DRAM Calculator for Ryzen & Gigabyte MB

funkydmunky

funkydmunky

2[H]4U
Joined
Aug 28, 2008
Messages
3,871
I once had some very nice RAM timings on my first gen Ryzen done by a friend. Using the included Dram calculator benchmark I scored 120'ish. Now after a BIOS update and the wiping of timings (yep should have wrote them down :cry:) with standard XPM profile with RAM listed in sig I get 170! Wow what a diff.

Now I am playing around with recreating those tight timings but the first thing I discover is very little of the excellent DRAM calc listings match what my Gigabyte BIOS lists. Some I can logically figure out what they mean, but many (most) others I have no idea.
So my question is. Is there a guide or any help available for me to learn what will help me figure out DRAM Calc to Gigabyte BIOS (X370) memory guide out there?
I almost exclusively play VR games and have noticed my all so important 1% lows have really taken a hit.
Any and all help is appreciated.
Stay [H]ard fellas ;)
 
You'll likely need to try and stability test those on your own.

I learned a lot overclocking with both 1usmus' DRAM calculator (https://www.techpowerup.com/download/ryzen-dram-calculator/) and via the following:
https://github.com/integralfx/MemTestHelper/blob/master/DDR4 OC Guide.md

I used memtest86 to stability test which was 3-4 hours per try but other people have used other, faster running programs with success.

Some quick notes:
Ryzen, assuming you have your MHz dialed in, in general likes tweaked timings and especially:
CAS Latency (and the other primary timings including the next two)
TRAS
TRC
TRFC
TFAW
TWR
The 2 SCL timings
Command Rate 1T > Geardown Mode > 2T

The rest have smaller impact on gaming speed.
 
mda said:
You'll likely need to try and stability test those on your own.

I learned a lot overclocking with both 1usmus' DRAM calculator (https://www.techpowerup.com/download/ryzen-dram-calculator/) and via the following:
https://github.com/integralfx/MemTestHelper/blob/master/DDR4 OC Guide.md

I used memtest86 to stability test which was 3-4 hours per try but other people have used other, faster running programs with success.

Some quick notes:
Ryzen, assuming you have your MHz dialed in, in general likes tweaked timings and especially:
CAS Latency (and the other primary timings including the next two)
TRAS
TRC
TRFC
TFAW
TWR
The 2 SCL timings
Command Rate 1T > Geardown Mode > 2T

The rest have smaller impact on gaming speed.
Click to expand...
Thanx for the advice. DRAM calc is a great tool, but my Gigabyte BIOS is speaking a different language.
But I will focus on the setting you suggest.
 
1800 RAM Timings.jpg


This was derived from someone else's settings on a Gigabyte board and they work for me on my Asus Crosshair VIII with Samsung B-Die memory. No memory timings from the DRAM Calculator work for me. YMMV.

3950x 3600 mem 1800 fclk 08-17-20.png


Here are the corresponding AIDA scores with other crap running in the background.
 
kamikazi said:
View attachment 270613

This was derived from someone else's settings on a Gigabyte board and they work for me on my Asus Crosshair VIII with Samsung B-Die memory. No memory timings from the DRAM Calculator work for me. YMMV.

View attachment 270621

Here are the corresponding AIDA scores with other crap running in the background.
Click to expand...

It'd be easier to sort if it was in a table but.

tRAS (Cycle Time): 32
tRC (Row Cycle Time): 48
Trfc: Is fine
Trfc2: 215
Trfc4: 132
Trrds: 4
Tfaw: 16

That's just from a quick glance over what you posted.
 
Priller said:
It'd be easier to sort if it was in a table but.

tRAS (Cycle Time): 32
tRC (Row Cycle Time): 48
Trfc: Is fine
Trfc2: 215
Trfc4: 132
Trrds: 4
Tfaw: 16

That's just from a quick glance over what you posted.
Click to expand...
I'm not the OP, I was just offering up some timings for him to try out. For the record, I think I let Trfc2 and Trfc4 auto fill themselves. I went through lots of 8 hour memtests verifying these settings were pretty solid. ;-) I would like them to be a little tighter, especially since it's an over $300 set of B-Dies, but I don't know if I have the patience.
 
kamikazi said:
I'm not the OP, I was just offering up some timings for him to try out. For the record, I think I let Trfc2 and Trfc4 auto fill themselves. I went through lots of 8 hour memtests verifying these settings were pretty solid. ;-) I would like them to be a little tighter, especially since it's an over $300 set of B-Dies, but I don't know if I have the patience.
Click to expand...

Yeah I know you're not the OP. Just using what you posted as an example. Everything I've found for TRFC2 etc gives the following.
tRFC 2 = tRFC / 1.346
tRFC 4 = tRFC 2 / 1.625

I have my tRFC at 260. I've picked up a good bit just from lowering tRFC alone on my 2700X. The 3700X picked up a little bit from mind 300s down to 290 but seemed to stop gaining anything in benchmarks after that even though it would go into the 270s without issue.

