Overclock.net banner
9181 - 9200 of 12942 Posts

·
Registered
Joined
·
2,122 Posts
Trying out 1.2.0.3 again,

so far after several reboots its behaving at 4133/2067.

No stability tests as of yet, audio is great, no USB issues yet, waiting for a WHEA 18

🤣😄

** EDIT **
Back to BCLK I go......

:)
 

·
Registered
Joined
·
62 Posts
Do you guys believe that the "holes" or whea 19s are ever going away with new agesa / gpu drivers / windows versions? I am running 3600 cl 14 right now I can't run 3666 cl14, 1900 MHz flck is not even posting. 2000 MHz cl 16 runs smooth but troughs whea 19s.

I believe 2000 cl 16 would be my best setup but I don't feel comfy tuning everything with the wheas. I would be fine running 3800 cl 16 if the hole was gone.

Right now it just feels like wasting your time ocing ram if the infinity fabric isn't stable or troughing whea 19s ...
 

·
Registered
Joined
·
2,122 Posts
Do you guys believe that the "holes" or whea 19s are ever going away with new agesa / gpu drivers / windows versions? I am running 3600 cl 14 right now I can't run 3666 cl14, 1900 MHz flck is not even posting. 2000 MHz cl 16 runs smooth but troughs whea 19s.

I believe 2000 cl 16 would be my best setup but I don't feel comfy tuning everything with the wheas. I would be fine running 3800 cl 16 if the hole was gone.

Right now it just feels like wasting your time ocing ram if the infinity fabric isn't stable or troughing whea 19s ...
I think we are all in the dark!

Each agesa that passes nothing that us overclockers want bear fruition

:(
 

·
Registered
Joined
·
729 Posts
Do you guys believe that the "holes" or whea 19s are ever going away with new agesa / gpu drivers / windows versions?
I don't. I beleive it's 99% hardware limit of current DF implementation.
 
  • Rep+
Reactions: Mach3.2

·
Registered
Joined
·
134 Posts
DF-States, AMD CBS - NBIO - SMU , DF-States ~ has to be disabled in order to prevent, sleep to wake up overboost
Since GIGABYTE uploaded new BIOS for my motherboard - F61c i can finally try it too. There are so many new options...

Also, i still didn't tested [email protected], sice i was really busy lowering tWTR_L from 12 to 8 (i founded interesting what you told before that _S is just base and later used muliplier x2 or x3). So, i used my [email protected] and just lowered one value:

tWTR_L=8
Errors 5, [email protected] (explosion)

tWTR_L=8 + vDIMM 1.42V
Error [email protected] (at this stage i thinking that it's simply overheat, but now i can say it was wrong)

tWTR_L=8 +vDIMM 1.42V + Fan
Error [email protected] (So, 6 more cycles, but still error, but not 13, so no overheat?)

tWTR_L=8 + tCKE=9
Errors 10, 5 (explosion)

===== Some play with tCKE and Error list
tWTR_L=8 + vDIMM 1.42V + tCKE=9 + tRDWR=8 + tWRRD=1
Error [email protected]

tWTR_L=8 + vDIMM 1.42V + tCKE=9 + tRDWR=8 + tWRRD=2
Error... Can't run ZenTimings in Windows - Error in Kernel.dll in Event viewer

tWTR_L=8 + vDIMM 1.42V + tCKE=9 + tRDWR=8 + tWRRD=3
Error [email protected]

tWTR_L=8 + vDIMM 1.42V + tCKE=9 + tRDWR=9 + tWRRD=1
Error [email protected]

tWTR_L=8 + vDIMM 1.42V + tCKE=9 + tRDWR=9 + tWRRD=2
Error [email protected] (No Kernel error at ZenTimings run)

tWTR_L=8 + vDIMM 1.42V + tCKE=9 + tRDWR=9 + tWRRD=3
Error [email protected]

tWTR_L=8 + vDIMM 1.42V + tCKE=9 + tRDWR=10 + tWRRD=1
Error [email protected]

After that "pattern" look obvious and repeatable. Wrong path.
=====

tWTR_L=8 + vDIMM 1.44V
Error [email protected] So... it was only because of voltage. After i added +0.02V more - it was O.K. No overheat and no Fan used.
Rectangle Font Screenshot Software Technology
So, i play a bit with tRFC lowering and then try CL15. I need that [email protected] for base (default) profile. I want to use something when i not in a mood to play with TM5 and all this "memory stuff" since i can't use XMP-profile with DDR-4133. I have no boot with it.
 

