[Sammelthread] Samsung 8Gbit DDR4 C-Die K4A8G085WC - OC-Ergebnisse im Startbeitrag! [updated]

[Dubstec]

@Dubstec Schau mal auf den Aufkleber der Module, ob der Lot Code wirklich auf -8810C endet oder vielleicht nicht doch eher auf -8810D. Für S8C wäre schon recht ungewöhnlich, dass 4000 @1.4V so problemlos laufen, für S8D nicht so sehr. Außerdem Glückwunsch, wenn du 4000 1:1 tatsächlich stabil und ohne WHEA nutzen kannst

Danke dir für deine schnelle Einschätzung werde ich direkt später mal schauen.
Aber 1.4V wären kein Problem bei 4.000 cl18? Ich werde nochmal einen Test machen die Tage auf errors etc bisher gabs keine Abstürze oder FPS Einbrüche usw.
Finde meine 38-40 grad auch recht gering ehrlich gesagt wenn ich so was von 50-60 grad teils lese. Jetzt im Multimedia Betrieb mit chrome 10gb Belegung etwa 36 grad.

 

[emissary42]

Für S8D wäre die Spannung unproblematisch. Wenn es tatsächlich S8C sind, dann muss man schon zwei Module mit sehr hoher Spannungstoleranz erwischen, um 4000 bei 1.4V zu betreiben. Es ist also durchaus möglich, nur eben sehr unwahrscheinlich. Solange die Einstellungen aber stabil und die Temperaturen im grünen Bereich sind, brauchst du dir deswegen keine Gedanken machen.

 

[Dubstec]

Also, laut Sticker neue Zahlen im Raum 8410D (?) was haben wir denn nu?
Hab noch ein „verwaschenes“ Bildchen wollte nicht aufmachen aber auch da 8410D..

 

[emissary42]

In jedem Fall bist du im falschen Sammelthread unterwegs. Die 4 verwundert mich etwas, aber hier geht es generell um S8D:

[Sammelthread] - & OC-Test mit Samsung 8Gbit D-Die (K4A8G085WD)

[Sammelthread] & OC-Test mit Samsung 8Gbit D-Die (K4A8G085WD) Einleitung Ähnlich wie zu Anfang auch sein Vorgänger ist dieser IC Typ bisher fast ausschließlich auf OEM Modulen von Samsung selbst zu finden. Im Gegensatz zu 8Gbit C-Die besitzen D-Die (wie ich sie im Folgenden abkürzen werde)...

www.hardwareluxx.de

 

[Dubstec]

Ok also ganz wilde Zahlen Sache.. D Die der als C Die gelesen wird und potentiell B Die gewesen wäre..

Gut zu wissen werde dort lesen, kurze Einschätzung wie ist B C D zu Staffeln von „gut zu naja“ und meine Werte dafür ok? CL 18-22-22-22-42 4000mhz 1.4V ca 40 Grad für scheinbar D Die?

Danke und dann werd ich mal da weiter lesen und zur Not da weiter fragen!

 

[emissary42]

Deine Werte sind absolut typisch für S8D. Schau dir z.B. mal die Ergebnisse von @Reous auf AMD an, findest du direkt im Startbeitrag.

 

[snck]

Hi, sorry for posting in english, but i dont really know german, even tho ive been learning it for many years now :P (school).
I've been struggling for almost a year now, to get a stable system with c-die kits with various of OC settings, including xmp setting too. My problem is kinda weird, never seen anything like that reading this thread or reading reddit, or anywhere else. I can pass many cycles in tm5, either 1usmus or anta777 config, can pass 14h of HCI and p95 large ffts for 2h (since that was the recommended amount of time in post i read.). But something is off, sometimes it crashes, for example, after closing Valorant, then after 3 mins it will once again crash while typing in discord and doing nothing else. So far, had two types of crashes, one is freeze+looping sound and after some seconds it restarts, and now what is happening, black screen+buzzing sound for couple of seconds and then it restarts . U might say its not ram. Thing is, I cleared CMOS and been running everything fully stock for almost a week now without any crashes (before it were crashing almost everyday), so I assume it might be really those rams. But cant get them fully stable on either xmp or my settings. Things i tried are formatting windows, changing bios settings (updating and rolling back), disabling all services and removing start up apps, and ofc, various of voltage and timings settings that has been checked from other users from this forum or reddit too (ik that something that has been working for someone else doesnt mean it will work for me, but i wanted some starting point). So after my long essay (which im sorry for ), here are my current specs:

