, can you choose both a low CL and a high speed such as CL 9 and DDR2133 with an APU, thank you in advance for sharing your wonderful knowledge..

m
0
l
November 4, 2013 11:15:45 AM

Well that used to be normal. But there is no more FSB per se. in modern CPUs

But in general, the lower the timings the better. For example 2400 with long latency could be worse than 1866 with low latency.
m
0
l
a b } Memory
November 4, 2013 11:18:42 AM

gman97005 said:

good stuff, you mention that an APU can sacrifice some CL in favor of more speed but do you actually (need) to make a compromise, can you choose both a low CL and a high speed such as CL 9 and DDR2133 with an APU, thank you in advance for sharing your wonderful knowledge..

Very rarely you could overclock RAM to reach those specs. I know that some G.skill chips are known to do this(Gskill snipers and Ripjaws).

But getting them there is not that simple.
m
0
l
a b } Memory
November 4, 2013 11:19:11 AM

Novuake said:
I don't see how you could say that?
Increasing clock rate does not improve the actual time it takes for the memory controller to read a specific column?

The latencies are counted in clock cycles. Clock cycles are shorter at higher clocks, so when you convert it into the actual time, things look different.

You can calculate the time the way I described above, or check the Wikipedia page for a few examples.
m
0
l
a b } Memory
November 4, 2013 11:20:17 AM

@Sakkura

Just to make it clear what I said above, while increase frequency DOES mitigate the higher CAS to a degree, its not a 1 to 1 thing. I really do not want to go read up on the math again, so if you know better, feel free to post it.
Would not mind a refresher on the math at all.

EDIT : POsted a bit late. But yes, what you say is true, but higher CAS does translate to a sacrifice. Like I said, not sure how much. Hate doing math unless forced to.

EDIT 2 : Not to mention that chart sucks.
According to it a 400MHz bump on CAS 10 vs CAS 9 pretty much improves the time, which does not seem right.

Damn, seems I will have to do the math when I get home. Still 400MHz is not quite ONE bump from 1600MHz on modern systems. So still sacrifice needed for higher frequency.

Agree Sakkura?
m
0
l
a b } Memory
November 4, 2013 11:28:52 AM

Novuake said:
@Sakkura

Just to make it clear what I said above, while increase frequency DOES mitigate the higher CAS to a degree, its not a 1 to 1 thing. I really do not want to go read up on the math again, so if you know better, feel free to post it.
Would not mind a refresher on the math at all.

EDIT : POsted a bit late. But yes, what you say is true, but higher CAS does translate to a sacrifice. Like I said, not sure how much. Hate doing math unless forced to.

Not to mention that chart sucks.

What matters is how many nanoseconds it takes for the memory to respond to a command. Whether 10 nanoseconds is made up by 8 cycles at 1600 MT/s or 12 cycles at 2400 MT/s does not matter, the CPU will be waiting for 10 nanoseconds either way.
m
0
l
a b } Memory
November 4, 2013 11:30:24 AM

Sakkura said:
Novuake said:
@Sakkura

Just to make it clear what I said above, while increase frequency DOES mitigate the higher CAS to a degree, its not a 1 to 1 thing. I really do not want to go read up on the math again, so if you know better, feel free to post it.
Would not mind a refresher on the math at all.

EDIT : POsted a bit late. But yes, what you say is true, but higher CAS does translate to a sacrifice. Like I said, not sure how much. Hate doing math unless forced to.

Not to mention that chart sucks.

What matters is how many nanoseconds it takes for the memory to respond to a command. Whether 10 nanoseconds is made up by 8 cycles at 1600 MT/s or 12 cycles at 2400 MT/s does not matter, the CPU will be waiting for 10 nanoseconds either way.

Yes, but for example 1866MHz CAS9 will be less than 2133MHz CAS 11, correct?

Form the top of my head.
m
0
l
a b } Memory
November 4, 2013 11:34:26 AM

Yes, 1866 CL9 is slightly lower latency than 2133 CL11.
m
0
l

Best solution

a b } Memory
November 4, 2013 11:37:24 AM

Sakkura said:
Yes, 1866 CL9 is slightly lower latency than 2133 CL11.

Cool. Stimulating conversation.   Thanks.

Back to OP..

You get what you need from this whole conversation gman97005?
Share
November 4, 2013 12:49:23 PM

Novuake said:
Sakkura said:
Yes, 1866 CL9 is slightly lower latency than 2133 CL11.

Cool. Stimulating conversation.   Thanks.

Back to OP..

You get what you need from this whole conversation gman97005?

Absolutely, learned to get G.skill DDR2133 Ripjaws X and the lowest CL i can afford
m
0
l
a b } Memory
November 4, 2013 1:50:55 PM

Good to hear.

Keep in mind that Ripjaws are high profile and MAY interfere with some aftermarket coolers.
m
0
l
November 4, 2013 4:24:39 PM

Novuake said:
Good to hear.

Keep in mind that Ripjaws are high profile and MAY interfere with some aftermarket coolers.

Won't be a problem with my Corsair Hydro H55 liquid-cooler, plenty of room in there. Gave up on the traditional heat-sink and fan over 6 years ago, with liquid cooling temps are never a question and it makes for a very tidy system..
m
0
l
a b } Memory
November 5, 2013 12:32:51 AM

Very well then.
m
0
l
!