Overclock.net banner

10141 - 10160 of 10217 Posts

·
Registered
Joined
·
1,046 Posts
Thanks, I can't tell if it helped, got to 45 minutes Error 7 and then Error 6 about a minute later.
View attachment 2466106
Again ZenTimings is wrong, BGS is Enabled, BGS Alt is Disabled. 4xSR sticks.
I put tRFC2/4 back to Auto just to see, as that's what it was until a couple days ago.
This stuff is too complicated, I have no ideas

Edit:
vSOC to 1.125
vDDG CCD to 1.05
tRFC/2/4 to 336/250/154
Error 7 within 5 minutes
set tRCDRD to 16

see if your errors go away
 
  • Rep+
Reactions: bohemiocaa

·
Registered
Joined
·
1,046 Posts
Thanks for you tips bro, with

GDM Off
2T
Norm: 7
Wr: 1
Park:5

I have a lot of errors...... +800.

For the moment my best result (lower errors, was with this settings) With 1.39v , if i go with the same settings to 1.4.... more errors.
View attachment 2466100
Are these your sticks


Are these definately b-die ? I know "B-die Finder " resource says they are.

They could be a very bad bin, use tRFC set to 485 and see if the errors disappear
 
  • Rep+
Reactions: bohemiocaa

·
Registered
Joined
·
26 Posts
Was I supposed to select dual rank when using four single rank sticks for the purposes of the calculator?
I've heard people say you're supposed to, and that's what I'm doing, but I haven't seen anyone like 1usmus confirm it.
It really doesn't make sense though, because technically you're telling it you have four dual rank sticks, for quad rank, so I guess the calculator just doesn't actually support or have a setting for actual quad rank systems, if this is how it works.

1usmus really should clarify if that's the case, because anyone that actually does the correct thing like you is not getting the right settings, apparently
 

·
Premium Member
Joined
·
5,240 Posts
I've heard people say you're supposed to, and that's what I'm doing, but I haven't seen anyone like 1usmus confirm it.
It really doesn't make sense though, because technically you're telling it you have four dual rank sticks, for quad rank, so I guess the calculator just doesn't actually support or have a setting for actual quad rank systems, if this is how it works
Four Single Rank DIMMs run Dual Rank the same as two Dual Rank DIMMs do. It has to do with the memory controller. That is why two Dual Rank DIMMs are better than two Single Rank DIMMs.

Might take some Google to find out exactly why.


Edit: Ryzen Above: Best Memory Settings for AMD's 3000 CPUs, Tested

  • The best practical configuration is four ranks of DDR4-3600 at the lowest stable latencies. Reminder: Four ranks can be achieved from either four single-rank or two dual-rank DIMMs. And almost all X570 boards are Daisy Chain so two 2x16GB Dual Rank b-die is the way to go.
 

·
Registered
Joined
·
26 Posts
Four Single Rank DIMMs run Dual Rank the same as two Dual Rank DIMMs do. It has to do with the memory controller. That is why two Dual Rank DIMMs are better than two Single Rank DIMMs.

Might take some Google to find out exactly why.

Edit: Ryzen Above: Best Memory Settings for AMD's 3000 CPUs, Tested
I know, but we're talking about when you have four single rank dimms, you can still choose Memory Rank 1 or 2 in the calculator. Which if you say 2 with 4 DIMMs, you're technically saying you have four Dual Rank DIMMs. Quad Rank. A lot of people say that's what you're supposed to do even with four single rank dimms.

You can't really have 4 DIMMs and not run in dual rank, can you? So what is it asking when you can say 4 DIMM 1 Rank.
 

·
Registered
Joined
·
1,046 Posts
I know, but we're talking about when you have four single rank dimms, you can still choose Memory Rank 1 or 2 in the calculator. Which if you say 2 with 4 DIMMs, you're technically saying you have four Dual Rank DIMMs. Quad Rank. A lot of people say that's what you're supposed to do even with four single rank dimms.

You can't really have 4 DIMMs and not run in dual rank, can you? So what is it asking when you can say 4 DIMM 1 Rank.
It looks likes its not been factored into the calculator, you have understood it right though, as KedarWolf has explained.

4 sticks of single rank need same settings as 2 sticks of dual rank.\

Re brute force with air and voltage, generally does not help tRCDRD, that why i said to drop to 16 for testing purposes

:)
 

·
Registered
Joined
·
26 Posts
It looks likes its not been factored into the calculator, you have understood it right though, as KedarWolf has explained.

