[Sammelthread] Intel DDR5 RAM OC Thread

[acoustic]

Also had one since Release. Booting 8800 wasnt any problem.
If u cant boot 8533 its another Problem, Not the board🤣

not every board is the same, that’s the problem with poor QC. The later production batches tightened quality up a lot more.

 

[tibcsi0407]

By pulling an individual stick? That’s what I meant in that post — with the sticks on water, it’s too time consuming with the small amount I have to spend on this stuff these days.

If there’s an easier way to do it.. that could be possible, like disabling the entire channel via BIOS, but I haven’t looked into that.

You can disable the channels, like I did.

Beitrag automatisch zusammengeführt: 20.02.2024

Also had one since Release. Booting 8800 wasnt any problem.
If u cant boot 8533 its another Problem, Not the board🤣

Believe me, I tried with 3 CPU, and 3 kit.
The same can do 8600 in my Encore.
That board was B-Ware from Alternate. Probably that's why.

 

[snakeeyes]

Broken one. One slot Not ok, got also one from Amazon.

But a normal apex can boot 8600 ez. Got 3 apex all first batch and all boot ez 8600+

 

[tibcsi0407]

Broken one. One slot Not ok, got also one from Amazon.

But a normal apex can boot 8600 ez. Got 3 apex all first batch and all boot ez 8600+

I guess I was unlucky. But 8266C34 was pretty fast. 😊

 

[amd_man_bavarian]

Bisher scheint das 1901 Bios bei mir zu funktionieren. Beleuchtungsänderungen interessieren mich eher nicht, da alles deaktiviert ist da ich auf das leucht Klimbim keinen Bock mehr habe.

SA 1,191V , VddqTx 1,35V , IMCVdd 1,35V , VDD 1,45V , VDDQ 1,43V

Leider hab ich nur einen testshot behalten bei den Stabilitätstests .Warscheinlich haben die vibrationen des 140er lüfters @ 1600U/min der den Ram gekühlt hat auch zu sehr mein Hirn Vibrieren lassen so das ich die Konzentration verloren habe






8200 C36 kann ich zwar benchen und läuft auch kurz im stabilitätstest aber das ist alles andere als stabil. ca. jeder 6. Boot ist damit ein Fail.
Ich denke für den I5 sind die 8000 C36 ein super ergebnis und das ganze stabil mit im Alltag passiver Kühlung.

Hat jemand mal die Teamgroup T-Force Xtreem 8000er getestet? Hab hier nicht alles mitgelesen.
Gemeint sind diese :

TeamGroup T-Force XTREEM Narvik Black DIMM Kit 48GB, DDR5-8000 ab € 299,36 (2024) | Preisvergleich Geizhals Deutschland

✔ Preisvergleich für TeamGroup T-Force XTREEM Narvik Black DIMM Kit 48GB, DDR5-8000 ✔ Produktinfo ⇒ Typ: DDR5 DIMM 288-Pin, on-die ECC • Übertragung: 8000MT/s (8000MHz effektiv - abgeleitet von 4000MHz I/O… ✔ Speicher ✔ Testberichte ✔ Günstig kaufen

geizhals.de

Wenn jemand die hier getestet hat die Frage : sind die Passiv stabil betreibbar?
Ich will nichts mit RGB und auch keine software installieren müssen um sie zu deaktivieren,daher meine Frage.

 

[snakeeyes]

Die 14600k können schon was ^^


Sry für die Bildqualität, aber war nur auf die schnelle mal eben abgeknippst.

 

[amd_man_bavarian]

Nice ! So gut ist meiner nicht.
Meiner ist dagegen ne Gurke und nur gut für Salat ^^

 

[CarSalesman]

By pulling an individual stick? That’s what I meant in that post — with the sticks on water, it’s too time consuming with the small amount I have to spend on this stuff these days.

If there’s an easier way to do it.. that could be possible, like disabling the entire channel via BIOS, but I haven’t looked into that.

You can leave the sticks in and disable MC1 or MC0 in BIOS.

I don't know if it is causing your issue, but tCKE should be 21 for 8267. I can't see what tXP is set to but it should be at 31. Your tCPDED is correct at 21.

 

[acoustic]

I will try disabling a channel and testing that way if I can't find something stable by tomorrow. tCKE is set to Auto right now, just letting the board train it how it wants. tXP is 26 apparently; I will try changing that to 31.

