Vengeance LPX Overclock Settings (Trash B-Die)

DAM20 · Mar 16, 2020

Hi all, thanks to my bro I was able to put my hands on some Vengeance LPX DDR4 3200 C16 2x8GB, so I oc the crap out of it.
This kit is kinda trash, it's B-Die (or C-Die?!) but works like Hynix MFR with minor tweaks.
Anyway wanted to share my settings to save you all the frustration.

A little note tho, these settings won't work for everyone coz not all chips are the same, also can vary between system and power settings.

System tested: 2600X, X470 Strix, ANTEC HCG-750-GOLD, Win10 Pro.

UPDATE: The x470 Strix-F have a fixed 100 bclock only after pressing F11 and setting gaming with tower cooler.

Now the cool stuff:

3466Mhz
Voltage-1.44----------------------------------------------------------(Errors if lower, latest bios needs 1.445)
SOC-1.05
VDDP-0.900/1.050
VPP-2.525
PLL-1.8

tCL-16 ---------------------------------------------------------------------(no post if lower)
tRCDRD-20 ----------------------------------------------------------------(post at 19 with errors, GM off)
tRCDWR-18 ----------------------------------------------------------------(no post if lower)
tRP-19
tRAS-39 -------------------------------------------------------------------(gives errors if lower)
tRC-60 --------------------------------------------------------------------(gives errors if lower)
tRRDS-4
tRRDL-6-------------------------------------------------------------------(gives errors if lower)
tFAW-16
tWTRS-4
tWTRL-8
tWR-12------------------------------------------------------------------------(gives errors if lower)
tRDRD SCL-4---------------------------------------------------------------(Blue screen on startup if lower)
tWRWR SCL-4--------------------------------------------------------------(Blue screen on startup if lower)
tRFC-555 -----------------------------------------------------------------(Lower give errors sometimes, under 500 is no post)
tRFC2-412
tRFC4-254
tCWL-16-------------------------------------------------------------------(odd no post, even give error if lower than tCL)
tRTP-8
tRDWR-8
tWRRD-1
tWRWR SC-1
tWRWR SD-7
tWRWR DD-7
tRDRD SC-1
tRDRD SD-5
tRDRD DD-5
tCKE-1
procODT-53
Command rate-1T
Gear Down Mode-Disabled
Power Down Mode-Disabled
RTT_NOM-RZQ/7
RTT_WR-OFF
RTT_PARK-RZQ/5
CAD_BUS-24 (ALL FOUR)

*Not good at formatting text

Yeah I know userbench is kinda a joke nowdays, gonna add aida when I got the time.

Dasa · Mar 16, 2020

Try
tRRDS, tRRDL, tFAW 6, 6, 24
Also see if you can lower tRFC

Veii · Mar 16, 2020

tRFC 420-312-192 will work on this set without any change
If not, 540-401-247 is the highest that will run alright

Lower both SCL to 4, and use tWRRD 5 (need to try if 4 runs)
tFAW should be 36 or 28 (unless you change tRRDS)

Interesting would be rather SiSoftware Sandra Multi-Core Efficiency test (detailed result)
Next to Aida64, to showcase effective memory perf , while SiSandra would also have to do with your ryzen

Without changing much tRAS down to 36 is possible, although not lower so far
There's a lot you can do still

EDIT:
Might want to attach a thaiphoon burner readout here too ?
Interesting might be PCB Layout ~ A0,A1,A2 bottom left corner

yrelbirb · Mar 16, 2020

Looking forward for your experiments. I have lpx kits as well (rated at 3000 mhz cl16)

Currently using these weird timings at 1.40v

https://prnt.sc/rh5wwc

It works (currently prime95 at 40 min mark @ smallfft) and gives 67 ns latency on Aida64.

Don't know how wonky or okay my timings are. They say it can do exceptional tRAS but why is that actually ? Same goes for tRP (a report i saw put it at 14).

It passed memtest86 too... Weird but will look forward to it!

