Ram Performance formula

[moggeman]

I once read about a RAM performance formula:

cl/MHz * 1000

example cl10 / 2400 MHz * 1000 = 4.1

Something about measure the performance in milliseconds. Can anyone confirm this method? good/bad or just a over simplified way to dumb down all numbers down to one?

 

one reason why I wonder is that im comparing the following:

(my usecase would be a am4 mobo+cpu+rx 390 running win10, 3 EVE clients, yuotube, TS3, chrome and sometimes other titles like Doom or Battlefield 1.

CMD16GX4M4B2400C10    cl10/2400MHz = 4.1 for 2099$(Swedish krona)

CMD16GX4M2B3200C14    cl14/3200MHz = 4.3 for 2499$(Swedish krona)

 

the cl10 RAM is the clear winner, more performance for less cost, but only if this was of calculating is valid.

think the lower clockspeeds of the CL10 also helps with overall system overclocking. Or is that also a wild assumption from my imagination?

[pwn_intended]

RAM speed is largely irrelevant. I'd opt for the cheaper one.

[porina]

The calculation only gives the latency expressed as time as opposed to cycles. It is one dimension of performance, and like most cases of simplifying complex things to a single number, doesn't tell you the whole story.

 

Some tasks benefit from short latency over high speeds. I'm not overly familiar with this area, but believe some competitive benchmarks are in this camp.

Some tasks benefit from high bandwidth over latency. This applies for my interest area of Prime95-like compute tasks. Latency doesn't do much, but for a fast quad CPU with dual channel ram, performance is often significantly ram bandwidth limited, not CPU limited.

Some tasks really don't care about either (most other things).

[moggeman]

35 minutes ago, wrathoftheturkey said:

DEFINE PERFORMANCE

 

Sounds like someone pulled these numbers out of thin air, m8. Short story: RAM's effect on performance doesn't vary nearly enough to ever* justify spending more on it.

Well. As I wrote that I'm running 3 EVE online clients, talking on TS3, watching YouTube. All at the same time. Performance in this case is what type of RAM would help me run it all as smoothly as possible without stuttering.

 

Any idea about how ram speeds affects CPU overclocking? Does it matter if I go with a low MHz and CL instead of higher when it comes to clocking the CPU? 

Running a i7 4820K right now and to go as high as I can on the CPU I gotta hold the ram back a little... 

[App4that]

1 hour ago, pwn_intended said:

RAM speed is largely irrelevant. I'd opt for the cheaper one.

How is this still a misconception? This has been disproved over and over again. But, here we go. 

 

 

[pwn_intended]

17 minutes ago, App4that said:

How is this still a misconception? This has been disproved over and over again. But, here we go. 

Did you actually watch the video? The FPS numbers are pretty much the same in most of the examples they show. Faster RAM has the POTENTIAL to increase framerates, but its only really noticeable when you are quite a bit over 60fps, AND you have a GPU that is capable of pushing those high ass framerates, and is one of the few games that it actually makes a difference for. So yea, RAM speed being LARGELY irrelevant still stands.

[App4that]

6 minutes ago, pwn_intended said:

Did you actually watch the video? The FPS numbers are pretty much the same in most of the examples they show. Faster RAM has the POTENTIAL to increase framerates, but its only really noticeable when you are quite a bit over 60fps, AND you have a GPU that is capable of pushing those high ass framerates, and is one of the few games that it actually makes a difference for. So yea, RAM speed being LARGELY irrelevant still stands.

No, have you tried it yourself?

 

Overclocking your Ram has a large impact on fps dips, not every game but if the game is CPU intensive faster RAM has a bigger chance of making an impact. Also when the graphics card/s are pulling away from your CPU, the impact can be substantial.

 

Easiest game to see it is Fallout 4. Set the shadow distance to ultra, find a spot where the fps dips bad from it and save. The drop the RAM speed to stock, go back in. Now overclock your RAM. Go back in. I've seen differences of over 20fps. 

 

Don't be a parrot. 

[coalminer071]

2 hours ago, moggeman said:

I once read about a RAM performance formula:

cl/MHz * 1000

example cl10 / 2400 MHz * 1000 = 4.1

Something about measure the performance in milliseconds. Can anyone confirm this method? good/bad or just a over simplified way to dumb down all numbers down to one?

 

one reason why I wonder is that im comparing the following:

(my usecase would be a am4 mobo+cpu+rx 390 running win10, 3 EVE clients, yuotube, TS3, chrome and sometimes other titles like Doom or Battlefield 1.

CMD16GX4M4B2400C10    cl10/2400MHz = 4.1 for 2099$(Swedish krona)

CMD16GX4M2B3200C14    cl14/3200MHz = 4.3 for 2499$(Swedish krona)

 

the cl10 RAM is the clear winner, more performance for less cost, but only if this was of calculating is valid.

think the lower clockspeeds of the CL10 also helps with overall system overclocking. Or is that also a wild assumption from my imagination?

Actually you would need to divide the MHz by 2 because it is DDR (double data rate)
So It should be.
 

cl10/2400MHz = 9.167ns
cl14/3200MHz = 9.375ns

Corsair tends to use uneven timings (14-16-16 or 10-12-12) which acts closer to a latency of 15 and 11 respectively 

[moggeman]

26 minutes ago, coalminer071 said:

Actually you would need to divide the MHz by 2 because it is DDR (double data rate)
So It should be.
 

cl10/2400MHz = 9.167ns
cl14/3200MHz = 9.375ns

Corsair tends to use uneven timings (14-16-16 or 10-12-12) which acts closer to a latency of 15 and 11 respectively 

thx(for staying on topic)

 

any idea about CPU overclocking? would the CL10/2400MHz or CL14/3200MHz make any differens?

[coalminer071]

3 hours ago, moggeman said:

thx(for staying on topic)

 

any idea about CPU overclocking? would the CL10/2400MHz or CL14/3200MHz make any differens?

That would totally depend on AMD's potential to overclock the memory.
Maybe Ryzen would only be capable of running Jedec speeds.

You would really need to wait it out and see benchmarks before purchasing kits that are outside of your platform's spec.