4 sticks of single rank need same settings as 2 sticks of dual rank.\

Re brute force with air and voltage, generally does not help tRCDRD, that why i said to drop to 16 for testing purposes

:)
Yea so Memory Rank 2 with 4 Single Rank DIMMs for the calculator, which can be counter intuitive, like it was for this guy, and me.

I just set that fan up, so I'm going to be heartbroken if I see another error. It is depressing to be constantly let down all day no matter what you do...
If I have to throw tRCDRD back to 16 that loses a lot of latency if I remember correctly, more than maybe any other setting?
I know this is stable:
2466141

I can't get FCLK 1900, would be nice to kinda make up for it
 

·
Overclock the World
Joined
·
1,984 Posts
@Veii Do you know what Error 3 is? Just got that for the first time 11 minutes in.
View attachment 2466118
BGS Enabled, BGS Alt Disabled.

Edit: Changed tRFC/2/4: 294/218/134 = Error 13, 11 mins in
Upped voltages and got 3 Error 10's in first 10 minutes
I'm looking at your case, much posts to process
Let's start with the acknowledgement, that you use 4x A2 B-dies on a Daisy-Chain Board
Where the slave set get's 1/3 of the ampere (current but not voltage one) ~ of the first set

Meaning you will either have to increase ClkDrvStrengh , or increase procODT (one of both)
And then look how to balance current

Mid-re'edit:
Anta777 and 1usmus_3 "error numbers" are apparently different
Many are identical but some tests are completely different
I am not sure what you run and what the errors would mean

Simple Test 0mb usually are refresh tests. A crashing memory would mean that either the resitance is too low with high current flowing through it
Or by distance/bad signal integrity ~ impedance is too low/unstable

Burst Test 4mb,
Are voltage related or cutoff related ~ timeout issue
i am really not sure what test file you let loop or run, but each test file has a different order of "test numbers"
They are not comparable - only the test types could be to some extend

Else,
Many tests which are not Simple Test 0 or Burst Test 4mb ~ are progress
EDIT:
They will not appear, if memory already fails "refresh stable" or Simple Test 0mb" tests
But i don't think you could drop tRCD to 14 on a Daisy Chain Board
The 2nd link is just too weak ~ maybe something can be done

GDM is helpful, because the internal Multiplex Driver -> Data Signal (MUX) is running at half speed with it enabled
Same goes for the Sense Amp -> I/O Gate
but this is one of the reasons why many timings are rounded in the hidden or "1.5T" exists
It will hide tho instability, same as higher than tRAS+tRP=tRC will
Better get GMD off 2T enabled, soo the "whole array of timings" will loop once ~ while MUX run at normal speed
Because you can optimize 2T sets to be lower and still outperform 1.5T rounding
If you take a look at part of the illustrative example
You can see that Data Line is Sinus Curve like. Up to exterior state the signal will pass or not pass ~ but it is variable IF it will pass and can corrupt
The Driver Signal has no sinus behavior ~ only a "falloff current = remain current in cells" behavior

Meaning,
Bad signal integrity (noise), not enough impedance, too strong reistance cuttoff
will not allow the data-signal to pass. soo you should work your RTT values
(probably lowering the strengthness of them)
Then increase procODT a bit with more SOC voltage but a loadline with stronger vdroop
And increase ClkDrvStrength between 40-120ohm
Usually for Threadripper and similar "many dimms" setups , you move in the area of >60 ClkDrvStrength
----------------------------------------------------------------------------
I would here also double tWR & tRTP for now
And start with SCL of 5 & tWRRD 1 + higher tRDWR 10?
Also increasing both tRRD_ & tWTR_ by +1 each
tRRD_ 5/7
tWTR_ 5/14
tFAW 20

tRFC up to *8
384-285-176
tWR double 16( 32! )
tRTP double 8 ( 16 )

VDIMM near 1.48 should be fine
2nd set should get around 1.41-42ish
(current will be applied but amperage will be different, just illustrative example)
Show proof that you can get tRCD_RD 14 stable with these higher settings and changed CAD_BUS + RTT
Then we can continue the talk :)
I feel like giving up and going back to 16-16-16-16-32-48 lol..
View attachment 2466069
Ran for over an hour fine before I left to sleep.
I don't know what to change, haven't messed with voltage related settings much cause I'm new and don't want to damage anything with things I don't understand, RTT/Proc/DrvStr all Auto.
Also ZenTimings is wrong, BGS is Enabled, BGS Alt is Disabled. 4x8gb B-Die Patriot.
But don't worry all too much
4x A2 kits are already a big challenge on T-Topology.
@fcchin might be able to share some experience and advices
on Daisy-Chain it's nightmare :D
but you got to learn from hard stuff :coffee:
 

·
Registered
Joined
·
293 Posts
Uh? It's only one page, one post, 6 steps, that I linked:

How to do Stable DDR4 Tuning on Ryzen: Finding the Appropriate Values of the Termination Resistance (Rev. 2)

I don't see other stuff...
Dunno, I counted 20 pages scrolling, didn't count the rows :))
 

·
Registered
Joined
·
31 Posts
I'm starting to feel really really bored about my current memory ram kit and trully thinking of replacing it by a new set to see if I can get rid of all the instability issues and maybe get a bit better performance.