None of my changes today have resulted in anything better than what I did last night; if anything it's made it worse. I reloaded the 10hr/3error profile, and raised DIMM VDDQ 10mv. I am now at a 40mv in-mem delta of 1.50v VDD/1.46v VDDQ. We will see if this helps.

Edit— it does not help. Error 0 @ 16min, lol..

 

[y668]

Some progress with eventual DRAM voltages since the Vdroop (errors within 10 seconds of test) was killing my runs.
1.16 SA
1.51 VDD Mem (Training -100mV, MSI only has presets in 50mV increments)
1.41 VDDQ Mem
1.27 VDDQ CPU
1.39 VDD2

Bandwidth looks healthy now (still waiting on heat sinks), air cooled.
No idea if I'm at the optimal VDDQ deltas yet.

 

[CarSalesman]

I will try disabling a channel and testing that way if I can't find something stable by tomorrow. tCKE is set to Auto right now, just letting the board train it how it wants. tXP is 26 apparently; I will try changing that to 31.

None of my changes today have resulted in anything better than what I did last night; if anything it's made it worse. I reloaded the 10hr/3error profile, and raised DIMM VDDQ 10mv. I am now at a 40mv in-mem delta of 1.50v VDD/1.46v VDDQ. We will see if this helps.

Edit—it doesn't help. Error 0 @ 16min, lol..

The boards trained timings are wrong quite a bit. I would imagine the tXP being too low has more of a negative impact than tCKE being too high, but who knows?

 

[acoustic]

The boards train timings wrong quite a bit. I would imagine the tXP being too low has more of a negative impact than tCKE being too high, but who knows?

It is the strangest thing. @tibcsi0407 @Veii

Around ~1hr, Error 0 popped up. I am busy and didn't have time to reboot and try new settings, so continued to let it run. It's now at 2hr 45min, and no further errors, just that single Error 0. Just doesn't seem to be any rhyme or reason to it. On this run, I raised VDDQ TX to 1.235v from 1.230v..

Last time I seen this kind of behavior, it was the Unify-X limiting me, causing wacky behavior that was unsolvable. I'd be a bit surprised to be stuck at 8200, but I guess it's not impossible. My 14900K is an average sample.

 

[perlibu]

Die 14600k können schon was ^^
Anhang anzeigen 973228

Sry für die Bildqualität, aber war nur auf die schnelle mal eben abgeknippst.

Tm5, karhu und yc wäre interessant.

 

[Veii]

It is the strangest thing. @tibcsi0407 @Veii

Around ~1hr, Error 0 popped up. I am busy and didn't have time to reboot and try new settings, so continued to let it run. It's now at 2hr 45min, and no further errors, just that single Error 0. Just doesn't seem to be any rhyme or reason to it. On this run, I raised VDDQ TX to 1.235v from 1.230v..

Very likely VDDQ_CPU issue.
Self refresh or clock halt issue.
One timing is not fine and VREF(DQ) is not fine

+/- 5mV change, is the correct approach for now.

Beitrag automatisch zusammengeführt: 21.02.2024

Any word on VDDQ deltas for 48GBx2 greens to start off with?
I’m unable to train beyond 7600 C36. Tested 1.41-1.5 Mem VDD and they all boot. SA around 1.16 (tried 1.05-1.2, they all boot) gives me the least training inconsistency. I used the 60mV VDDQ delta rule, but this might not be the correct application for dual ranks.

Inside mem, voltage is ok to stay 60mV apart ~ higher the better, depends on how good powering is.
IVR VDDQ & MVDDQ
VDDQ_CPU & VDDQ_MEM

That part depends on Boardtuning and bios.
VDDQ Training off to force you figure this out.

Test till minimum stable (without #0) and maximum stable (without #0)
Use the middlepoint of both , or use the edge (max delta & work on ODT, to get it there)

VREF exact middlepoint, is a DDR3 topic (@unifyx)
DDR4 VREF was minimum half or bigger of VDIMM ~ some people said half or lower, but thats DDR3
DDR5 VREF is slightly shifted upwards

^ DDR4 onwards ~ ([DDR5] Pseudo Open Drain example)

DDR4 & 5 ~ POD illustration

Differential signals are symmetrical, but single-ended signals are not. Rise & Fall Time differ.