I'd be curious what those adjustments do for you if you try them. Im also on B-Die.
 
Those settings won't fly with a first/2nd gen Ryzen chip.

My 2700X won't clock my RAM past 3000mhz. I'm sure my board has something to do with this since the board also clocked RAM like crap with the 2600X, but even on my Crosshair 7, the 2700X/RAM combo wouldn't do past 3200mhz.

My 3600 though would clock RAM like crazy, and would probably do higher if I tried harder (AFRs at 3600mhz C16). Except I got lazy and called it a day a few hours in including stability tests heh.
 
mda said:
Those settings won't fly with a first/2nd gen Ryzen chip.

My 2700X won't clock my RAM past 3000mhz. I'm sure my board has something to do with this since the board also clocked RAM like crap with the 2600X, but even on my Crosshair 7, the 2700X/RAM combo wouldn't do past 3200mhz.

My 3600 though would clock RAM like crazy, and would probably do higher if I tried harder (AFRs at 3600mhz C16). Except I got lazy and called it a day a few hours in including stability tests heh.
Click to expand...

Your 2700X might have a bad IMC or maybe a voltage setting is off somewhere. Im able to do 3200 CL14 with tight timings on 4x8GB B-Die on my 2700X and C7H. When I only had 2x8GB it would do 3466 CL14 and tight timings with the only change being ram voltage at 1.4 volts.
 
Priller said:
Your 2700X might have a bad IMC or maybe a voltage setting is off somewhere. Im able to do 3200 CL14 with tight timings on 4x8GB B-Die on my 2700X and C7H. When I only had 2x8GB it would do 3466 CL14 and tight timings with the only change being ram voltage at 1.4 volts.
Click to expand...
The 2700X would do 3200 C14 with the CH7 but no higher.

The same RAM would clock to 3600 C16 on the CH7 and the 3700X.

The 2700X now won't do any higher than 3000 C14 on the GB X470. That board sucks but for the price I got it at, it's not bad.

I also do know that the 2700X / 3700X official speeds on 64GB/48GB are quite low... so I'm happy with what I have.
 
OP ..once you get your RAM timings figured out again .. be sure to save your profile in your BIOS (that's usually an option) .. and then just reload your profile after the next BIOS update (if there is one)..

I've experienced massive memory speed gains going from 2000 series to 3000 series .. on my wife's rig (B350) and a 2200g. I could not get 4 sticks to run stably at any speed. 2 sticks would run 2933mhz fine. Upgrade to a 3400g, all 4 sticks can run XMP 3200mhz just fine.

On my daughters rig (B350).. 2200g could only do 2933mhz om a 3000mhz kit .. upgraded to 2700x and same ram could only do 2866mhz .. upgraded her to a 3600x .. same memory does 3200mhz

I'm currently running a 3700x on an x570 mobo with 2 x 16GB Hynix HyperX Predators 3200mhz 16-18-18-36 stuff @ 3800 18-22-22-38 1.41v fclk 1900 (1:1)

the DRAM calculator never worked for me .. I just went off of XMP and then bump up speed see if stable . .if not , add a couple to xmp's numbers and try again
 
ThreeDee said:
OP ..once you get your RAM timings figured out again .. be sure to save your profile in your BIOS (that's usually an option) .. and then just reload your profile after the next BIOS update (if there is one)..

I've experienced massive memory speed gains going from 2000 series to 3000 series .. on my wife's rig (B350) and a 2200g. I could not get 4 sticks to run stably at any speed. 2 sticks would run 2933mhz fine. Upgrade to a 3400g, all 4 sticks can run XMP 3200mhz just fine.

On my daughters rig (B350).. 2200g could only do 2933mhz om a 3000mhz kit .. upgraded to 2700x and same ram could only do 2866mhz .. upgraded her to a 3600x .. same memory does 3200mhz

I'm currently running a 3700x on an x570 mobo with 2 x 16GB Hynix HyperX Predators 3200mhz 16-18-18-36 stuff @ 3800 18-22-22-38 1.41v fclk 1900 (1:1)

the DRAM calculator never worked for me .. I just went off of XMP and then bump up speed see if stable . .if not , add a couple to xmp's numbers and try again
Click to expand...
ThreeDee How do you decide that your system is stable or not?
 
x509 said:
ThreeDee How do you decide that your system is stable or not?
Click to expand...

...of course, if my rig doesn't successfully boot into windows, then I know right away it's NOT stable :)
But first and foremost .. I play Hunt:Showdown .. it's a good indicator of stability .. I crap out pretty quick if I'm not stable
Plus running vidcoder re-encoding video files
Plus running Aidi64 stress tests
Plus running THIS with custom 1usmus preset for 25 runs without any errors
Plus running various other random stress tests here and there in conjunction with just using my setup .. to see if any wonkiness comes about ..
 
