Is there any advantage for IL-2 in RAM 2x8Gb vs. 2x16Gb or 4x8Gb?

[chiliwili69]

It is well known that IL-2 in monitor or VR doesn´t need more than 16Gb of RAM. Having more RAM should not give better performance.

 

The IL-2 RAM requirement has been analyzed here, and it is around 8-9 Gb:

 

But some people (like me) went to some 2x16Gb configuration believing that the Dual rank thing was going to provide better performance (higher bandwidth).

Other people went to 32Gb or more just because they play other games or sim or profesional applications which require more RAM.

 

But, if your only interest is IL-2 in monitor or VR (my case), then What is the advantage of a 2x16Gb (or 4x8Gb) over 2x8Gb?

 

I have been distilling some tests from the SYN_VANDER test results table and tried to compare CPUtests with no AMD GPU cards, Windows 10, same CPU model, same IL-2 version, similar RAM speeds, etc. So, basically I ended comparing only the 5600X and 5800X tests.

 

For the 5600X the comparison is this (grouped by Il-2 version):

 

And for the 5800X the comparison is this:

 

Looking every case in detail, I would say that there is not performance difference in the 2x8 or 2x16 (or 4x8) for a given frequency.

 

The test performed by DBCOOPER011, just removing 2 sticks, allowed him to go to higher RAM speeds.

 

Therefore, for a given budget, it would be better to go for 2x8Gb of higher frequency (and lower Latency) rather than spending the money in 32Gb.

 

Or also, spend the money in a better Mobo (Asus, MSI, EVGA) of X570 line (rather than B550) which will support faster RAM models (default XMP profiles).

 

Note: @SCG_Wulfe You were experimenting some microsttutering when playing in VR. And you have indentical system than me except 2x8Gb. I am not having microstuttering at all. Would that perhaps explain your issue?

[SCG_Fenris_Wolf]

That's on the right track. The biggest drops were found in the low 1% or 0.1% and regarding thread management, and a rise in latency in other games as well, when using 4x8GB on Daisy chain topology.

 

Whether this has been fixed by now due to the many BIOS and chipset updates released for the Ryzens since 2020, is the right question to ask.

 

It was absolutely true in December 2020 when I benchmarked in IL-2, and others in MSFS. There are even YouTube videos still existing on it, wondering about the difference in results and later updating via commentry.

 

 

Whether this is still true in December 2021 may be the more precise question, putting date and updates into the equation.

 

P.S. This is referring to VR.

Edited November 27, 2021 by SCG_Fenris_Wolf