---

Maybe it can be easier calculated
If you sample minimum delta and max delta
Then use like 75-80% of the value as "optimal level"
I guess, its near 75-100mV delta for 48gb dimms. Probably min 50mV max 150mV. Test and report back
Test in 30mV steps, make a cold-boot (psu off) between VDDQ_CPU to VDDQ_MEM changes.
15mV steps for fine granular, 30mV steps for raw.
! It will not boot if value is off by 15mV. It will not be stable if value's are off by 5mV !
// unless board keeps adjusting~
// Clock near irrelevant. For 8000MT/s stable system, sample at 7800. For 7600MT/s stable system, sample at 7200.

I may need some rework on CKE and clock-halt. Especially rework on CK_L & CK_H.
But i also think i remain right for CKE & clock-halt.
It definitely should not be 7.5ns but smaller.
While XP stays at 7.5ns & CPDED stays at 5ns.

 

[acoustic]

@Veii

Thanks. I will try adjusting VDDQ TX downwards. I've tried higher values and they don't seem to help, or they make it worse. I didn't expect below 1.23v to be viable, but at this point, it's worth trying I suppose.

 

[Veii]

@Veii

Thanks. I will try adjusting VDDQ TX downwards. I've tried higher values and they don't seem to help, or they make it worse. I didn't expect below 1.23v to be viable, but at this point, it's worth trying I suppose.

Use https://www.hardwareluxx.de/communi...-ram-oc-thread.1306827/page-442#post-30268285 as orientation example ~ for PoC
Mind you, each capacity has different VDDQ to VDDQ delta.
Same as 1 DPC differs from 2 DPC.

Removing dimm sticks, changes slot impedance.
For isolated testing, disable a full channel via bios.
Or a full MC-Link for cpu sided testing.

Ok yeah i was pretty sure 24/4 was correct which in the end i did have to swap it back to that to get my memory stable again so maybe it just brute forced its way past it in the first rounds of 1usmus? otherwise no clue but careful on 0904, the ME firmware def changed something for me and @tibcsi0407 - idk fully for him but i had to change a couple v/f points around, the main one being v/f #11 needed to go down or up 0.003 can't remember now, even reverting back to 0080 i had to change my v/fs to what was needed on 0904 again, prob will swap back to 0904 again tonight to test some mem stuff although I'm still on air so only so far I can go

0081 is a good bios, exception is randomness on what feels like MR10 or RTT ~ one of both. Against that randomness "final" 1801 can be used.
New Bios has retuning for 256GB sets ~ soo very likely it was played again with VREF and RTTs.
Very much suspect RTTs, due to errors i can read last 1-2 pages.

ME update is permanent ~ tho doesn't override ME FW between EEPROMs. Only one EEPROM is accessed at a time.
If CPU curve needs remake, so be it
Reverting the Bios will not do much, except increase instability on the CPU side, with exchange of older voltage scaling on Mem Side.

If one has updated, might as well deal with it~
Its unlikely that the team pushes two updates with negative changes, and also unlikely that Intel-HQ pushes bad ME.
If change was needed ~ its ok to adapt * . Might as well stay on the new bios and figure new delta's out.

Like, it wont get reverted. Any ME shipping was pretested for weeks or a month.
Mostly/usually a month, before there is a chance to be supplied to customer as Beta.
It being updated has no chance of revert. Unless you swap eeprom's.
Downgrading bioses will make the chance of (CPU) issues bigger, as ME cant be reverted.

EDIT:
* @tibcsi0407 if curve got messed up , just increase AC_LL higher and problem is resolved.
Its that easy, hence we use a mix of both. So we can scale if whole substrate for memOC needs more voltage or not.
Much easier to linearly scale that way. Cleaner too.

 

[Mtorrent]

Ok so did some quick tests before I gotta leave again, messed with some v/f points - think I got the hang of it for the most part (hopefully), though I did remove the 6400 for now cause I figured it would be easier to tune/learn the v/f stuff with just normal tvb for now, maybe i'll go back for the 6400 eventually but really wasn't needed

here is my cpus v/f and new updated curve after messing with some stuff
Anhang anzeigen 965413
Anhang anzeigen 965508

, Goo

Hi @Vhypur, I hope you are doing well. I'd like to ask you if you can share me your V/F Offset. We have a very similar V/F factory curve, in my case I have a 14900k SP105, but the V/F are pretty similar. Veii suggested me to use your values. Thanks

 

[y668]

Maybe it can be easier calculated
If you sample minimum delta and max delta
Then use like 75-80% of the value as "optimal level"
I guess, its near 75-100mV delta for 48gb dimms. Probably min 50mV max 150mV. Test and report back
Test in 30mV steps, make a cold-boot (psu off) .

