Is clock speed more important than latency?

jedinegotiator

Distinguished
Apr 2, 2011
141
0
18,690
Currently I have G.SKILL Ripjaws X Series 8GB (2 x 4GB) 2133Mhz
they have 9-11-10-28 timings.

Newegg currently has G.SKILL Sniper Gaming Series 16GB (2 x 8GB) 1600Mhz RAM on sale for $89.99
They have 9-9-9-24 timings.

Which RAM do you think would be better? The GSkill has they higher clock speed but does it really matter much after 1600Mhz?

I am wondering if they lower timings on the sniper series would benefit me at all. I would also have two free RAM slots with a 2x8GB kit which would be nice. What do you guys think? Its a good deal but would there be any difference?

Thanks!
 
Solution
One formula you'll see thrown around is this one:

CAS / Speed x 1000 = ns

That is how many nanoseconds it takes RAM to do something

9/2133 x 1000 = 4.2 ns

You are not going to see a significant difference between 9-11-10-28 and 9-9-9-24

As for what faster RAM does....well not much if whatever you are running is bound by another component's performance. In gaming, most of the time it's your GFX card..... but ya ever get to the point in a game where ya screen gets jumpy and your toon is stutter stepping ? That **can** be a RAM limitation.

While tests really only show a very small increase in performance in average fps.... when reviewers bother to look, we do see greater increases in minimum frame rates. So let's say ya see a 5%...

inherendo

Honorable
Nov 1, 2012
48
0
10,540
I've read a some on my first build regarding ram. There's not much tangible improvement besides benchmarks when you overclock ram or tighten the timings, etc. Mostly, people do it just for kicks. Don't take my word for it though.

It does help with APUs though.
 

jedinegotiator

Distinguished
Apr 2, 2011
141
0
18,690


Which RAM would you pick?
 

PyjamasCat

Honorable
Mar 20, 2013
874
0
11,360
Timings will affect only specific latencies that you change. Clock speed will more or less adjust everything (or need everything to be adjusted, which the mobo can do automatically). The clock speed and latencies won't really affect your gaming performance. The difference will only be seen in benchmarks. And inherendo is right. Most people play with these settings for fun (me) or bragging rights. Occasionally people do this for thier work PC's to get the extra performance from MatLab or something like that.

The RAM you choose depends on what you are using it for.
Is this a purely gaming PC? Then the upgrade won't be worth it.
 

inherendo

Honorable
Nov 1, 2012
48
0
10,540
Do you actually use up all 8 gb or plan to? If not, 90 bucks for something that will have no effect on your system is pointless. You could down clock your sticks and make a ram disk for fun I guess. Ram prices jumped for a while but are climbing down again, though I don't think 90 bucks for 2x8gb is that great. Again haven't needed ram so haven't followed trends carefully.
 
One formula you'll see thrown around is this one:

CAS / Speed x 1000 = ns

That is how many nanoseconds it takes RAM to do something

9/2133 x 1000 = 4.2 ns

You are not going to see a significant difference between 9-11-10-28 and 9-9-9-24

As for what faster RAM does....well not much if whatever you are running is bound by another component's performance. In gaming, most of the time it's your GFX card..... but ya ever get to the point in a game where ya screen gets jumpy and your toon is stutter stepping ? That **can** be a RAM limitation.

While tests really only show a very small increase in performance in average fps.... when reviewers bother to look, we do see greater increases in minimum frame rates. So let's say ya see a 5% increase in minimum frame rates, is that worth an increase in RAM costs from say $133 to $159 ..... well the obvious answer is no..... that's a 20% increase in RAM cost for a 5% performance gain .... or is it ? Wasn't that a $1600 system ya playing on....or $1626 with the faster RAM ? If so that's a 1.6 % increase which does make sense for a 5% performance improvement. I won't go out and buy DDR3-3000 any time soon but 1866 memory has been roughly the same price as 1600 for a while now and 2133 is starting to drop fast.
 
Solution

Tradesman1

Legenda in Aeternum

-------------------------------

+1

then you take in the freq yet again, 1600 theoretically can move 12,600 MT/s (roughly MB per second) vs say 2133 which can move (again theoretically) 17,000 MT/s, so the faster it can complete an operations the more it can move when using higher freq DRAM
 

MJM87

Distinguished
Aug 28, 2008
10
0
18,510
Still, I think no one actually answered to the basic question.
Which is faster, speed of ram in mhz or latencies (Mostly people talk about Cache latency).
Please don't go for offtopic, please do not talk about CPU or GPU etc.. We supposed to talk about rams, that is why google points me here

In my comparison I am wondering which is better, I supposed to choose between 8GB rams of: CL7 1333MHz or CL11 1600Mhz.
What I'm doing on my computer? -Everything, playing games, running skype and web browser behind, making videos with video editors, building maps for games, packing and extracking rar files, akykind of processing data..

I'd like to compare differences in Mhz and latency in percentages. Which is more importand, Mhz or cache latency?
from CL11 to CL7 the drop is 36,36%, while from 1300Mhz to 1600Mhz the increase is 23,08%.

So may I ask again, which one is more importand, Mhz or Cache Latency, lets think in percentages. Or are they egual importand to each other?
 

Tradesman1

Legenda in Aeternum
When comparing sticks, it's not an either or thing, it's a combination of the two. Generally you would look at 211 and 1600 and most just say the 2133 is the better, but if it's 2133 with a very slow CL of 11 or 12 the it will be out performed by a set of 1600/7 or 8 (i.e. with good sticks 1600/7 can OC to 2133/9)
 
I'd like to compare differences in Mhz and latency in percentages. Which is more importand, Mhz or cache latency?
from CL11 to CL7 the drop is 36,36%, while from 1300Mhz to 1600Mhz the increase is 23,08%

To borrow from "My Cousin Vinny" .... there is no answer, it's trick question

The correct answer was given to the correct question. It's not either or that matters; it's the product of the two that matters.

CAS / Speed x 1000 = ns

If you are asking what's faster 1333 CAS 7 versus 1600 CAS 9 versus 2400 CAS 10

7 x 1000 / 1333 = 5.251 nanoseconds

9 x 1000 / 1600 = 5.625 nanoseconds

10 x 1000 / 2400 = 4.167 nanoseocnds.


Again, how much the extra speed gets ya is almost always misrepresented. It's basically game / application dependent. It's somewhat true in most games in that the % increase is small ..... however since the difference in system price is even smaller, the extra dough is easy to justify. Also where it affects performance, in games it';s generally a greater impact in minimum rather than average fps that we see the most impact.

Finally, some applications / games benefit greatly. How's 10% improvement strike ya ? Going from 159 for 1600 / 177 for 2400 is an 11+% improvement.

image006.png
 
The first set has lower latency (4.375) ... that's the "rest" it takes between executing in instruction ... the 1866 executes it faster. How that whole jumble ends up is more complicated than simple multiplication which is why I rely on actual test results like in the charts above.

http://www.tomshardware.com/reviews/memory-scaling-gaming-haswell-richland,3593-18.html

We found that DDR3-2133 performed best on both AMD and Intel platforms, and Mushkin’s 997121R won the performance race by being latency-optimized at that data rate. It gets our most prestigious and rare Elite award for being the best product in its class