Does RAM nanoseconds delay make a difference?

[steffeeh]

So I've done some investigation on RAM speeds and RAM latency, and I've realized they in cooperation affect the real world latency in nanoseconds.
I've also done some investigation on Newegg to find the various best speeds in addition to latency cycles and compared the real world latency in nanoseconds.

For instance, my Corsair Vengeance memory with 2133 MHz and CL13 gives a real world latency of 6 nanoseconds.
Comparing with a Corsair Vengeance LED memory with 3466 MHz and CL16, it would give a real world latency 4.6 nanoseconds.
So there is a difference by 1.4 nanoseconds. [OBS! Considering I've done all the math correctly]

But how much performance difference would I notice in different tasks with a computer if I would in theory upgrade to the latter memory?
Tasks such as CPU heavy games, RAM heavy games, medium heavy games for CPU/RAM, video editing software, video rendering, audio rendering, audio software/DAW, photo editing, etc.

Does it really matter how much latency the RAM has? By googling you find lots of discussions and tests on wether it has an effect, but it feels like the tests are poorly done, only focusing on either speed or CAS latency, and not comparing different real world latencies in nanoseconds. 
So there's not much to go on there.

[Enderman]

If you look at benchmarks of ram speed, there are several games that benefit from higher frequencies.

[The_Stig]

RAM speed makes almost no difference, I think linus has a video on this.

[Tycho]

no. just get the cheapest ram you think looks good. (from a reputable brand ofcourse)

[typographie]

26 minutes ago, steffeeh said:

Does it really matter how much latency the RAM has? By googling you find lots of discussions and tests on wether it has an effect, but it feels like the tests are poorly done, only focusing on either speed or CAS latency, and not comparing different real world latencies in nanoseconds.

It seems like you're trying to boil down frequency and CAS latency to one oversimplified figure, when in reality different applications benefit from each differently. I'm not necessarily saying your way of looking at it is entirely incorrect (I have no idea where those nanoseconds are coming from or how accurate they are, tbh), but it seems highly theoretical. The benchmarks tend to test frequency and latency independently because that's a better representation of how real software behaves.

 

How much of a difference memory performance makes depends on what you're doing. It does seem to help a bit in some games, but enough to justify an expensive DDR4-4333 kit? Probably not. I think it's worth going up to about 3200 MHz, based on the prices and benchmarks I've seen.

[SLAYR]

Unless you need those 6E-9 seconds, but it is literally impossible for a human to notice that small of a change.

 

Ram speeds only help if you have your cpu pinned at 100% constantly.

[Gonio]

As with anything: there is always a scenario where there is benefit. However in most cases for RAM it's considered a bad investment for the performance gain. Thats the reason most don't recommend it, myself included. Not saying it doesn't help, but there options in your build you could better spend that dosh on that deleviers mroe overall performance.

 

As for your question: a difference of 1.4 nanosecond you will not notice. Ram speed hasn't be the biggest chokepoints for years.

[SnowyMus]

RAM speed does affect CPU performance (and GPU performance if you use some sort of integrated graphics). While on some games it can give a nice bump which can be useful if you're below 60 FPS, anyway, you could always spend the extra $20 on, say, a K-sku i5 instead of a non-K i5, which CPU frequency will almost always have a much greater impact on CPU performance, as you're making the processor itself faster rather than the data.

 

If you want, you can still overclock RAM with RAM that isn't rated to run past 2133 MHz, as the only differences are binning and XMP profiles. You aren't given any guarantees that anything will work, but it can mean the difference between something like an average of 55 FPS and an average of 50 FPS on lower-end hardware and give you something to overclock if you can't overclock your CPU any higher.

[typographie]

19 minutes ago, wrathoftheturkey said:

We have a consensus

Do you see us reach consensus on this forum often? Do you find this unusual?

 

It's very difficult to isolate memory performance on a gaming benchmark, and it's absolutely done incorrectly sometimes. If there's any hint of a GPU-bound framerate, the influence of memory will probably be invisible. But if the benchmark suite used game sequences that were heavily CPU-bound, memory frequency should usually make some consistent difference.

 

I recommended 3200 MHz mainly for pricing reasons. That is a frequency I've seen make a couple of frames' difference in some games, and DDR4-3200 kits are available for < $10 more than DDR4-2133 kits. I wouldn't pay a huge premium for faster memory, but for $5–10 out of the cost of a gaming PC I think it's a fairly reasonable way to go.