I tend to drain residual power (power button with PSU off) before a cold boot between adjustments. Is that sufficient or do I have to cut power for a minimum time?

 

[Veii]

I tend to drain residual power (power button with PSU off) before a cold boot between adjustments. Is that sufficient or do I have to cut power for a minimum time?

A psu off (switch) is plenty
The PSU cuts power cleanly
Vampire current is a residue effect of capacitors ~ reverse leakage some might call. I forgot the exact description for it.
Its also part of the decision why a Pseudo Open Drain loop is used, or why one has filtering on VSS (Ground, memory)

EDIT:
Well vampire current itself is idle powerdraw.
I didnt phrase it well. But i mean its not gonna leak over if (switch off)
Vampire current will happen between powerplug and psu. Same as example above
But if switch is off, it is off. Thats what i mean

Beitrag automatisch zusammengeführt: 21.02.2024

verstehs nicht.

@perlibu
Wie ist deine CPU montiert.
Alles liest sich nach einem Socket-Pressure Problem.
Von meiner Sicht aus

Frame ja, nein ?
Welcher Wasserblock ~ Federschrauben/Abstandshalter?

 

[Mtorrent]

Hello, Good timezone~
So so
You ?

Ok i see.
In such case you can just copy it from both either tibcsi or zebra_hun.
And slightly reshape it because mid-low differs too much between CPU binns.

That Blue bios looks soo strange.
Unusual

30mV step on P3 is huge

Hi @Veii! I'm also so so, but still pushing. Long story short, there was once a very happy OCN member with a good setup for his 13700k and 7200 with a Hero . But well, the desire to improve components and performance made me change everything. I bought an Encore, I managed to reach 8266MT/s with false stability, since by mistake, like many of us, I limited the voltage, in my case with an AVX offset, that's why YC never failed. Then I bought some 8000 2x16 Gskill A-die memories, to lower the voltages and temperatures a little, and until then, everything was fine. Then I bought the 14900k, which to my surprise and because it was a first attemp of purchase on Amazon US, I got a good one, SP105.

From there all my headaches started. First, the memory OC failed, and all the configuration problems it had came to light. I had set CPU L2 at 1.2v, Ring 48/50, PLLs SA, MC and VTT. I realized from the progress of @tibcsi0407 what had advanced and one of the things that surprised me and you mentioned to me in this last post, that my SA was very high, and I didn't understand why so many users used such a low SA with an Apex. Meanwhile, I have two childs, they played Fortnite or Roblox, every so often an error 41 would appear and it would crash or reboot. From then on, I realized how many errors 41 I had. I ended up uninstalling my 4090 that was underclocked with the MSI Afterburner, then deactivated the interrupt priorities with the MSI tool, and little by little I returned to zero. I forgot to mention, I also changed my PSU, given the Kernel Power 41 BSOD problems, I thought that my 1000W Seasonic Titanium did not support the spikes, and I bought a Corsair Gold RM1200 Shift. I'm not very convinced if it's really good, now when I turn off the PC, it turns off completely, before with the Seasonic I had the Mobo light on for a few seconds, as if the capacitors held more charge. It will be better? Same with this Corsair 1200, I still have Error 41, but for me for other reasons, more on the CPU side than the memory. There is something in my VRM or CPU Voltage settings that is not right.

Spoiler: Nightmare Error41

My VF curve that I showed you in my previous post did not seem so stable, I think starting at 700mV is too low, I left that 15mV that you suggested. But in P3 I need to go up to at least +25 so it's not so flat. From then on it seems to me that the curve is well "rounded", but I don't know if 1320v for 6000 is not enough. I see the other users reach 6200 or 6400 in the end. Should I apply a positive offset to the rest? This would be my curve and now I'll tell you my nightmare with the OC of the memories.