  • Like
Reactions: x509
like this
Priller said:
Yeah I know you're not the OP. Just using what you posted as an example. Everything I've found for TRFC2 etc gives the following.
tRFC 2 = tRFC / 1.346
tRFC 4 = tRFC 2 / 1.625

I have my tRFC at 260. I've picked up a good bit just from lowering tRFC alone on my 2700X. The 3700X picked up a little bit from mind 300s down to 290 but seemed to stop gaining anything in benchmarks after that even though it would go into the 270s without issue.

I'd be curious what those adjustments do for you if you try them. Im also on B-Die.
Click to expand...

I may give them a try when I have some spare time. Thanks.

x509 said:
ThreeDee How do you decide that your system is stable or not?
Click to expand...

For memory stability, I boot to Memtest86 and let it run overnight. For 32GB, it takes about 5 hours to complete. Lots of attempts failed after several hours while dialing in RAM and SOC voltage.
 
Priller said:
It'd be easier to sort if it was in a table but.

tRAS (Cycle Time): 32
tRC (Row Cycle Time): 48
Trfc: Is fine
Trfc2: 215
Trfc4: 132
Trrds: 4
Tfaw: 16

That's just from a quick glance over what you posted.
Click to expand...
Priller said:
Yeah I know you're not the OP. Just using what you posted as an example. Everything I've found for TRFC2 etc gives the following.
tRFC 2 = tRFC / 1.346
tRFC 4 = tRFC 2 / 1.625

I have my tRFC at 260. I've picked up a good bit just from lowering tRFC alone on my 2700X. The 3700X picked up a little bit from mind 300s down to 290 but seemed to stop gaining anything in benchmarks after that even though it would go into the 270s without issue.

I'd be curious what those adjustments do for you if you try them. Im also on B-Die.
Click to expand...

First, I tried the rRFC 2 and tRFC 4 settings you mentioned and scored a little worse in AIDA64:

Pre-change:

3950x 3600 mem 1800 fclk 08-17-20.png

Post-change:
3950x 3600 mem 1800 fclk tweak 08-20-20.png

So, I reverted those settings back and tried the main timings changes you suggested (tRAS, tRC, tRRDS, tFAW). Even a little worse. I think I remember going through this before and scratching my head:


3950x 3600 mem 1800 fclk tweak main timings 08-20-20.png

Of course, we're not talking huge differences in any of this, fractions of a nanosecond, but you would think it may make more difference. Also, in BIOS, the auto settings on tRFC 2 and tRFC 4 show differently than what Ryzen Master shows after boot. However, Ryzen Master shows the same settings as BIOS if I manually set those two. Weird.

For the record, here's what this set of memory can do at 3733 with high memory voltage and a 3900x.

3733 mem 1.45v 1866 fclk soc 1.1v cpu -0.1 offset DRAM Calc Fast 082119.png

Both 3900x chips and the one 3950 I have now are not capable of fclock of 1900.
 
kamikazi

Try all of them at once and see what that does. I should've looked closer and seen you're on 3K Ryzen instead of 2K.

Im used to changing things in bios so try doing it as follow:
tRCD+tRP=tRAS
tRP+tRAS=tRC
tRRDS x4= tFAW (Bring tFaw down to 24 or a combo of 4/16, 5/20, 6/24)

If it is B-Die try to make both 4. You should be able to go down to 2 on both.
tRDRD SCL 4
tWRWR SCL 4
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Priller said:
kamikazi

Try all of them at once and see what that does. I should've looked closer and seen you're on 3K Ryzen instead of 2K.

Im used to changing things in bios so try doing it as follow:
tRCD+tRP=tRAS
tRP+tRAS=tRC
tRRDS x4= tFAW (Bring tFaw down to 24 or a combo of 4/16, 5/20, 6/24)

If it is B-Die try to make both 4. You should be able to go down to 2 on both.
tRDRD SCL 4
tWRWR SCL 4
Click to expand...

Here you go:

Ryzen Master all tweaks 08-21-20.jpg



3950x 3600 mem 1800 fclk all tweaks 08-21-20.png


Amazingly, still a little slower than my original looser RAM timings. Only thing off from your recommendations is TWRWR SCL is 5 instead of 4.
 
kamikazi Very interesting. The only last thing i would try is your original timings with both SCLs down to 2. It shows your tRFC is still at 290 I'd try to experiment to bring it down to 270s to see if that changes anything.

Im going to experiment with my 3700X more now to see if there is some combo it prefers. On the 2700X i saw big gains with only changing a handfull of settings and lowering tRFC was a nice step up.

EDIT: After playing around with AIDA64 on my 2700X, the benchmark does not seem completely consistent. It could be Windows running different things in the background but I've seen decent swings in latency.

With my original timings I saw some variance between 2 runs. So I manually controlled every memory timing on my board and nothing is left on auto. First run was 67.1 ns, second run was 65.7 ns.
Bringing both SCLs down to 2 did result in increasing L3 Cache Read by 6 GB/s tho.
 
Last edited:
Great thead folks. That for the replies to my OP and all the branches it took. Great stuff. I hope to find time to reply with my tests soon.
 
You must log in or register to reply here.
Back
Top