[Sammelthread] Intel DDR5 RAM OC Thread

Wenn Du diese Anzeige nicht sehen willst, registriere Dich und/oder logge Dich ein.
Karhu ist 45Minuten durchgelaufen, das langt mir erstmal an stabilität da ich keinen Sponsor habe der mir den Strom für ne 24 Stunden Session bezahlt :ROFLMAO:
Zum Memtimings test gehört ein CPU test~
Wenn nicht 60-90min, dann wenigstens 30min.
 
Zum Memtimings test gehört ein CPU test~
Wenn nicht 60-90min, dann wenigstens 30min.
Das ist mir bewußt. Die Aussage von mir bezog sich rein auf die Ramstabilitätsseite. Ich habe Außer Karhu auch TM5 Extreme1 @anta777 eine Stunde laufen lassen, den Aida stabilitätstest eine Stunde , Cinebench R23 Throttlingtest 30 minuten ,. Damit es mir nicht langweilig wird hab ich ein Paar mal den Corona Benchmark laufen lassen und UL CPU Benchmark. Dann als Abschluß Gaming im CPU Limit und im GPU Limit. Letzteres hab ich mir angewöhnt da ich mehrfach hatte das alle stabilitätstest gesagt haben "läuft" und einige Games haben sich dennoch mit freezer verabschiedet.
Zwischendrin habe ich geschaut ob Coldboot bugs auftreten , habe bei neustarts das warmboottraining aktiviert um zu sehen ob es hierbei Probleme gibt. Ich denke das langt für nicht ganz zwei Tage an tests. Und bisher hate ich ... toitoi ... keine Probleme. Scheint eine CPU zu sein deren Controller nicht so nervös ist. Ich konnte auch 7800 und 8000 Booten, aber leide nur benchstable, was mir nicht langt.
 
I'm now sharing an experience about how sensitive the LGA1700 socket is.
I changed CPUs because I bought another one. At the moment I have also converted the old CPU, which has passed many tests.
Of course it is not stable, it is impossible to install it the same way twice. I don't have a torque screwdriver (Nm), I think I tighten it normally by hand.
Until now, no LGA socket has been so sensitive.
What was the error phenomenon? Y Cruncher was unstable. He didn't run for a single minute. The socket and bottom of the CPU are clean, no dirt. Because of the pipes, it is not easy to remove.
I tried for the first time with the IMC voltage, 1.35750V or 1.35650V instead of 1.35V (next step), error in 3rd step.
SA didn't help either.
Interestingly, increasing TX to 1.21-1.215V solved the problem immediately.
I won't take it out and put it back in as the new CPU will be installed in 1 week and then with it.
In any case, when I take it out, I get a different MCSP but a stable value. Now it is currently 74 76 76 76.
If I remember correctly, there were also 84. I don't know whether it's because of the high frequency or the bad contact. In fact, the LGA1700 is too sensitive to pressure. Bad, I've never had that before.
Just tested it for an hour and it's working fine again.
Hi @zebra_hun, how are you? I had worked very hard since Veii gave me some hits. I will put some results in a next post :)

How did it go with the new CPU? What is your VF Factory Fuse? My biggest problem is on the CPU side due to my lack of knowledge with Intel tuning. I put more focus on mem, but the CPU side has a lot of relevance in stability. I still use your 54x configuration to protect the CPU. Maybe we have similar curves.

Thank you!
PS: I'm going to install an old BIOS to see if I can read the CTLs, I think with 0080 I could read them. At least to see how he trains 8000 and 8200.
 
@AndreasP1981 ich weis nicht ob dir das @Veii schon gesagt hat aber es gibt auch sowieso wie ein "SA loch"

das heißt das es sein kann das 1.25v geht -> 1.2v nicht geht - aber zb 1.1245v wieder geht , deshalb musst du weiter testen und nicht bei 1.2v aufhören nur weil das nicht funktioniert.
 
Zuletzt bearbeitet:
Hi @zebra_hun, how are you? I had worked very hard since Veii gave me some hits. I will put some results in a next post :)

(y)

Hello!