What do you think? My goal is undervolt and stability above all things, whether in the CPU or in the memories, I am not an Overclocker, I like to see what they do, but I prefer your definition of a researcher more, I think we learn more, take longer, but it's real. Everything else is to show a number and is unreliable, my POV

Regarding the memories, I started from scratch. What did I do? Bios almost by default, activate XMP 1 @8000, adjust it a little on this basis, and disable VDDQ Training, with VDDQ at 1.45v which is the achieved another false stability, which was not stable to retraining

Spoiler: #1 8000XMP a little tight, Delta VDDQ-VDDQCPU-Low SA

After reading several posts here on the forum, I made these changes. I removed all the PLLs, also the CPU Input Voltage, the CPU L2 that I used, ring in Auto, TVB in Auto (I don't know if it's right), Multicore Enhancement in "Enable Remove All limits" and SVID Behavior in Trained. Cstates by defaul in auto. In the mems, TWR in 30 and WRPDEN in Auto. Delete reference CTL0 and RTTs, memory trainings all in Auto except VDDQ Training disabled
This configuration at 8000 was not stable, and I passed a little YC (maybe that's the error) and then TM5, then Karhu 9hs... and cold boot... I went to 8266 taking one of your recommendations from a previous post

Spoiler: 8266 Crash after 10th cycle

What I did was, I lowered VDDQCPU to 1.22 and SA rised to 1.21 I realized that at 1.18 it failed quickly, perhaps TX at 1.22 is too low for VDDQ 1.45. I know SA can't get above TX, that's why they are so close. Perhaps my mistake is modifying SA by 10mV, it should be 5mV. MC was set to 1.46. I think the error must be in my SA and MC configuration, if VDDQ_Mem is a fixed variable, and the delta is between 180 and 230mV, it would be at the maximum of that delta, I could raise it, but there are three variables. Stability in retraining is a big issue. I always save a configuration and clear CMOS, and load it again. I don't know if I'm close, at least I see that I can have a good delta without debinding the RONs, nor the RTTS, nor CTL0... with a low SA, the opposite of what I had before. It is very rare that TM5 gives me an error. They are almost always voltage problems, which Karhu can sometimes catch, but not always. YC is the key.

What do you recommend me? I'm a little lost and frustrated...

Thanks like always!

 

[tibcsi0407]

* @@tibcsi0407 if curve got messed up , just increase AC_LL higher and problem is resolved.
It's that easy, hence we use a mix of both. So we can scale if whole substrate for memOC needs more voltage or not.
Much easier to linearly scale that way. Cleaner too.

The Vmin was the same on new BIOS, but something was messed.
I tried hours to find what the hell happened but I couldn't stabilize the 8533 there.
I will try it later too to find out what did they change, but atm I had to come back to 0081. It works like charm.
Actually ME version is the same afaik, so it's not a problem to bounce back without BIOS switching. CPU Microcode is updated in new BIOS.
Maybe they are preparing for KS launch.

TBH I didn't want to buy KS, but now I am thinking on it. A little bit afraid that I would get worse MC.

Beitrag automatisch zusammengeführt: 21.02.2024

My VF curve that I showed you in my previous post did not seem so stable, I think starting at 700mV is too low, I left that 15mV that you suggested. But in P3 I need to go up to at least +25 so it's not so flat. From then on it seems to me that the curve is well "rounded", but I don't know if 1320v for 6000 is not enough. I see the other users reach 6200 or 6400 in the end. Should I apply a positive offset to the rest? This would be my curve and now I'll tell you my nightmare with the OC of the memories.

You can try my current curve. I have AC_LL 0.65, so you can just lower that to reach your Vmin.
You want to see ~flat line on the bottom and a double S shape after. For the bottom I like to set 730mV to have a little headroom. With lower Voltages I have stability issues.

P 2x60 5x59 7x58 +2 TVB E 8X46 16X45 R 0-50X

Core VID max is hugely differ from Vcore max, it's because my Pcore0 requests for higher Voltage for some reason. I have no idea why, that's not even getting the highest ratio.

 

[perlibu]

@perlibu
Wie ist deine CPU montiert.
Alles liest sich nach einem Socket-Pressure Problem.
Von meiner Sicht aus

Frame ja, nein ?
Welcher Wasserblock ~ Federschrauben/Abstandshalter?

heatkiller, kein Frame. Alles normal. Eingespannt, Kühler mit Federschrauben, u-Scheiben. Hatte nie Probleme bei dem.
es ist komplett Random. Einmal so, ander mal so. Dann mal ein 55er.
Hab jetzt noch einzelne timings bisschen verstellt. Kommen manchmal verschiedene Errors. Nach einem Neustart wieder andere Errors. Wie soll man da wissen bzw vorgehen.. Spannungen erhöht, gesenkt, kein Erfolg.
Habe bemerkt, dass der frei verfügte Memory im tm5 sehr niedrig ist. Bei 1050mb. Ist das nicht zu wenig?Wie erhöhe ich den?