MOBO: B450 Tomahawk (current AGESA - 1.0.0.6 <still crashes on oc | before rollbacking - 1.2.0.2 <was worse than 1.0.0.6 in my feeling)
CPU: Ryzen 5 2600
Ram: F4-3600C18-8GTZRX (tried various speeds on those rams, starting from 3000 ending at 3466 | voltages on dram were 1.35 in every settings, VSOC was around 1-1.1 and VDDP 0.9-1. ProcODT ive tried were 40 up to 60)

So what I'm asking for is, what should i do ofc . I feel like after those long months ive come to starting point which is fully stock settings. What should i do first? What freq, timings, voltages etc? I know that posting my full settings would be helpful, but there was many of them, and none turned out to be stable apparently .



Edit: Aight, rollbacked to 1.0.0.4 B, and set my OC. I'll see if its gonna be stable or not. Also, lemme know if there is something that could be changed for stability or performance wise. No info in ZenTimings, but also set LLC to level 3.
Edit 2: And it crashed today. Passed 12h of tm5 anta777, but not playing valorant. It's making me sick.

 

[athosdewitt]

Hi, sorry for posting in english, but i dont really know german, even tho ive been learning it for many years now :P (school).
I've been struggling for almost a year now, to get a stable system with c-die kits with various of OC settings, including xmp setting too. My problem is kinda weird, never seen anything like that reading this thread or reading reddit, or anywhere else. I can pass many cycles in tm5, either 1usmus or anta777 config, can pass 14h of HCI and p95 large ffts for 2h (since that was the recommended amount of time in post i read.). But something is off, sometimes it crashes, for example, after closing Valorant, then after 3 mins it will once again crash while typing in discord and doing nothing else. So far, had two types of crashes, one is freeze+looping sound and after some seconds it restarts, and now what is happening, black screen+buzzing sound for couple of seconds and then it restarts . U might say its not ram. Thing is, I cleared CMOS and been running everything fully stock for almost a week now without any crashes (before it were crashing almost everyday), so I assume it might be really those rams. But cant get them fully stable on either xmp or my settings. Things i tried are formatting windows, changing bios settings (updating and rolling back), disabling all services and removing start up apps, and ofc, various of voltage and timings settings that has been checked from other users from this forum or reddit too (ik that something that has been working for someone else doesnt mean it will work for me, but i wanted some starting point). So after my long essay (which im sorry for ), here are my current specs:

MOBO: B450 Tomahawk (current AGESA - 1.0.0.6 <still crashes on oc | before rollbacking - 1.2.0.2 <was worse than 1.0.0.6 in my feeling)
CPU: Ryzen 5 2600
Ram: F4-3600C18-8GTZRX (tried various speeds on those rams, starting from 3000 ending at 3466 | voltages on dram were 1.35 in every settings, VSOC was around 1-1.1 and VDDP 0.9-1. ProcODT ive tried were 40 up to 60)

So what I'm asking for is, what should i do ofc . I feel like after those long months ive come to starting point which is fully stock settings. What should i do first? What freq, timings, voltages etc? I know that posting my full settings would be helpful, but there was many of them, and none turned out to be stable apparently .



Edit: Aight, rollbacked to 1.0.0.4 B, and set my OC. I'll see if its gonna be stable or not. Also, lemme know if there is something that could be changed for stability or performance wise. No info in ZenTimings, but also set LLC to level 3.
Edit 2: And it crashed today. Passed 12h of tm5 anta777, but not playing valorant. It's making me sick.