Veii · Mar 16, 2020

yrelbirb said:
Looking forward for your experiments. I have lpx kits as well (rated at 3000 mhz cl16)

Currently using these weird timings at 1.40v

https://prnt.sc/rh5wwc

It works (currently prime95 at 40 min mark @ smallfft) and gives 67 ns latency on Aida64.

Don't know how wonky or okay my timings are. They say it can do exceptional tRAS but why is that actually ? Same goes for tRP (a report i saw put it at 14).

It passed memtest86 too... Weird but will look forward to it!

Might you attach the timings and a taiphoon burner report ?
If you have hynixMFR - the timings in the signature can be reused for you
They where from 2666C16 vengeance kits
67ns sounds like ryzen 3rd gen - on 2nd gen unless you rock 3400MT/s , this low latency won't be possible
For Zen 2 use this tool:

EDIT:
Or are you actually on intel ?

yrelbirb · Mar 16, 2020

Veii said:
Might you attach the timings and a taiphoon burner report ?
If you have hynixMFR - the timings in the signature can be reused for you
They where from 2666C16 vengeance kits
67ns sounds like ryzen 3rd gen - on 2nd gen unless you rock 3400MT/s , this low latency won't be possible
For Zen 2 use this tool:

EDIT:
Or are you actually on intel ?

im on ryzen 2700x and a b450 board

Veii · Mar 16, 2020

yrelbirb said:
im on ryzen 2700x and a b450 board

Ooh Micron B-Die, this is interesting
I haven't worked with them that much - but they are still 20nm, soo your 1.5v vDIMM range counts

This are your timings but fixed a tiny bit

And this is some set you can try if you're bored ~ 1.48v vDIMM, 1.075v SOC, 30-24-24-24 CAD_BUS, 53ohm ProcODT

For both sets tRTP 7 & 8 works - you need to play find out what your kit likes
Even 6 should be possible but it's unlikely

yrelbirb · Mar 16, 2020

thank you, i will try your suggestions

i've tried large fft test and tried to use my computer in the same time so got a crash around 13 min mark (a bit overkill, downloading call of duty from battle.net and watching a youtube video. i dont know if it can destabilize stress test however)

then tried again, this time using no programs in background and it went 40 min without errors but i got bored

i tried touching the stickings while lots of stress tests and they seemed very cool so i can up the voltage i guess? but a lot of comments i saw suggested these b die micron kits do not like high voltage or something like that

Veii · Mar 16, 2020

yrelbirb said:
thank you, i will try your suggestions

i've tried large fft test and tried to use my computer in the same time so got a crash around 13 min mark (a bit overkill, downloading call of duty from battle.net and watching a youtube video. i dont know if it can destabilize stress test however)

then tried again, this time using no programs in background and it went 40 min without errors but i got bored

i tried touching the stickings while lots of stress tests and they seemed very cool so i can up the voltage i guess? but a lot of comments i saw suggested these b die micron kits do not like high voltage or something like that

Then stick till 1.48v max
Usually i notice not voltage liking kits have a range of 1.47ish
1.46v shouldn't make issues on 20nm at all

Timings i posted had yours as inspiration but couple of stuff was messed up predicted
Only worried that first one's tRFC is too low - but you'll try, math says it will work ~ let's see, else i'll look for ~350ns tRFC @ this MT/s

2nd set has a lot of headroom, i would've need to mess up tRDWR/tWRRD to give it post issues or too low tRTP
Please try, for science

About the stresstest issue, haven't you tried 1usmus's TM5 or HCI Memtest before ?
Normaly on memory stresstests you shouldn't use memory :ninja:
Attached TM5 with changed test duration to 20 cycles, taken from here

yrelbirb · Mar 16, 2020

Veii said:
Then stick till 1.48v max
Usually i notice not voltage liking kits have a range of 1.47ish
1.46v shouldn't make issues on 20nm at all

Timings i posted had yours as inspiration but couple of stuff was messed up predicted
Only worried that first one's tRFC is too low - but you'll try, math says it will work ~ let's see, else i'll look for ~350ns tRFC @ this MT/s
2nd set has a lot of headroom, i would've need to mess up tRDWR/tWRRD to give it post issues or too low tRTP
Please try, for science

About the stresstest issue, haven't you tried 1usmus's TM5 or HCI Memtest before ?
Normaly on memory stresstests you shouldn't use memory :ninja:
Attached TM5 with changed test duration to 20 cycles, taken from here

Okay, i can go for usmus membench as well, which mode would be best suited for gaming stability? easy/memtest/defaults ?

extra question,

20-24-24-24 CAD_BUS

what are these values actually representing?