Edit: gester lief tm5 26min, screen gemacht, direkt danach freeze. Heute gleiches Setting, nach 30 min 300 errors. 🙄
Edit2: 40 min, dann freeze. Nochmal gleiches Setting, 8 min 20 errors und freeze ihne was zu ändern.

 

[Holzmann]

Hat das 1001 schon jmd. getestet?

ROG MAXIMUS Z790 APEX ENCORE | Gaming motherboards｜ROG - Republic of Gamers｜ROG Global

The ROG Maximus Z790 Apex Encore supercharges builds with 24+0+2 power stages, hyperspeed DDR5 memory slots, latest PCIe 5.0, and Intel® Wi-Fi 7

rog.asus.com

 

[Felix the Cat]

Falls jemand noch ein weißes Apex sucht - ich hätte da eines in taktfreudige Hände abzugeben.

 

[X909]

Hhm... vielleicht hat jemand tips. Hab die CPU getuscht. Alt: 13900KS, 1.488V 6 GHz VID - der hat 8400 easy mit 1.195V VCCSA, 1.375V VDDQTX, 1.45V VDDQ und 1.56VDD gemacht. ycruncher und retrain stable.
Jetzt ist ein neuer 13900KS drin, 1.422V 6 GHz VID und auf den P Cores auch taktfreudiger (e Cores und Cache etwas schwächer als er alte). Mein Speichersetting läuft nun überhaupt nicht mehr, was ich soweit sehen kann, ich brauche mehr VCCSA.

In welche Richtung würdet ihr gehen zum finden der Spannungen?

 

[Digg]

Hhm... vielleicht hat jemand tips. Hab die CPU getuscht. Alt: 13900KS, 1.488V 6 GHz VID - der hat 8400 easy mit 1.195V VCCSA, 1.375V VDDQTX, 1.45V VDDQ und 1.56VDD gemacht. ycruncher und retrain stable.
Jetzt ist ein neuer 13900KS drin, 1.422V 6 GHz VID und auf den P Cores auch taktfreudiger (e Cores und Cache etwas schwächer als er alte). Mein Speichersetting läuft nun überhaupt nicht mehr, was ich soweit sehen kann, ich brauche mehr VCCSA.

In welche Richtung würdet ihr gehen zum finden der Spannungen?

Da wirst du nicht drum herum kommen, alles neu auszuloten.. also SA und CPU VDD/VDDQ. Womöglich war der IMC vom alten KS einfach gutmütiger?
Ich habe das Gefühl, dass mein 14600kf beim Mem Spannungstoleranter ist als der 13900kf. Sprich beim 13900kf müssen (obenrum) die Spannungen genau passen während der 14600kf mit verschiedensten Spannungen obenrum läuft.

 

[Mtorrent]

You can try my current curve. I have AC_LL 0.65, so you can just lower that to reach your Vmin.
You want to see ~flat line on the bottom and a double S shape after. For the bottom I like to set 730mV to have a little headroom. With lower Voltages I have stability issues.

P 2x60 5x59 7x58 +2 TVB E 8X46 16X45 R 0-50X

Hi @tibcsi0407, how are you? I couln't change my nickname here, I'm lmfodor@OCN.
You had sent me a BIOS config file with 8400, but the curve was not defined at that time. Now I see how you get to 6200MHZ, with 60x2 and TVB +2. Have you set the curve with or without IA_AC/DC?

For now I'm using the standard V/F curve trying to stabilize the memory at 8266 or 8200 or even 8000. I still cannot find a stable configuration of delta VDDQ - TX and SA. In each scenario I managed to achieve some stability, but not retrain stable. I'm still using 0080 BIOS.

The truth is I'm lost, the only progress was achieving a good delta with RTT and RON in Auto, as I had asked you about a while ago on OCN. I thought that the BIOS couldn't train a large delta without defining a RON and RTTs. I used to use RON 40-40 or 40-34/34-40 that Veii had suggested to me, and RTT 40-34-34-34-240-0-0-60-40-40 40-40. I think I'm going to put together a spreadsheet with the values for each scenario, to see where I'm going wrong. What I could do would be to load the XMP profile I, and try to optimize the CPU. It would be 8000 at 1.45V

Regarding IA_AC/DC_LL, did you define it after setting the curve? I read that Veii recommended first setting the curve and then moving it with AC.