Thank you everything is good. I hope you too.
I like the CPU, the hope attached to it has been fulfilled.
Old cpu needs 1.09V, new cpu needs 1.05V for 54x.
One thing was not built, the direct die.
At the last moment, I lost interest in dd.
The old proven version remains. Now I'm waiting for the cpu to be ready. (Delid/Relid)
I really don't like what I wrote above, that every time I take it out and rebuild it, it's always different.
It is also possible that this will remain in this machine and I will put the new cpu in my ddr4 machine instead of the 13700k.
I also like ddr4, I have that too.

CPU stability is important, but don't just identify and stabilize the 54x, but also the 55, 56 and 57x. It's good to know what the exact, stable Vmin is for each frequency.
Your cpu is better, so your Vmin value will be lower.

Can you show a picture of Y running with Hwinfo?
I would check how much is required Vcore for 54x.
My guess is ~1.04V.
 
Zuletzt bearbeitet:
@AndreasP1981 ich weis nicht ob dir das @Veii schon gesagt hat aber es gibt auch sowieso wie ein "SA loch"

das heißt das es sein kann das 1.25v geht -> 1.2v nicht geht - aber zb 1.1245v wieder geht , deshalb musst du weiter testen und nicht bei 1.2v aufhören nur weil das nicht funktioniert.
neue Info für mich, lässt sich da eingrenzen wie "groß" dieses loch ausfällt?
 
neue Info für mich, lässt sich da eingrenzen wie "groß" dieses loch ausfällt?

naja nicht wirklich , es gibt setups da gibts sowas garnicht , dann gibts wieder setups da ist es so das die sa auf x.xx10v genau sein muss.

ich würde mal sagen versuch es zwischen ~1.1v -> 1.28v , das ist so das meist verwendete

und wenn du eh gerade im bios bist stell mal das ein (das undervoltet dein chipsatz etwas) :

  1. PCH 1.05 ->> 1.000v
  2. PCH 0.82 ->> 0.800v




Beitrag automatisch zusammengeführt:

es gibt sowas übrigends auch bei vdd/vddq - bei zu hohen werten kann der ram "zu machen" und auch bei zu hoher imc spannung kann es sein das der speichercontroller zu macht und garnix mehr geht.
 
Wie weit kann ich bei der IMC spannung ungefährlich gehen?

Denkt ihr /habt Ihr ne Idee wie ich das hier stabil bekomme?

Sa zur Zeit 1,2V , VDD 1,45V , VDDQ 1,39V , IMCVDD 1,4V, CPU VDD , 1,37V

1.png2.png3.png4.png
5.png

Das ist/wäre doch mal ne Herrausvorderung :ROFLMAO:
 
(y)

Hello!

Thank you everything is good. I hope you too.
I like the CPU, the hope attached to it has been fulfilled.
Old cpu needs 1.09V, new cpu needs 1.05V for 54x.
One thing was not built, the direct die.
At the last moment, I lost interest in dd.
The old proven version remains. Now I'm waiting for the cpu to be ready. (Delid/Relid)
I really don't like what I wrote above, that every time I take it out and rebuild it, it's always different.
It is also possible that this will remain in this machine and I will put the new cpu in my ddr4 machine instead of the 13700k.
I also like ddr4, I have that too.

CPU stability is important, but don't just identify and stabilize the 54x, but also the 55, 56 and 57x. It's good to know what the exact, stable Vmin is for each frequency.
Your cpu is better, so your Vmin value will be lower.

Can you show a picture of Y running with Hwinfo?
I would check how much is required Vcore for 54x.
My guess is ~1.04V.
Hi @zebra_hun, I’m doing fine. Thank you!
I had sent you in my last post #1,114 ycruncher while it was running at 54x and then many tests with the CPU at stock values, CB23, 15, 24

Do they work for you or do you need other ones? Do you have a screnshot of the new VF to see if they look alike? more than anything because it is not easy for me to put together the VF well. The reference points are very helpful to me.

Why aren't you going back to DD? What problems did you find? I was just thinking about making a deli and putting some DD...