I'm thinking on replacing current kit https://www.kingston.com/dataSheets/HX432C16FB4AK2_32.pdf which is only causing me troubles by this other one F4-3200C14D-32GTZ-G.SKILL International Enterprise Co., Ltd.

Do you think it's worthy? What can I expect from that kit? Better clocks, better latency?
 

·
Registered
Joined
·
6 Posts
Are these your sticks


Are these definately b-die ? I know "B-die Finder " resource says they are.

They could be a very bad bin, use tRFC set to 485 and see if the errors disappear
Yes bro, they are B-Die. I know they dont are the best bin o the b-die , but you think i cant get 3733 stable ?

I trow the towel with 3800, but 3733 ? :unsure:

Edit: A little better results with trfc 485, but errors.
 

Attachments

·
Registered
Joined
·
26 Posts
2466210

I was running this before your post @Veii
I think I'm just going to relax now for a bit, try to lower vDIMM over the next couple nights, 1.52v in BIOS unstable but 1.53v or 1.54v should be fine. Then I'll probably see if I can lower vSOC, vDDG, vDDP at all.
53.5ns AIDA.
They just need active cooling, doesn't matter how good your case airflow is
20201120_233056.jpg
Despite having 2x 140mm Noctua NF-A14 intake pretty close (not a long case) and 140mm CPU fan pulling air right off (it's practically touching leftmost DIMM) it's not enough.
Have to have one blowing right on
 

·
Registered
Joined
·
176 Posts
I think I'm stable now with the 4x8gb djr using dual rank fast settings 3800. 1.42v. Passed 20 cycles tm5 and no reboot corrupted file blue screens. Crazy how changing just a few settings fixed everything. Still have no idea why reboot issues were non existent for 2 days and then blue screen on every reboot.
 

·
Registered
Joined
·
1,345 Posts
So running Aida in safe mode makes a difference. I guess I have too much RGB bullshit running in the background.

AIDA64 3800MhzCL16 11212020.png
3800CL1611192020 1730HRS LATENCY.png
 

·
French loving Japan!
Joined
·
536 Posts
Any people having a combo running with 3500X and corsair 3200CL16 LPX with hynix chip from 1st gen ryzen times. Just played a bit with xmp in the bios of my B550 board but not tried Dram cakculator yet so looking for advice.apologize if i am too specific
 

·
Overclock the World
Joined
·
1,984 Posts
Any people having a combo running with 3500X and corsair 3200CL16 LPX with hynix chip from 1st gen ryzen times. Just played a bit with xmp in the bios of my B550 board but not tried Dram cakculator yet so looking for advice.apologize if i am too specific
If hynix MFR, check this picture
you can replicate RTT values and maybe find scaling similarities
This was under 1.56v, but it could need 1.62v on some cases

If HynixAFR, your max clocks move near 3600MT/s , maaybe 3700 ~ with rec voltage not beyond 1.45
If CJR , you can pretty much max out the IMC at 1900=3800MT/s ~ same rec voltage, peak aruond 1.48

MFR is 20nm , and it loves voltage ~ more than b-die soo you can safely push it to 1.62v
The remain voltage rulesets, well pretty much everything remains on Matisse with the 3500X
View attachment 2466210
I was running this before your post @Veii
I think I'm just going to relax now for a bit, try to lower vDIMM over the next couple nights, 1.52v in BIOS unstable but 1.53v or 1.54v should be fine. Then I'll probably see if I can lower vSOC, vDDG, vDDP at all.
53.5ns AIDA.
They just need active cooling, doesn't matter how good your case airflow is
Despite having 2x 140mm Noctua NF-A14 intake pretty close (not a long case) and 140mm CPU fan pulling air right off (it's practically touching leftmost DIMM) it's not enough.
Have to have one blowing right on
Wow that close.
It would question here if the thermal pads on these really touch the heatsink and probably remove one side of it like:
Keep in mind, they often use double sided-glue thermal pads
very sticky ones at least and you don't want to rip appart the ICs
Else , alone comparing both A2 vipers ~ if they really are A2 dimms - would be a worthful thing to do
Although once removed from the DIMM Slots, i'd probably wipe them once with isopropanol
As strong current causes an oxidation layer on their pins. Same goes for older 3xx series boards (CPU and DIMM pins)

tWR 12 is awkward for this tRFC, 14 would be better - just a bit slower
But is this really a good set 🤔
It's a bit too unstable at this high voltage with yet "high" tRFC
My vipers run 1.45 without any airflow at all - open bench with their "heatsink"
Really questioning how well yours touch the thermal pads of them

