ryzen memory alculator

thefish · 8 Sep 2019 at 19:24

i have 16gig of 8pack samsung b die 4000mhz ram im trying to run it at 3800 with my ryzen 3600x but im getting worse performance in cinebench r20 2861.i cant seem to work out the tRFC ns value to calculate the tRFC2 and 4 values.any help would be much apreciated.when i run xmp in cinebench i get 3424

opethdisciple · 8 Sep 2019 at 22:46

Just run the Ryzen calculator.

MaXxBoulton · 9 Sep 2019 at 06:00

Just run xmp..

thefish · 9 Sep 2019 at 12:07

tRFC is showing 304 tRFC alt is showwing 448.but when i put into additional calculators the tRFC ns value of 350 that its showing in the calculator im getting tRFC as 665 tRFC2 494 and tRFC4 as 304.im confused as on the main screen of the calculator its reading tRFC 304 and on the additional calculators section its giving me tRFC as 665

tamzzy · 9 Sep 2019 at 12:13

thefish said:
i cant seem to work out the tRFC ns value to calculate the tRFC2 and 4 values

you need to get the values from thaiphon first

thefish · 9 Sep 2019 at 12:22

thats the values im getting from thaiphon.i exported the values to a html report and inported them into the dram calculator

chroniclard · 9 Sep 2019 at 12:25

have you actually tested for errors at 3800? Might not actually be stable.

thefish · 9 Sep 2019 at 12:25

im up to 3560 in cinebench r20 with if at 1900 and ram at 3800.im still way of the 3895ish scores ive seen so i guess somethings not quite right

thefish · 9 Sep 2019 at 12:37

chroniclard said:
have you actually tested for errors at 3800? Might not actually be stable.

yes with memtest 86 and with windows memory diagnostics both showing no errors

MaXxBoulton · 9 Sep 2019 at 15:56

thefish said:
yes with memtest 86 and with windows memory diagnostics both showing no errors

You said xmp was stable.. that's 4000mhz why are you trying to run at 3800?

If it will run at xmp then set xmp and play with timings if you want from there

G550 · 9 Sep 2019 at 23:55

thefish said:
i have 16gig of 8pack samsung b die 4000mhz ram im trying to run it at 3800 with my ryzen 3600x but im getting worse performance in cinebench r20 2861.i cant seem to work out the tRFC ns value to calculate the tRFC2 and 4 values.any help would be much apreciated.when i run xmp in cinebench i get 3424

I have the same kit, also with a 3600X, and run it at 3600C16 1.35V (likes 1.32V too) and with a tRFC of 345.
Ran the kit at 3.8something briefly, though de-coupled from the IF (at 3.6G). Don't recall what I used for tRFC, but probably somewhere around the 360 mark.

IIRC tRFC/2 and tRFC/4 are not used at all by the Ryzen Memory controller. You can ignore them.

If you're seeing erratic/non-linear performance scaling, it could be that you're on the verge of IF stability and data-transfers are failing/being re-sent.
Not sure if the data portion of AMD's IF is packet based with error control, though some parts of AMD's IF system definitely are.

Not something I've noticed myself, since IF @ 3.8G on my chip didn't POST at all without tinkering so never bothered going higher than 3.6G, but have read user reports of high IF causing erratic performance scaling.

thefish · 10 Sep 2019 at 00:44

G550 said:
I have the same kit, also with a 3600X, and run it at 3600C16 1.35V (likes 1.32V too) and with a tRFC of 345.
Ran the kit at 3.8something briefly, though de-coupled from the IF (at 3.6G). Don't recall what I used for tRFC, but probably somewhere around the 360 mark.

IIRC tRFC/2 and tRFC/4 are not used at all by the Ryzen Memory controller. You can ignore them.

If you're seeing erratic/non-linear performance scaling, it could be that you're on the verge of IF stability and data-transfers are failing/being re-sent.
Not sure if the data portion of AMD's IF is packet based with error control, though some parts of AMD's IF system definitely are.

Not something I've noticed myself, since IF @ 3.8G on my chip didn't POST at all without tinkering so never bothered going higher than 3.6G, but have read user reports of high IF causing erratic performance scaling.

thanks mate

Hardstyle_Shuffler · 20 Sep 2019 at 23:30

Someone above said tRFC2 and 4 aren't used by memory controller, but upon changing them from auto to values obtained from ryzen dram calc, my cinebench score went from 3676 to 3722. The only change was tRFC2 and tRFC4, so maybe they are used by memory controller? (but my bios has 3 redundant locations for ram timings, soc/vddg/vddp param, so maybe something else happened upon restart with trfc2/4 value changes.)

Used 3600 at 4.1ghz (manual oc), ram 3733 14-14-16-14-28-42 @ 1.515v, asrock ab350 pro 4 mobo with 6.0 bios

MaXxBoulton · 21 Sep 2019 at 08:30

Hardstyle_Shuffler said:
Used 3600 at 4.1ghz (manual oc), ram 3733 14-14-16-14-28-42 @ 1.515v, asrock ab350 pro 4 mobo with 6.0 bios

What ram are you running

G550 · 23 Sep 2019 at 23:16

Hardstyle_Shuffler said:
Someone above said tRFC2 and 4 aren't used by memory controller, but upon changing them from auto to values obtained from ryzen dram calc, my cinebench score went from 3676 to 3722. The only change was tRFC2 and tRFC4, so maybe they are used by memory controller? (but my bios has 3 redundant locations for ram timings, soc/vddg/vddp param, so maybe something else happened upon restart with trfc2/4 value changes.)

Posted by The Stilt in the Crosshair VI thread on oc.net: "tRFC2 or tRFC4 are not used unless the refresh mode is 2x or 4x (which normally never happens). Therefore they can be completely ignored (including the programming rule, i.e. tRFC > tRFC2 > tRFC4)."

https://www.overclock.net/forum/26135326-post18117.html

Minstadave · 23 Sep 2019 at 23:18

tamzzy said:
you need to get the values from thaiphon first

Cant you just use R-XMP to get the values now? Works ok for me.

tamzzy · 24 Sep 2019 at 00:31

Minstadave said:
Cant you just use R-XMP to get the values now? Works ok for me.

My understanding is that you can, but it depends on the ram bin quality. Rxmp gives the default values and thaiphon gives the exact timings - thats what I think happens...may or may not be accurate