·
Registered
Joined
·
159 Posts
Did a bunch of testing today. The results don't necessarily mean a lot to me, but may to some of you more experienced guys. You may be able to pick up a trend or pattern than I don't see. Here goes:

3733 Test 1
View attachment 2513540
3733 test 1CyclesError
Settings as pictured612, 10, 0 (all errors in cyycle 6)
tRCDRD 1610 cyclesno errors (pictured settings /w tRCDRD 16)
tRCDRD 14, Fix tFAW to 20 (4x tRRDS)-no test. only changed tfaw 21 to 20
tRCDRD 16, tFAW 20, tRDWR 8 / tWRRD 335, 14, 9, 14
tRCDRD 16, tFAW 20, tRDWR 9 / tWRRD 336, 13, 0, 15
Settings as pictured (tFAW 20)33, 5, 9, 0
pic, tfaw 20, 1.47 vdimm312, 2, 0, 3, 9
pic, tfaw 20, 1.48 vdimm315, 15, 14, 11
pic, tfaw 20, 1.45 vdimm211, 2, 4, 4,
pic, tfaw 20, 1.45 vdimm, ProcODT 40514, 15, 2, 55, 1, 4, 3, (1st 2 errors in cycle 4, rest cycle 5)

3733 Test 2
View attachment 2513541
3733 test 2CyclesError
Settings as pictured (procODT to 40)514, 15, 2, 5, 1, 4, 3, (1st 2 errors in cycle 4, rest cycle 5)
40 clkdrvstren, 40-20-24-2426, 10, 4, 3, 3, 3, 12, 0, 12, 2
+ csodtdrvstren, 40-20-30-2030, 14, 5, 5
Settings as pictured, 1.46 VDIMM315, 13, 11,
Settings as pictured, 1.44 VDIMM31000's of errors.

3800 Test 1
View attachment 2513542
3800 Test 1cycles2 complete cycles
Pictured settings20, 1, 4, 13, 0, 9, 14, 14, 1, 1,
VDIMM 1.5212, 2, 10, 0, 14, 14,
RttNom 621, 1, 11, 12, 13, 13, 15
VDIMM 1.48214, 10 - accidentally closed 1/2 way through 2nd cycle
tRAS 30, tRC 44, tFAW 42 (tCCD_l 7 x tRRDS)214, 11, 15, 0, 2, 10, 5, 5, 4, 14, 0, 15, 15, 6, 0, 6, 6, 6, 12, 0, 2, 2, 2, 10, 10, 10
returned tRAS, tRC, tFAW to prrior valuesdidn't test
Pic + 1.06 vddg iod, 1.48vdimm, 40 procODT26, 6, 12, 2, 2, 2, 10, 10, 5, 4, 4, 4, 3, 3,
RttNOM 6213, 9, 15, 6, 6, 6, 12, 12, 0, 2, 2, 2, 10, 5, 1, 4, 3, 13, 13, 9, 14, 0,
RttNOM 7, 1.1 VSOC16, 12,
1.125 vsoc, .9 vddp/vddg ccd,210, 6, 6, 2, 2, 2, 2, 10, 10, 10, 0, 5, 5, 5, 5, 1, 4, 4, 3, 3, 3,

3800 Test 2
View attachment 2513543
*note, cycles are ~9 minutes vs 6 (on 3733 or 3600 IF)
7 cycles / 65 mins. Errors: 0, 0, 12, 14
No errors until 45 min mark (in cycle 5), then 0, 0, back to back
then nothing until 58 min mark (in cycle 7), then 12 and 14.
max dimm temp was 42.8 c
Change your secondary timings, tRDRDSD and tRDRDDD to 5, tWRWRSD tWRWRDD to 7, I kept getting errors with these at 4 and 6.


Also: I get more fps from my gpu if I run the trfc = trc x8, ram runs a hair slower but the fps boost is real. You can use this handy dandy tool.


My timings for reference, in this case 42 x 8 = 336

2513602
 

·
Iconoclast
Joined
·
31,443 Posts
Indeed; the big latency gain was due to a very lucky training at boot seems... now it's at an average of 53.1ns, about 0.1ns better than Auto DPM LCLK..
Training variances make it almost impossible to pin down subtle changes in performance. Since there isn't any technical reason for the LCLK to influence memory performance--unless one's SoC is running into a TDC/EDC limit, or one is running out of memory and paging--the safest assumption is that it doesn't.
 

·
Registered
Joined
·
729 Posts
Has anyone got B1 revision chip?
 

·
Premium Member
Joined
·
6,378 Posts

·
Registered
Joined
·
729 Posts

·
Registered
Joined
·
3,105 Posts
Change your secondary timings, tRDRDSD and tRDRDDD to 5, tWRWRSD tWRWRDD to 7, I kept getting errors with these at 4 and 6.


Also: I get more fps from my gpu if I run the trfc = trc x8, ram runs a hair slower but the fps boost is real. You can use this handy dandy tool.


My timings for reference, in this case 42 x 8 = 336

View attachment 2513602
This genuinely sounds like you are just more stable than you were previously So you don't have auto-correction kicking in as much as opposed to 8 x trc = tRFC being better for games.
 