Well anywho, go for GDM off 2T now :D
Then we shall see how stable the CAD/RTT setup really is :)
Your tRFC alone with GDM on, should go one step further down at least

Probably get them out first and make a close up picture on the bottom of the dimm - the traces
OC'ing T-Force 4133 cl18 visible like here, just a bit further away
Who knows, some old 4400 vipers apparantly where A0 not A2
It would be bad luck overall, but you might have lucked out - as A0 are less sensitive and can be placed on the slave set
I'm starting to feel really really bored about my current memory ram kit and trully thinking of replacing it by a new set to see if I can get rid of all the instability issues and maybe get a bit better performance.

I'm thinking on replacing current kit https://www.kingston.com/dataSheets/HX432C16FB4AK2_32.pdf which is only causing me troubles by this other one F4-3200C14D-32GTZ-G.SKILL International Enterprise Co., Ltd.

Do you think it's worthy? What can I expect from that kit? Better clocks, better latency?
How much would you pay for that C14-14 set
Have you looked for similar 3600CL16-16 or CL15-15 sets ?
 

·
Registered
Joined
·
1,046 Posts
Yes bro, they are B-Die. I know they dont are the best bin o the b-die , but you think i cant get 3733 stable ?

I trow the towel with 3800, but 3733 ? :unsure:

Edit: A little better results with trfc 485, but errors.
Frendo,

Whats the reason you are running the Rtt values you have in the screenshot ?

I said previously to start with

RttNom / RttWr / RttPark @ 7/5/1

You wrote this previously (I missed this before)
GDM Off
2T
Norm: 7
Wr: 1
Park:5
You mixed up the settings, you have RttWr and RttPark the wrong way round ??

ProcODT 43.6 ohms -- You have this set
vSOC 1.065 - 1.075 -- You have your quite a bit lower, is there a reason for this? You must fix the Rtt first though!
vDDP 0.900V -- You are good here
 
  • Rep+
Reactions: bohemiocaa

·
French loving Japan!
Joined
·
536 Posts
If hynix MFR, check this picture
you can replicate RTT values and maybe find scaling similarities
This was under 1.56v, but it could need 1.62v on some cases

If HynixAFR, your max clocks move near 3600MT/s , maaybe 3700 ~ with rec voltage not beyond 1.45
If CJR , you can pretty much max out the IMC at 1900=3800MT/s ~ same rec voltage, peak aruond 1.48

MFR is 20nm , and it loves voltage ~ more than b-die soo you can safely push it to 1.62v
The remain voltage rulesets, well pretty much everything remains on Matisse with the 3500X

Wow that close.
It would question here if the thermal pads on these really touch the heatsink and probably remove one side of it like:
Keep in mind, they often use double sided-glue thermal pads
very sticky ones at least and you don't want to rip appart the ICs
Else , alone comparing both A2 vipers ~ if they really are A2 dimms - would be a worthful thing to do
Although once removed from the DIMM Slots, i'd probably wipe them once with isopropanol
As strong current causes an oxidation layer on their pins. Same goes for older 3xx series boards (CPU and DIMM pins)

tWR 12 is awkward for this tRFC, 14 would be better - just a bit slower
But is this really a good set 🤔
It's a bit too unstable at this high voltage with yet "high" tRFC
My vipers run 1.45 without any airflow at all - open bench with their "heatsink"
Really questioning how well yours touch the thermal pads of them

Well anywho, go for GDM off 2T now :D
Then we shall see how stable the CAD/RTT setup really is :)
Your tRFC alone with GDM on, should go one step further down at least

Probably get them out first and make a close up picture on the bottom of the dimm - the traces
OC'ing T-Force 4133 cl18 visible like here, just a bit further away
Who knows, some old 4400 vipers apparantly where A0 not A2
It would be bad luck overall, but you might have lucked out - as A0 are less sensitive and can be placed on the slave set

How much would you pay for that C14-14 set
Have you looked for similar 3600CL16-16 or CL15-15 sets ?
Thanks so much! Its MFR and i pushed the voltage at 1.4V so far in my light attempts to go beyond my current 3333mhz
Will try and report back, Danke nochmal
 
10141 - 10160 of 10217 Posts
Top