Do you do the delid/relid, or do you send it to someone? thank you!😊
 
es gibt sowas übrigends auch bei vdd/vddq - bei zu hohen werten kann der ram "zu machen" und auch bei zu hoher imc spannung kann es sein das der speichercontroller zu macht und garnix mehr geht.
das hab ich schon letzte Nacht festgestellt, über ~1,42 mag der Speichercontroller nicht... vdd/vddq hab ich bisher nix gemerkt, aber denke in den berreichen beweg ich mich nicht...
Vielen dank für die Tips
 
You will notice that I have a 60x boost on all cores. That's because I opened the Asus OCTool before. For some reason it must go core by core. I am using your same BIOS settings to limit the cores, 90C, Current to 420A, 300/320W and SVID Trained. Y-Cruncher runs very well, at least I don't have thermal throttling.
Sorry, I wasn't careful enough.
Indeed 54x, only the Vcore is high. You can move it to your heart's content with AC LL as long as it is stable. Me is 0.25

I don't have a picture of it at the moment, I'm at work. Night shift.
This cpu is 40mV better.
Yeah, I don't do it. I leave it to a routine expert who does a lot every day. You can laugh, I'm afraid of LM. That's also why I didn't finish DD.
I put it together with paste, but the surface of the chip is very small, and I got horribly bad results in terms of temperature.
LM is certainly excellent, but it will never happen for me.
 
das hab ich schon letzte Nacht festgestellt, über ~1,42 mag der Speichercontroller nicht... vdd/vddq hab ich bisher nix gemerkt, aber denke in den berreichen beweg ich mich nicht...
Vielen dank für die Tips

vdd/q kann ab 1.45v probleme machen , aber bei deinem speicher (mdie der 2. gen) ist das eher unwarscheinlich , es gibt ja auch ram der gelockt ist ... (ka ob du das wusstest)
 
vdd/q kann ab 1.45v probleme machen , aber bei deinem speicher (mdie der 2. gen) ist das eher unwarscheinlich , es gibt ja auch ram der gelockt ist ... (ka ob du das wusstest)
das der PMIC locked sein kann wusste ich (y)
 
kommt drauf an ich würde nicht über 1.45v gehen. (24/7), aber meist läuft es eh schon eher nicht mehr (~1.425v)
Infinite ~ 1.62-1.65v max
// bis ~1.42 problemlos.
Ist keine IMC Spannung.
MC_Link (CPU ausgehend) Data Rail

Ich hab es gerade mal ausgetestet. Die IMC_VDD kann maximal (auf meinem Board) auf 1,41V gesetzt werden. CPU_VDD geht auf maximal 1,5V.
 
Ich hab es gerade mal ausgetestet. Die IMC_VDD kann maximal (auf meinem Board) auf 1,41V gesetzt werden. CPU_VDD geht auf maximal 1,5V.
IMC_VDD2
Gesetzt = Boardpartner erlaubt
Gerannt = startet nicht

Keines der beiden bedeutet (das Limit)

MC-Link VDD2 (Data-Link) IMC_Voltage
ASRock benennt CPU_VDD & CPU_VDDQ komisch.
Data & Data Strobe - Rail
Weiterhin IMC_Voltage
1711665172205.png

IMC Voltage~
Nun auch nicht ganz.
Ist automatisiert.

CPU ↓
CPU Data Rail (VDD)
CPU Data Strobe Rail (VDDQ)
↓ 60 Traces ↓
Board resistance
Slot resistance
↕↕↕↕ sync ↕↕↕↕
UDIMM
Mem Data Rail (VDD)
Mem Data Strobe Rail (VDDQ)
↑ SWA, SWB, SWC, SWD ↑
PMIC
↑ 5V (24Pin Supply) ↑

Data & Data Strobe Rail sync~
Ampere Sync
 
Zuletzt bearbeitet:
ich hab windows halt gerade erst neu gemacht und ich nenne es mal "custom view" der wheas noch nicht eingerichtet um abzukürzen

das es nicht toll/produktiv ist jedes mal dahin manuell zu navigieren ist mir bewusst :)
 
das es nicht toll/produktiv ist jedes mal dahin manuell zu navigieren ist mir bewusst :)
Eher dass es das nicht hier logt. Nur das, sorry.
Du brauchst den View von beiden und die Erlaubnis "Verbose" zu loggen.