·
Registered
B550 gaming edge, 5800x, 5600x, 3600xt, 1600AF, 2070S, 4x8 3200c14, 2tb adata 8200sx PRO, 500gb 970e
Joined
·
575 Posts
Sorry for being vague, I meant CPU 19h family revision
are you speaking on the second line on the CPU heatspreader?

or would CPUz tell this?
(heat spreader says BG
(B0 but family is 19)
sorry for responding thought i might have just what you were asking for.
2513643
 

·
Registered
Joined
·
43 Posts
I have a new problem with my rams. My system wont boot up after it has been shut down for a while. Memory error led stays lid and I have to remove the sticks, clear cmos and put them again to be able to boot. This even happens on docp 3200 mhz. I have a strix b550-f, 5900x and 16x2 3200 cl14 samsung b-dies. I tried increasing cl and trcd to 15, tried increasing procodt. Any advice?


Sent from my iPhone using Tapatalk
 

·
Registered
Joined
·
159 Posts
This genuinely sounds like you are just more stable than you were previously So you don't have auto-correction kicking in as much as opposed to 8 x trc = tRFC being better for games.
Idk why it works, but I went from 315 to 327 fps in furmark with only that change. Now it seems to have taken and works for other speeds.
 

·
Registered
Joined
·
21 Posts
Hello everyone, I'm new to this forum. As one of many I have changed from intel to amd.
I'm still trying to get the most efficient and stable OC, so I looked through most of the pages in this thread.

My current platform is 5900x + ASUS Crosshair VIII Dark Hero + 32GB (2x16) G.Skill F4-3600C16-16GTZN.

Voltage:
DRAM: 1.41
SOC: 1.05
VDDG CCD: 0.900
VDDG IOD: 1.050
CLDO VDDP: 0.880

Here you can see my results based largely on these community advice - special thanks to Veii for his advice

2513655
2513656


I'am aware that my settings are on 1T (not 2T). However, on these settings my system is rock-solid (tm5 - 20 cycles, karhu 8000% + working heavily 3D/multi-app over 12-16h per day).

Looking at the experience of the people on this forum Iam convinced that I could get better results. The most important thing is that my goal is to get the most efficient but stable configuration that I can work on for up to 12 hours a day (I don't want to do OC to get the best benchmark results).

I will be grateful for any hints and advice on what I can do better with my configuration. BTW - of course I also play games sometimes 😉 ..in 1440p.
 

·
Registered
Joined
·
39 Posts
Yes, my 16-16-16-36 3600 B1.

View attachment 2513625
Hello, since my ram is B1 too, I shamelessly took your build. Now, I got 1 error in test #6. My TM5 is set as you recommended, changed to 1000%, 8 cycles.
According to @Veii 1usmus errors explanations, one single error in test #6 = "fix RTT values or give it +1 VDIMM". I would like to leave increasing VDIMM as the last possible option.
Please, any chance you could see if the RTT values are wrong in my case, or how to fix it?
 

·
Overclock the World
Joined
·
3,164 Posts
About DPM , LCLK and similar
Let me drop you this little tool
Bios Modding - Google Drive Tool1007.zip (tool.exe)

Have fun with it :)
Single CCD can run

To override allcore EDC throttle
You still should set Bios PBO EDC limit higher soo cache can boost up
But this overrides another stage of allcore throttling

This is interesting for you , just the plotting is a bit strange to use
And this is what you surely will like to see


Enjoy experimenting
Each of the 8 "Frequency Limiters" Sensors can be the cause of frequency loss
FIT PPT throttle usually appears because of something else, still searching for the value

DPM doesn't show up correctly and the ASRock ITX misses LCLK optionality to enable it. The Asus board had it
Some values are under a wrong multiplier but there are valuable things to check and read out
Effective Clock is correct for example
Also it shows fully how many CCDs and Cores you have ~ for all these mysterious samples out there


Copyright belongs to the ASUS Forums as XOC tool
I forgot who published it and which team maintains it.
No credit belong to me, just want to make it easier accessible to people :)

Waiting for 1.2.0.3A for the ITX board, then i'm gonna check out what can be done for the public
113 Binning value is already good, 118 is fantastic - haven't seen 120+ so far


DPM 2-1-1-2 doesn't show it
testing 2-2-1-1 too ~ but this is intercore stuff. SiSoftware Sandra is your best friend for such
This one is neat, but only ASUS boards let access to RAW VRM controller

Apparently there should be an I²C port on the ITX , but i haven't checked
=========================================
MEM-OC thread, soo here are some valuable RTTs for Rev.E@1.6v or XMP under 56.50
2513725

I lack a lot of options on the ASRock bioses - but 1203A update needs to be out, before it has any modding potential or consideration
* on the Drive link above, AMI_FLASH Win can flash unsigned bioses ~ as long as the DeviceID remains identical. It bypasses AMDs SPI lock but is not able to cross-flash bioses
Backups of it are sector based and tiny ~ do no use it for backup purposes, unless you can stitch bioses together (HEX)
 
9181 - 9200 of 12942 Posts
Top