Hi, i know it's not fully a fix for your issue but can you try for a while 2933MHZ cr1 14-17-17-35 tRC 52 GDM off 1,35v while leaving everything else on auto? Stock cpu and gpu preferably, black screen+buzzing could be related to a faulty gpu too u might want to check that in first place with ram at stock settings.

 

[snck]

Hi, i know it's not fully a fix for your issue but can you try for a while 2933MHZ cr1 14-17-17-35 tRC 52 GDM off 1,35v while leaving everything else on auto? Stock cpu and gpu preferably, black screen+buzzing could be related to a faulty gpu too u might want to check that in first place with ram at stock settings.

Hi, thx for reply . I did something else, and followed once again guide about ocing rams and finding baseline on higher freqs (https://github.com/integralfx/MemTestHelper/blob/oc-guide/DDR4 OC Guide.md). Right now im on 3400 16-20-20-40, everything else auto timings. So far, its good. Also, might found out why its only happening in Valorant. Turns out it's bad timings (and not freq) on app with AVX instructions, and since Valorant is probably only app that uses it in my example, it explains why PC crashed only in this game. There is reddit thread that explains it, and those crashes can look, as you said, like faulty gpu, psu and cpu (https://www.reddit.com/r/VALORANT/comments/hcnhcv/valorant_crashes_probable_cause/ - "debunking" comment is where it starts to be interesting, since OP explained it a bit more there.). So far, it passed 2h of y-cruncher on AVX2 and OCCT AVX2 as well (ofc did HCI and other as well, but AVX2/AVX is my main concern rn). Dunno if tm5 and hci is sensetive to those, apparently not, since i could pass them overnight and couldnt play Valorant . Haven't crashed yet, but will wait a week at least to tight some timings. Coming back to CPU and GPU, gpu passed heaven, furmark and occt like a champ, cannot be it, and CPU passed many hours in p95 and OCCT as well. Would be cool if u let me know if there is any settings i should change (i know that tRC and tRFC is super high rn, but would like to wait with them).

Edit: well maybe i was wrong xd crashed in first 30 mins of Valorant. Maybe Val hates c-dies. Gonna try your suggestion then.

 

[NemesisR55]

Hello all,
I'm not entirely sure my ram kit is C die but I have stability issues when XMP enabled.
I did try manualy overclock with XMP settings with1.37v but still getting a bsod after few days. (I did 8 pass memtest86 with no errors.)

My system:
- Ryzen 5800x
- MSI b550 a pro with latest bios AMD ComboAm4v2PI 1.2.0.5
-2x16gb a2 b2 slots

How can I make my ram more stable?

EDIT: I made it stable with

VDIMM :1.32V
VSoC: 1.1V
VDDG CCD: 0.950V
VDDG IOD: 0.950V
CLDO VDDP: 0.900V

other settings are same as xmp profile 1.

 

[emissary42]

In general my recommendation is to not use the "calculator". You also shouldn't have to increase the Vdimm for 3200 CL16, since S8C can typically clear that with less than that (and even OEM modules don't need more). So just start over with manual OC from scratch and try to keep Vdimm low, maybe even starting with 1.3V instead of 1.35V. Testing both modules individually can also help if you suspect one of them might be faulty. The first post has settings from other others that you can use as pointers / for orientation, but esp the required voltages can and will differ.

 

[snck]

Ok, crashes are probably from my house electricity, since my pc crashed in same moment as mibox (also other devices dont really work as they should). Sooo, what settings should I change first?

 

[emissary42]

Well maybe fix your house electricity first, then?

 

[snck]

Rn I'm looking for a UPS, no worries :P

 

[RockThePylon]

Just wanted to drop a thanks for everyone who shared their experience here.

I have a kit of Ripjaws V F4-3200C16D-32GVK, Thaiphoon Burner identifies them as B-die, and after having a very hard time getting any positive results, I pulled off the heatspreaders to check.

And that is definitely C-die! So I found my way here, and made some progress.

