Question Huge difference in Multicore MC write score compared to other average benchmarks (same memory)

Stocky1984

Distinguished
Dec 4, 2010
32
1
18,535
Recently I built a new system with a 5800x. At first I got the memory only up to 38 percentile (userbenchmark) with standard docp enabled. After some research with dram calculator I got it up to 54th percentile however I think there's still room for improvement. Especially when I compare the memory benchmark with others that have same memory. For some reason my multicore MC write score doesn't get higher than 27 while others with the same memory get it up to 46-47. Is there any setting in the bios that I still have to adjust to get those numbers as well? Below you see the (timing) values I put into the bios using dram calculator:

sqe.png


I copied all those settings into the bios and it runs stable. Went from 38th percentile (standard docp) to 54th percentile (manual) on memory userbenchmark.
There must be some room for improvement when I compare my multicore MC write score to other average benchmark with the same memory:

qdqsdqsd.png


Motherboard: X570 Asus E-gaming
Memory: G.Skill F4-3600C16D-32GTZN

Must be some bios setting somewhere that I'm overlooking. Maybe someone here knows? Thanks.
 
have you tried to test your ram speed? i use aida64 extreme to test my ram speed + latency.

besides on tweaking primary timing on ram, you shoul explore more about secondary and teritary timings, mostly on Read and Writes (R/RD and W/WR as indicator)

here's a link to make you understand about these timings
 
Last edited:

Stocky1984

Distinguished
Dec 4, 2010
32
1
18,535
Just talked to a friend, the explanation might be that these other benchmarks (with higher multicore write score) are all done with cpu's which have 2 ccx's. The 5800x has only 1 ccx so this might be the reason why there's such big difference in scores. Maybe someone can confirm this.
 
Just talked to a friend, the explanation might be that these other benchmarks (with higher multicore write score) are all done with cpu's which have 2 ccx's. The 5800x has only 1 ccx so this might be the reason why there's such big difference in scores. Maybe someone can confirm this.
The higher percentage on ram is likely from people running the test multiple times while tightening timings. It doesn't state the cas latency of those other 3600MT/s tests, just that they were 3600. People run userbenchmark a lot when overclocking and tweaking their systems and it's been speculated that it's why the results don't match most peoples results when compared to the average.

Honestly though, unless you are doing memory sensitive tasks, don't worry about your ram unless your programs are crashing alot, which would indicate your ram is not stable at it's current settings. You are unlikely to notice any difference tweaking it beyond what DRAM calc is showing.