@zebra_hun I haven't been answering you for a bit.
There is something going on :-)
Some days ago, i got another victim of the same APEX issue that you have, but much worse.
The Channel B (2nd towards 24 pin) does have design issues, and some traces or through filtering, do leak charge.
Hence by this the Training algorithm completely derps out in anything above 5000MT/s. 5600MT/s barely boots on auto.
It does boot with my legacy GROUPs+RONs (lucky hit) and it does remember the settings after the board got some charge. Then it stays and trains fine.

I'm currently waiting and watching how things develope with ASUS.
Given Consumer Law.
The thing you do not notice is, that 5V Input completely stutters towards the memory side.
If you use RGB dimms you would see them constantly flickering.

Soo because of a Hardware issue (non socket related) , i may take myself a look if it can be fixed.
But its unclear and without a promise.
Mostly definitely your two Boards are guaranteed defective and we'll see from now.
Like thats without a question.

Just thought i'd mention this, before you fixate on "global behavior" testet by your Board.
I may have seen 3 of such now , by the early researcher too. But yes;
It may be batch related (well surely is), hence i wanted to ask if you can give me all sticker pictures that you see and the description of your APEX Box (that sticker too)
One sticker is near the 24pin, one is above the PCI slot, and there should be 1-2 more. But those two are plenty.

In PM please !
If ASUS doesnt want to cooperate with the consumer, i'll get his Board to measure, and may would like to borrow yours too
So with 3 White APEX our team can inspect and measure traces better.
We will see, and i will not say more ~ given i wish first that person gets a repair they deserve.
Not me running after companies hiding completely defective batches. I'm already disliked enough;
The thing you do not notice is, that 5V Input completely stutters towards the memory side.
I think i can say its without danger to send 5V to the CPU. It shouldn't do that.
But both slots are passively affected by this and one is completely derping out - because training algo fails due to bad noise.
It may not be without danger to the UDIMMs tho.
Beitrag automatisch zusammengeführt:

Mostly definitely your two Boards are guaranteed defective and we'll see from now.
Like thats without a question.
Its a wonder you got it stable at 8000MT/s :)
But yes, lets see how things develop and what guy decides to do.
Wish i could just "buy off" both boards, because thats very valuable content (learning from the mistakes).
But lets see.

Looking at the time and "situation"
I would very much love it to be repaired, and we'll move on from this.
Alternative option is borrowing somebody as complain person. Roman or Steve.
I've been annoying the team enough so far, but lets see what happens;

Any case, maybe attempt to sell yours at a normal price, while it functions decently-ok.
His worked for couple days and just stopped.
 
Zuletzt bearbeitet:
I don't have an RGB kit, I only have this one.
What I can say about it is that 6400MHz is the last frequency that auto RTT/ODTs can boot.
I emphasize that this is a Klevv kit that I have, it is possible that it is faulty. I don't rule it out.
It is certain that starting with DDR5 is not the best for me. I also managed to buy the worst cpu in the world.

After a weak start, a strong decline :d

Anyway, we already have the secrets and tricks of how it works well. I think Asus is proud of their wonderful product, just as Intel is proud of the SP93.
I know you don't like the undervolt and frequency reduction, but everything remains dynamic. I just downgraded everything as much as possible.

Of course, I will send photos in PM if I will be near a PC in the evening.

Thx

Edit:
It's matter? 13187 13190


Edit 2:
Es kann user error auch sein. Seit november habe ich ddr5. Schon zufrieden 8000c36 46 46 wenn es stabil bleibt. Schaut schon so aus. Kompliziert, aber geht's.
 
Zuletzt bearbeitet:
Its a hardware defect
DIMMs are fully fine.

But i need to measure and compare.
Could be within PCB layers, but chance is high to be an/couple SMD.
Slots are fine, socket was fine.
It leaks charge & 5V_IN→PMIC stutters.
 
Zuletzt bearbeitet:
Hardwareluxx setzt keine externen Werbe- und Tracking-Cookies ein. Auf unserer Webseite finden Sie nur noch Cookies nach berechtigtem Interesse (Art. 6 Abs. 1 Satz 1 lit. f DSGVO) oder eigene funktionelle Cookies. Durch die Nutzung unserer Webseite erklären Sie sich damit einverstanden, dass wir diese Cookies setzen. Mehr Informationen und Möglichkeiten zur Einstellung unserer Cookies finden Sie in unserer Datenschutzerklärung.


Zurück
Oben Unten refresh