Edit: Your first set of suggestions didnt work

pc didnt boot and had to reset the cmos from mobo

my theory is about tRC. whenever i meddle with this setting it cannot boot

Veii · Mar 16, 2020

yrelbirb said:
Okay, i can go for usmus membench as well, which mode would be best suited for gaming stability? easy/memtest/defaults ?

extra question,

20-24-24-24 CAD_BUS

what are these values actually representing?

Nah just use TM5, 16gb take about 1h:30min
https://www.overclock.net/forum/11-...therboards/1627407-asrock-x370-taichi-overclocking-thread-698.html#post28234888 i've talked here a bit about it
Findings belong to 1usmus, Reous & cm87

• CLKDrvStr - Helpful gainst Post Problems with low ProcODT (recommendation less Ω)
• AddrCmdDrvStr - Does influence stability quite a bit (recommendation 30 or higher - find it out single handed)
• CsOdtDrvStr - It's quite similar to how RTT_NOM works (try out what runs better - to be known, higher values increase the chance of POST issues)
• CKEDrvStr - against waking up from sleep issues (more is better, but does increase thermals by quite a bit)

It's also to keep in mind, higher vDIMM requires higher CAD_BUS resistance
But that starts to matter with cheaper boards, or very high MT/s
1usmus explained it as "Data Eye" values - it's hard to explain in short words
The guide on TechPowerUp should cover it
Overall 24-24-24-24 runs always well, 30-24-24-24 on lower end boards when you have more vDimm on
20-24-40-30 for unstable HynixMFR kits and 24-20-20-24 still for b-dies, although you can go to 120-20-20-24 on 32gb dimm (dual rank) b-dies with >3733 or 48-20-20-24 for >4200MT/s SingleRank b-dies

yrelbirb · Mar 16, 2020

Another question, is it important to enter all the values one by one? Also, max tWTRL for my board was 15 while you suggested 16

I tried 3 possibilities,
c
tRC at 48 and tRFC at 287 = no boot
tRC at 58 and tRFC at 287 = no boot
tRC at 58 and auto tRFC = boot
only tRF at 287 and others default = no boot

in conclusion this ram cant boot at tRC at 48 and tRFC at 287 for time being

i tried upping the voltage to 1.45 as well but to no avail

DAM20 · Mar 16, 2020

Dasa said:
Try
tRRDS, tRRDL, tFAW 6, 6, 24
Also see if you can lower tRFC

Last time I change tFAW didn't post, gonna try again tho, had different settings.

I played with tRFC a bit, stock is 560, 500 booted a few times but anything less was a nope.

My actually good B-Die is running tRFC 264, the vegeance didn't like it.

DAM20 · Mar 16, 2020

Veii said:
tRFC 420-312-192 will work on this set without any change
If not, 540-401-247 is the highest that will run alright

Lower both SCL to 4, and use tWRRD 5 (need to try if 4 runs)
tFAW should be 36 or 28 (unless you change tRRDS)

Interesting would be rather SiSoftware Sandra Multi-Core Efficiency test (detailed result)
Next to Aida64, to showcase effective memory perf , while SiSandra would also have to do with your ryzen

Without changing much tRAS down to 36 is possible, although not lower so far
There's a lot you can do still

EDIT:
Might want to attach a thaiphoon burner readout here too ?
Interesting might be PCB Layout ~ A0,A1,A2 bottom left corner

Stock tRFC is 560 and I think something like 416-256. Anything less than 500 doesn't post or give errors, I'll try anyway but maybe 540-401-247 could work.
I'm not the confident about changing tFAW lul, last time had no post. Let's go anyway!