Notes on my RAM;

1.28-1.30v is the sweet spot. Lower worked okay, higher would immediately cause errors.
3466mhz was the best it would do. It would post up to 3800mhz, but errors in testing.
360ns was as low as tRFC would go. Default of 350ns would not post above 3200mhz.
I needed to increase voltages for VCCSA/IO as timings got tighter. VCCSA = 1.20v VCCIO = 1.15v, I will test lower but default would fail testing.
tREFI might be able to go higher, but my goal to be below 50ns latency was met, so I am done for now.
tRFC may also go lower, I suspect to poor heatspreader contact (see above) is trapping heat. I will look for a solution and test further.

Overall, good improvements and I had fun. Thanks again!

 

[emissary42]

tREFI and tRFC are two timings that are temperature sensitive. So min-maxing these can require active cooling of the memory modules.

I am a bit surprised, that you had to settle with 3466. In most cases 3600 18-20-20 or 18-21-21 is pretty easy to achieve. While the Gigabyte Z390 motherboards are not exactly great memclockers, in general they shouldn't have issues with that frequency. From my experience with the ASUS ROG models dual rank modules can sometimes require some BIOS juggling, to find a version that works best at high frequencies.

 

[RockThePylon]

I am a bit surprised that you had to settle with 3466. In most cases 3600 18-20-20 or 18-21-21 is pretty easy to achieve.

In order to boot 3600 18-21-21, I had to change "Memory Enhancement Setting" to "DDR-4500+". So I think you are right, and the Gigabyte board isn't very good at determining secondary timings automatically.

When I improve the cooling, I will try 3600 again. I may have missed something on the secondary timings that caused errors.

 

[Gabbygytes]

Hello there, greetings from the Philippines! (Pls note: Im new to ram overclocking)

I found out about this thread when people from reddit told me that my ram was actually a C-Die, not B-Die. For 3 months, I've been running my ram voltage at 1.45v. Its kinda funny ngl.
Thaipoon Burner reports that my ram has a B-Die but its not. I'm glad someone pointed it out that thaipoon isn't accurate.

Then, recently I overclocked my ram from 3200 CL16 v1,35 to 3600 CL18 v1.35. I left all the subtimings on Auto. I also did 3800 CL18 v1.35 but random restarts occur so I reverted it back to 3600 CL18. I performed HCI Memtest + Testmem5 overnight and there were no errors. However, my concern is its latency. AIDA64 Benchmark shows its slow at 82.9ns. It's quite slow compared to what others done in this thread.

What do you think about this? I want to push its performance to its limit.

 

[Krümel]

Welcome Gabbygytes,

You could try to lower trfc = 560 (310 ns), tRAS = 38, tRC = 58
In Post No. 1 of this thread are various Screenshots of possible timings.
Have fun!

Greetz
Krümel

 

[Reous]

Also if possible disable PowerDown for better latency.

 

[Gabbygytes]

Welcome Gabbygytes,

You could try to lower trfc = 560 (310 ns), tRAS = 38, tRC = 58
In Post No. 1 of this thread are various Screenshots of possible timings.
Have fun!

Greetz
Krümel

Also if possible disable PowerDown for better latency.

I lowered tRFC and tRAS, disabled PowerDown but I didn't touch the tRC (kinda scared).
Latency seems a bit better now. Its at around 68-71ns.

 

[Krümel]

but I didn't touch the tRC (kinda scared).

Don´t be afraid, just do it.

 

[Gabbygytes]

Don´t be afraid, just do it.

It didn't boot but I saw two lights blinked on the EZ Debug Led. I think they're BOOT and RAM.

 

[Krümel]

That´s unusual. I thought tRC = 58 was on the safe side.
Try tRC = 70 and if you want try lower timings later.

 

[Gabbygytes]

That´s unusual. I thought tRC = 58 was on the safe side.
Try tRC = 70 and if you want try lower timings later.

Epic. tRC at 70 works.

 

[Gabbygytes]

That´s unusual. I thought tRC = 58 was on the safe side.

Try tRC = 70 and if you want try lower timings later.

I did tRC=62 earlier (anything below 62 didn't post ☹️), but it shutdown after an hour during testing.

