Please help me to understand Passmark ratings for memory

[octopus]

I have observed two different sets of memory whose benchmark's don't seem to corroborate well. On one system I have DDR3 that benchmarks higher than the DDR4 I have in another machine. I would've expected the DDR4 to benchmark at around double the rating of the DDR3 and yet it benchmarks lower. Do these benchmarks translate in a linear manner to MB/sec?

Here's what I've observed:

1) Micron DDR3 (PC3-12800) 1.5V 11-11-11-28 running at 1600MHz -- Passmark rating: 2493
2) G.Skill DDR4 (PC4-25600) 1.35V 16-18-18-38 running at 2933MHz -- Passmark rating: 2058

I don't know what the MB/sec are for either of these, but it would appear from the Passmark rating that the DDR3 memory is roughly 21% faster than the DDR4 memory. What is going on here?

For most other hardware--CPU, graphics, HDDs--the Passmark scores are relatively close to what I'd expect, but these memory marks (and I have done these several times over the past few years) are beyond my interpretation. Can the benchmarks be converted to average throughput (in MB/sec) by some sort of formula?

[David (PassMark)]

The PassMark rating is for the whole system (CPU, GPU, Disk, RAM). So unless all other components are identical you can't use the overall PassMark rating to compare memory speeds.

Instead look at the "Memory Mark" value. Or even better the sub-tests, "Memory write", "Memory Read Uncached", etc...

Even when just looking at the memory tests there are other factors to consider. e.g. dual channel or single channel, the amount of RAM installed, the CPU in use & BIOS settings.

[octopus]

Yeah, sorry, I shouldn't have said Passmark rating. These are the Memory Mark values obtained consistently over several tests from two different systems when I perform memory tests only. I'm having a really hard time understanding how DDR3 memory at half the clock rate of DDR4 memory can be ~21% faster. All other points from my original post are still relevant.

correction:

1) Micron DDR3 (PC3-12800) 1.5V 11-11-11-28 running at 1600MHz -- Memory Mark rating: 2493
2) G.Skill DDR4 (PC4-25600) 1.35V 16-18-18-38 running at 2933MHz -- Memory Mark rating: 2058

I seriously think that your algorithm for scoring memory has a bug unless somebody can justify how this can be so.

[David (PassMark)]

As mentioned, there are other factors to consider. e.g. dual channel or single channel, the amount of RAM installed, the CPU in use & BIOS settings. None of which you have detailed. Posting the individual benchmark test results and BIOS settings might also help.