For now, I'm on 8266 with VDD 1.52/ VDDQ 1.45 / MC 1.46 and SA 1.2v. (The minimum I managed to run SA was 1.15v). RX in Auto, CTLs, RTT, RON in Auto. Even though I can't run CB15 Extreme or CB24, it gives me WHEA19, but before I had BSOD with Service Exeption or Clock Watchdog Timeout errors. This happens to me for buying some A-Die, I should have bought some 8000Mdie and I would have a better pattern to compare since most have 2x24 or 2x16mdie with lower voltages.

About your curve, I didn't understand the double S, I see that the first graph looks pretty good, with a 45 degree angle between 3400 and ~5200. Then in the second graph I see that double S, why should it have that shape?

Can you share the screen of the OCTool Freq / target die Sense? That way I can see for each frequency what the target die sense is, and when applying it I automatically calculate the Offset. So I copy it and try it without OC from memory. I think it would be an interesting first test after so much chaos. . Thank you very much for your help!

Beitrag automatisch zusammengeführt: 21.02.2024

Core VID max is hugely differ from Vcore max, it's because my Pcore0 requests for higher Voltage for some reason. I have no idea why, that's not even getting the highest ratio.

Do you still have Multicore Enhancement set to Auto? Or have you selected Enable - Remove All limits? This changes quite a bit how the SVID behaves, in fact if I remember correctly it set PL1 to 253. Another thing I noticed is that if I disable Cstates, the processor is limited to 5700, otherwise, it boosts up to 6000.

Here I found your old BIOS configuration, from what I see in your HWInfo screenshot you achieve 6200 but a max effective clock of 5701, do you still use this configuration or did you upload 8x58 as you told me?
I also found that you had TVB Voltage optimization disabled. Should I disable them?

Cache Dynamic OC Switcher [Auto]
TVB Voltage Optimizations [Disabled]
Enhanced TVB [Disabled]
TVB Overclocking [+2Boost Profile]
Overclocking TVB Global Temperature Offset Sign [+]
Overclocking TVB Global Temperature Offset Value [5]

Maybe since I am limited by temperature with an AIO360, I should only enable TVB + 2 over 57x which is the maximum configuration that ASUS gives me if I remove all the limits. But for that I would like to have the curve right. Because if I run SFT, the CPU Package goes to 330W and I can't hold the 95 degrees, in fact it is the most difficult test because it always crashes, sometimes with a BSOD...

 

[X909]

Da wirst du nicht drum herum kommen, alles neu auszuloten..

Ei das wird lustig... VDDQTX kann offensichtlich viel niedriger mit dem neuen. Bin jetzt bei 1.275, also 100mV weniger als im alten Setup obwohl VCCSA 55mV höher ist. Mal sehen wie weit ich runter komme, vielleicht will der Chip einfach mehr Delta und ich kann SA dann auch wieder senken.

 

[Digg]

Für mich ist DDR5 Mem OC hauptsächlich wegen Benchmarks interessant. Und da auch eher die leichten wie SuperPI.
Ich kann mir vorstellen, ein daily Setting mit einem i9 und 8000+ hinzubekommen, ist nicht ohne. Wenn da alles an P/E-Cores aktiv ist + HT, ist das anscheinend viel härter für den IMC. Da hängt es wohl vom IMC ab, wie einfach es geht, oder nicht.

Bis ich gestern mal den i9 auf dem Strix zumindest mit 7800 durch den Y-Cruncher 2.5B bekommen hatte, verging eine Weile. Ein 32m Lauf mit 8000 dagegen war schnell erledigt.
Temps sind auch nicht so besonders... ist halt nur ne 60 Euro 280mm AiO.

 

[X909]

Aufm APEX sind 8000 kinderspiel, jedenfalls bei meinem Setup hier. Es kommt einfach der Punkt, ab dem die nächste Taktstufe ein Krampf wird und der kann bei 8.000, 8.200 oder 8.400 sein. Je nach Speicher, Board und CPU. Wie immer halt beim OC - die letzten 2% brauchen erhebliches Feintuning.