Aida and Typhoon coming right up! The rest I didn't even know they existed until now

Veii · Mar 16, 2020

yrelbirb said:
Another question, is it important to enter all the values one by one? Also, max tWTRL for my board was 15 while you suggested 16

I tried 3 possibilities,
c
tRC at 48 and tRFC at 287 = no boot
tRC at 58 and tRFC at 287 = no boot
tRC at 58 and auto tRFC = boot
only tRF at 287 and others default = no boot

in conclusion this ram cant boot at tRC at 48 and tRFC at 287 for time being

i tried upping the voltage to 1.45 as well but to no avail

You can skip tWTRS tWRTL then or try a lower one
Yes tRFC is too low then

Let's see
get back to tWTRS 4/tWTRL 12, tWR 12, tRFC 384-285-176, tRTP 8
Yes you need to put all stuff in, timings are interconnected

tRFC has a bit of wiggle room, as it's just the "world delay" or so called "refresh cycle"
But a tight one delivers faster perf, and let's you use lower SCL

Doublechecked everything, i actually messed up hard with tRFC oops
tRFC2 and 4 where wrong, because it mixed two different ways ~ sorry, glad i noticed before it's published
2nd preset for 3467 is correct, just edit tWTRS 5/tWTRL 14, tWR 16, tRTP 8

yrelbirb · Mar 16, 2020

Btw, despite setting those Cadbus values to 24 ohm, Ryzen timing checker shows them as 120 ohm. Any tips on that? (all 4 shows as 120 ohm from timing checker)

everything's mixed and tangled up

btw my board is gigabyte b450 gaming x. could it be considered mid end or would i assume it as a low end board? or in between?

Veii · Mar 16, 2020

yrelbirb said:
Btw, despite setting those Cadbus values to 24 ohm, Ryzen timing checker shows them as 120 ohm. Any tips on that? (all 4 shows as 120 ohm from timing checker)

everything's mixed and tangled up

btw my board is gigabyte b450 gaming x. could it be considered mid end or would i assume it as a low end board? or in between?

This is just a readout error, don't worry
It's SMU incompatibility (AGESA)

MSI B450-A Pro, Asrock Pro4
Most of the B350 boards, Asrock B450M-HDV, yours would fall in that B450 category
(depends on the pcb layout rly)
But as you aren't going for 3467 first gen ryzen, or 3700 2nd gen, there is no need to go over 30-24-24-24

DAM20 · Mar 16, 2020

Alright, tRRDS-6 tRRDL-6 tFAW-24 540-401-247 SCLs-4 tWRRD-5 boot! Bad news is that I can only run memtest this night, so will took a bit before knowing if is stable :/

Added the Typhon and Aida files, the benchmark is a month old, done with the same settings but at 1.45v.

DAM20 · Mar 16, 2020

Tested 2 times again in userbench, nothins as really changed, still in margin of error. Anyway I'll keep the lower tRFC if stable.

Veii · Mar 16, 2020

DAM20 said:
Tested 2 times again in userbench, nothins as really changed, still in margin of error. Anyway I'll keep the lower tRFC if stable.

That's why you should use SiSoftware Sandra

Even Aida64 will show margin of error

Test this later:

tRTP can be 8 or 7

Next step is then

As speed upgrade, tRTP can be 8 or 6, and tWR could drop to 12 technically
Move in the 1.46-1.48v region here
tRCDRD can be 19 only if you disable GearDownMode
Else odd timings will round up, and break tRAS & tRC, which desyncs tRFC at the end
EDIT: oh F , tWRRD is 5 on the last sample - by going tRCD 18 and SCL 3

Vengeance LPX Overclock Settings (Trash B-Die)

DAM20

Attachments

Dasa

Veii

yrelbirb

Veii

Attachments

yrelbirb

Attachments

Veii

yrelbirb

Veii

Attachments

yrelbirb

Veii

yrelbirb

DAM20

DAM20

Veii

yrelbirb

Veii

DAM20

DAM20

Attachments

Veii

Top Contributors this Month

Recommended Communities