Now, I'm testing at tRC=64 for 6 hours and nothing happened yet. (Tbh im kinda lost and. idk what i am doing xd)

 

[emissary42]

@Gabbygytes tRC 65/65 is within the normal range for DDR4-3600 with S8C. When min-maxing timings like tRC and tRFC properly testing stability is recommended, since these timings can be temperature sensitive and are on the slower side of triggering errors.

If you don't know what you are doing at all, I would recommend starting with a guide like the memtesthelper one:

MemTestHelper/DDR4 OC Guide.md at oc-guide · integralfx/MemTestHelper

C# WPF to automate HCI MemTest. Contribute to integralfx/MemTestHelper development by creating an account on GitHub.

github.com

Afterwards you should be able to find all the required information about timings etc from the screenshots in the first post.

 

[Gabbygytes]

I followed some of the advice from this thread and it helped me a lot.

Btw, my system is:

• Ryzen 5 5600g
• MSI A520M-A PRO
• Avexir Core 2 Memory

After tweaking the timings, I was able to hit 64-66ns in AIDA64. My membench results also got better. Before, it was at around 138-140, now it's 120-122. Here are the results:

Gaming performance also increased, especially in GTA V (High Settings):

Stock XMP (3200 CL16) - Average FPS: 43
This OC - Average FPS: 57

I also performed a few test, and it passed without any errors.

AIDA64 Stability Test for 2 Hours
Prime95 Large for 8 Hours
HCI Memtest 1000% Coverage
Testmem5 Extreme 3 Cycles
Testmem5 Default 20 Cycles
OCCT for 1 hour

Honestly, without you guys, I'd not be motivated enough to overclock my crap ram. I guess I'd end up buying a new ram if I wasn't able to find out about this forum.

I also learned so many things about overclocking ram during the process, so thank you so much guys!

Beitrag automatisch zusammengeführt: 03.02.2022

Also, I have a question...

What does tRDWR and tWRRD do? I left them on auto.
Are there any rules about them?

 

[emissary42]

Honestly, without you guys, I'd not be motivated enough to overclock my crap ram. I guess I'd end up buying a new ram if I wasn't able to find out about this forum.

That is something that happened a lot back in the days (maybe still is, even to this day). Many gave up on their "crappy" memory after figuring out that either it was not actually S8B like TB suggested and/or b) the recommended settings from the "calculator" for some reason did not work. Like with many things in life, the biggest obstacle is to adjust your mindset. Being willing to learn and the motivation to look for information is already half the way. The persistence to systematically go through with it might be time consuming, but can also be a lot of fun and very rewarding.

ps Good job and nice increase in FPS!

 

[jvu316]

Hello! Sorry for writing English.. I really need help. I just recently finished a custom loop 5900x/ Dark Hero. My RAM is not easily accessible due to a tight fit and I will need to disassemble my whole loop to get it out to replace. I would like that to be a last resort. I was guided here by Reddit /overclocking.

My RAM is F4-3600C18-16GTZN. It is not stable at default DOCP. It will hard lock (screen freezes) but sometimes the screen just turns black (NO BSOD) and DRAM LED lights will turn on indicating it is a RAM issue. PC stays on until I switch off the psu off/on. Unless my default PBO settings is causing the crash. Everything is auto, but I have FMAX enhancer disabled. . Should I leave it off while trying to get a stable RAM OC?


I am willing to even go below 3600 as long as I can get a stable system.

The most I've gone was 16 hrs+ then it would black screen randomly. Again, no BSOD. For this setting, I believe the voltage was 1.45 and everything was auto. 3600/1800. I've never had issues with DOCP/XMP on my previous build,s this is a first.



VSOC 1.15 and Power Down Mode is disabled.

Edit: I read that DIMM Voltage is 1.33-1.35 for CDIE, but I have been running @ 1.4-1.465+., VSOC 1.15.

Will do some tests with VSOC 1.15, DIMM voltage 1.33v, PBO OFF

TM5 extreme anta777 (Did 3 runs with PBO off) PASS
Cinebench PASS
OCCT PASS
AIDA64 PASS
Did another RUN of TM5, it failed. "no signal monitor" PC stays on, never restart/BSOD