[pshootr]

1440Mhz
9-9-9-27


1600Mhz
9-10-10-27


1600Mhz
9-9-9-27


I found these results kind of confusing. The copy and write are higher at 1440Mhz, but the read and latency are lower. This holds true even when compared to 1600Mhz with the same timings, although the CPU clock speed is slightly higher on the 1440Mhz. test. It seems looser timings scored better at 1600Mhz.

I have not done extensive testing yet, just this performance test, and 16M Super-Pi.

Obviously lower latency is nice, but I wonder which settings would overall be better suited for gaming considering the read, write, and copy as well?

 

[RyoConstantine]

Quote:

Originally Posted by pshootr

1440Mhz
9-9-9-27


1600Mhz
9-10-10-27


1600Mhz
9-9-9-27


I found these results kind of confusing. The copy and write are higher at 1440Mhz, but the read and latency are lower. This holds true even when compared to 1600Mhz with the same timings, although the CPU clock speed is slightly higher on the 1440Mhz. test. It seems looser timings scored better at 1600Mhz.

I have not done extensive testing yet, just this performance test, and 16M Super-Pi.

Obviously lower latency is nice, but I wonder which settings would overall be better suited for gaming considering the read, write, and copy as well?

Good question

it's often said in gaming speed of ram does effect fps rate, and it's true but that speed is a combination of CAS latency and frequency, not one ofthem alone

for example :
in the first tested ram :
1440Mhz
9-9-9-27

* all time are in nano second

first -> 1 / 1440*10^6 = 69.44 ns
tCAS 624.96
tRCD 624.96
tRP 624.96
tRAS 1874.88

1600Mhz
9-10-10-27

1 / 1600 *10^6 = 62.5 ns
62.5 * 9 = 562.5
tCAS 526.5
tRCD 625
tRP 625
tRAS 1687.5

1600Mhz
9-9-9-27

1 / 1600 *10^6 = 62.5 ns
62.5 * 9 = 562.5
tCAS 526.5
tRCD 526.5
tRP 526.5
tRAS 1687.5

* the reason behind difference in number is that the 1600 and 2400 aren't the actual testing speed

now some technical geek explanation about reading and writing procedure in DRAM
in DRAM two times are the most important tCAS and tRAS, tCAS for just accessing a row and tRAS for the whole accessing procedure (sensing, precharching, addressing, sensing, reading, closing)
the lower these two times are the faster the RAM is functioning (but unfortunately the tRAS time can't be controlled like the tCAS). as you can see reading in 2nd and 3rd is faster that's just not because of tCAS but also because of tRAS.
the writing procedure is kinda more sophisticated, that's why the writing time isn't the same of reading, at this point you might say if that's so then why writing in 1400 is faster than 1600 well that vastly depends on the internal RAM architecture, i mean it's not like it's always gonna be slower or different, but under normal circumstances the reading in DDR3 is faster and the ratio between reading and writing should stay the same as you go to higher frequencies but that is not always so (Architecture)
most us thinks copying is just reading + writing but that's incorrect, it'a a little confusing but reading and writing in the same page is always faster than in other pages so it's given that copying would be faster than reading and writing combined.

as for gamin, Reading and Writing speed is far more important than copying cause when you're gaming, constantly you're reading from memory and writing into the memory essentially you're not copying anything from memory to memory but you're reading and writing from memory and to memory and reading and writing from and into the SSD or HDD.

in gaming Reading and Writing bandwidth is more important than copying (copying bandwidth isn't really a good reference for choosing a RAM)

hope to be of some help

 

[pshootr]

Quote:

Originally Posted by RyoConstantine

Good question

it's often said in gaming speed of ram does effect fps rate, and it's true but that speed is a combination of CAS latency and frequency, not one ofthem alone

for example :
in the first tested ram :
1440Mhz
9-9-9-27

* all time are in nano second

first -> 1 / 1440*10^6 = 69.44 ns
tCAS 624.96
tRCD 624.96
tRP 624.96
tRAS 1874.88

1600Mhz
9-10-10-27

1 / 1600 *10^6 = 62.5 ns
62.5 * 9 = 562.5
tCAS 526.5
tRCD 625
tRP 625
tRAS 1687.5

1600Mhz
9-9-9-27

1 / 1600 *10^6 = 62.5 ns
62.5 * 9 = 562.5
tCAS 526.5
tRCD 526.5
tRP 526.5
tRAS 1687.5

* the reason behind difference in number is that the 1600 and 2400 aren't the actual testing speed

now some technical geek explanation about reading and writing procedure in DRAM
in DRAM two times are the most important tCAS and tRAS, tCAS for just accessing a row and tRAS for the whole accessing procedure (sensing, precharching, addressing, sensing, reading, closing)
the lower these two times are the faster the RAM is functioning (but unfortunately the tRAS time can't be controlled like the tCAS). as you can see reading in 2nd and 3rd is faster that's just not because of tCAS but also because of tRAS.
the writing procedure is kinda more sophisticated, that's why the writing time isn't the same of reading, at this point you might say if that's so then why writing in 1400 is faster than 1600 well that vastly depends on the internal RAM architecture, i mean it's not like it's always gonna be slower or different, but under normal circumstances the reading in DDR3 is faster and the ratio between reading and writing should stay the same as you go to higher frequencies but that is not always so (Architecture)
most us thinks copying is just reading + writing but that's incorrect, it'a a little confusing but reading and writing in the same page is always faster than in other pages so it's given that copying would be faster than reading and writing combined.

as for gamin, Reading and Writing speed is far more important than copying cause when you're gaming, constantly you're reading from memory and writing into the memory essentially you're not copying anything from memory to memory but you're reading and writing from memory and to memory and reading and writing from and into the SSD or HDD.

in gaming Reading and Writing bandwidth is more important than copying (copying bandwidth isn't really a good reference for choosing a RAM)

hope to be of some help

Interesting read, thank you for explaining. At 1440 The write is faster at the cost of read speed and latency. Seems that lower latency and better read speed would win, however that depends I guess on how important write speed is in comparison to read speed and latency.

 

[pshootr]

I may do another test tomorrow at 1333Mhz for comparison. Thanks for looking, and for all of your input. +1

 

[Cyro999]

Your memory read speed was almost 10% faster on 1600 with the slower timings. That shouldn't happen AFAIK, make sure other stuff is not using cpu/memory, do a fresh restart and then bench 5-10 times to make sure that you're always getting consistent results

 

[pshootr]

Quote:

Originally Posted by Cyro999

Your memory read speed was almost 10% faster on 1600 with the slower timings. That shouldn't happen AFAIK, make sure other stuff is not using cpu/memory, do a fresh restart and then bench 5-10 times to make sure that you're always getting consistent results

Ah, yes I was wondering about that. Your rite, I need to do more testing after closing all programs, and running multiple tests to get a clearer picture. That just does not make sense. I will spend more time testing tomorrow.

 

[RyoConstantine]

Quote:

Originally Posted by pshootr

Interesting read, thank you for explaining. At 1440 The write is faster at the cost of read speed and latency. Seems that lower latency and better read speed would win, however that depends I guess on how important write speed is in comparison to read speed and latency.

you're welcome
Glad to be of some help

Quote:

Originally Posted by pshootr

I may do another test tomorrow at 1333Mhz for comparison. Thanks for looking, and for all of your input. +1

looking forward to that, please post your result here, i wanna know them
i like these kinda testing infos, so please post as much as test result as you can, i wanna learn everything about RAMs

and thanks for +1