Memory Speed testing Results - Ryzen 5800 with fast 32Gb Ram comparison assistance required

[RAAF492SQNOz_Steve]

Hi fellow Ryzen 9 CPU users (with fast 32 Gb Ram fitted)

 

I have recently upgraded from 2x8 to 2x16 Gb fast RAM and am getting mixed results when tuning and running benchmark software. Passmark gives it a 98% score and UserBenchMark scores it at 44% ???????.  Interested to see what results others are getting using a simple free benchmark tool, Maxxmem2, if your 32Gb RAM is running at 3200 - 3600+ MHz or more (especially if you have managed to achieve CL14 timings). 

 

Would have preferred to use Aida64, for benchmarking, but my 30 day evaluation has run out  .  Results between the two benchmarking tools are not easily comparable so would request only Maxxmem2 results.

 

In my testing, I found that it is best to do three or so runs, with Maxxmem2, as you do get some variation in results.  Maxxmem2 give a result pretty quickly so testing should not require much time.

 

I have been able to do some tuning of the 2x16Gb G.Skill f4-4000C16-16GTZNA RAM from a default setting and gained about a 4 Gb improvement in the score but find them very inflexible when trying to reduce primary timings. When trying to reduce CL times I was running them at 3600MHz.

 

They clearly rely on operating frequency rather than CL timings to get their performance figures. Playing around with VDIMM and VSOC voltages yielded no benefit what so ever in trying to get CL times down which really surprised me. When tuning other B-Die RAM, increasing VDIMM voltage had always helped me with tightening primary timings. Happy to concede that I am still quite the amateur when it comes to tuning memory.

 

Here is the median result of tests once I had adjusted secondary memory timings. Trying to adjust primary timings was, generally, fatal with a BSOD or similar.

 

 

It is interesting to note that my El Cheapo 2x8Gb Samsung B-die RAM can get pretty close to the above figures ( 48.5 Gb) so the upgrade to 32Gb will not do much at all for IL2 performance. Should, however, help with MS Flight Sim when I get time to fly it.

 

This is my highest result below so you do get some variations..........

Edited July 7, 2021 by RAAF492SQNOz_Steve
Added variation example

[RAAF492SQNOz_Steve]

Looks like I have to answer my own question about the poor rating of my  2x16Gb G.Skill f4-4000C16-16GTZNA RAM when running the UserBenchmark tests.

 

This test score does not account for the test being run on either a 8 Core Ryzen CPU (i.e. 5800x) or 12-16 Core CPU ( i.e. 5900x or 5950).  The multi core result clearly benefits in either the larger 72Mb buffer or extra cores, or both with the 12 - 16 core test results.  All is not lost for 5600x or 5800x owners as they can apparently get their revenge in the single core test results with a bit of tuning.  

 

Top 3 results in the world for G.Skill f4-4000C16-16GTZNA RAM, all are from 5900x's (Note, when you say the top 3 in the world it gives the figures great "street cred" but I would observe that only 28 user tests have been conducted with this RAM, so a very small sample)

 

My 5800x result, Note: RAM is clocked at 4000Mhz which is apparently very difficult to achieve with the 5900x and 5950x

 

 

Not sure if IL2 relies on single core access for memory but speculate that it does..................................

 

 

 

 

Edited July 11, 2021 by RAAF492SQNOz_Steve
typo