Memory Latency

[Neutralus1492]

I've noticed that the latency tables for the DDR4 memory was just updated. The question I have is what caused the significant change? One of the Corsair modules I've been tracking was originally posted with a 16ns latency but now has been downgraded to 21ns. Also, why is there such a significant latency difference between Intel run DDR4 and AMD. It's at least a 35ns difference in the lowest latency options?

[Richard (PassMark)]

The charts are updated every night to include new submissions by our users. If a particular memory has low number of samples, then new submissions can have a bigger impact on the values swings.

We have split up DDR3/4 RAM benchmarks into separate AMD and Intel charts. The reason for doing this is because the current high end Intel CPUs are much more likely to be able to use the full bandwidth of current DDR3/4 RAM. Also the AMD chips have high latency compared to the Intel chips. Intel has supported DDR4 for much longer than AMD, while AMD is improving, they are still behind.

If there is a specific module you are looking at that you believe there is an issue, let us know which and we can take a look.

[Neutralus1492]

Is there a more neutral oriented benchmark list that builders can access to better gauge which equipment is top of the line and which isn't? When I shop for new equipment the benchmarks are an important resource in determining what to buy. With your explanation above, it would appear I have been misinterpreting the benchmark results on this site. I (incorrectly) assumed that the tops the lists were the fastest, most powerful, or whatever for it's category available to purchase. For example, the hard drive category shows the Samsung 970 Evo 1Tb beating out the Samsung 970 Pro 1Tb. That result, according to your explanation above would mean that whomever whether individual or group submitted the Evo benchmark probably has higher quality or more efficient computer equipment allowing them to run the Evo at a higher bench marked rate. Of course the manufacturers website shows the specs for both and it does state that the Pro is better, hence why it costs more. So with that knowledge and lacking the system specs of the benchmark submitter, how would I or any other builder be able to determine which gear can give me the best performance? The benchmarks listed seem to be more of a bragging center for users than an actual credible information center on a devices performance capabilities.

- Dazed and Confused

[David (PassMark)]

The benchmark is already "Neutral". But is just a fact that slow CPUs don't benefit from fast RAM as much as fast CPUs do.

Of course the manufacturers website shows the specs for both and it does state that the Pro is better

Actually it doesn't. Samsung claim 500,000 IOPS and up to 3500 MB/sec read speeds for both the EVO and PRO. There is a tiny difference in write performance, but this difference will be swamped by differences in motherboards, device driver versions and background activity. Those figures only apply at ridiculous queue depths in any case (and if buffers aren't exhausted).

So Samsung rates them about the same performance and so do we.
The big difference with the PRO vs the EVO is the endurance figure. The PRO has double the write endurance.