Du verwendest einen veralteten Browser. Es ist möglich, dass diese oder andere Websites nicht korrekt angezeigt werden. Du solltest ein Upgrade durchführen oder einen alternativen Browser verwenden.
Mmmm
Es ist schwer für mich hier mitzureden, da ich auch dem APEX Bios nicht vertraue.
Man muss selber Hand anlegen, damit Sachen laufen. Es ist traurig, aber so war meine gesamte Intel Erfahrung bis jetzt.
Von der Sicht eines Normalnutzers, ja ich kann es verstehen.
Ich finde es weiterhin peinlich dass Intels V/F Curve Points, bzw man beginnt schon bei CEP auf B760 ~ fast allen Boards fehlt, obwohl es Intel features sind.
Dass die Slopes Auswahl fast überall fehlt, oder eine Auslese der aktuellen antrainierten ODT/RTTs.
Es ist sehr peinlich.
Das ganze läuft solider bei AMDs ecosystem, für die Kunden.
Einfach da sich fast alle Mainboards nahezu identisch benehmen, da AMD ein gewissen Zwang und Druck ausübt.
// EDIT: Fast alle 6 Layer 2DPC cappen auf 7600MT/s.
// Die Oberklasse auf 7800MT/s. Und die 1 DPC zwischen 7800-8800MT/s. Straps zwischen 8200-8667 sind kapput.
// Sie benehmen sich alle nahezu Identisch da nicht nur die PCB design-targets erzwungen werden, sondern auch das Bios bzw die Biospredictions.
Das Naheliegendste an einem Apex/Extreme (bzw [Dark]Hero) ~ wäre womöglich das Tachyon X & als 2DPC dann das Aorus Pro X.
ASRock hat die Chance, aber die beiden Boards fehlen einfach im westlichen Markt. Das Tachi ist ok, alles eine einstell-Sache.
EVGA hatte die Chance, aber ja ~ was soll man machen.
MSI bietet soweit garnichts, nur das Z690 UnifyX.
Man kann es ASUS nicht übel nehmen.
Sie sind riesengroß. Mit sehr vielen angestellten Talenten.
I think you have a mounting issue & bad pins contact.
When those dropout.
Alternatively it may be VDD2, but i don't like your MC test results.
I think its overpressure.
Bei dem Tmings habe ich mich durchgearbeitet und 7400 laufen gut. Bekomme aber die 7600 nicht stable. Von daher habe ich, durch viel lesen hier, einige Spannungs-Regeln für mich zusammengestellt.
Ich denke das meine Spannungen noch "krum" sind.
VCCSA =CPUSA = System Agent Voltage über 1,250v (CPU) Maximal 1,35 V für die 12. und 13. Generation, da eine Spannung von über 1,40 V schnell zu einer Verschlechterung führt.
VDDQ TX = IVR Transmitter VDDQ Voltage
VDDQ_CPU (Speicherkontroller) gleich oder größer SA.
VDDQ_CPU & VDDQ_MEM Gehen hand in hand. Delta von 170 mV (75-200 mV) , wobei MEM höher ist
Höhere SA verrinert das Delta, Gerinere SA verkleinert das Delta
VDD_MEM gleich oder höher von VDD_CPU (MC) (VDD2_CPU)
VDD_MEM darf bis zu 100mV über VDDQ_MEM sein. Minimum 60mV laufen immer. Memory Spannung in 15mV Schritte
Bei schlechten Main Board kleines Delta und SA hoöher.
VCCSA =CPUSA = System Agent Voltage über 1,250v (CPU) Maximal 1,35 V für die 12. und 13. Generation, da eine Spannung von über 1,40 V schnell zu einer Verschlechterung führt.
SA bleibt Standalone
Höhere MemCapacity, Anzahl oder schlechteres Board ~ weniger VDDQ ⇔ VDDQ Delta
EDIT ~ V/F Curve:
14900K:
1.38-1.42 non leaky = niedrige Spannung
1.425-1.44 normal sample
1.45-1.47 leaky sample = mehr Spannung
1.48-1.51 very leaky sample = viel mehr Spannung
Dank@Veii.
So sieht es grad aus. Bin auch ganz ok damit, aber es sind 7800er GSkill und da wollte ich wenigstens 7600 schaffen. Auch mit einem 4 RAM SLOT Mainboard.
CPU ist auch ok. Board limitiert wohl.
#14 Trace signal dropout - CPU side.
"Dropout of CLK/CMD/ADR ~ DQ's"
Hmmmmm
Your memory sensors completely firetruck'd up.
Only V-SA will mess with ODT and so with mem.
Try on 1.17 SA & instead normal VDD2, try effective VDD @ 1.48.
Mmm, I don't like it
Not consistent enough.
Beitrag automatisch zusammengeführt:
I think you have a mounting issue & bad pins contact.
When those dropout.
Alternatively it may be VDD2, but I don't like your MC test results.
I think its overpressure.
Sorry, ich habe oft nochmal das Forum durchlese, jetzt schaue ich, daß du mir geschrieben.
Ich sage dir erlich, bischen Pause. Ich bin total zu frieden mit 8000c36.
Für Spaß, natürlich mache ich es weiter, aber wie jetzt ist, immer in der Arbeit...
Zu wenig Leute, Firma will immer, daß ich reinkomme... WE auch.
Mein Plan ist, daß den Kühler zurückbaue, aber wann... Es ist nicht 10 Minute.
Jetzt habe ich auf cpu Seite ein spezial Curve gemacht für Alltags, es ist downclocked.
Sehr stark - offset on Curve, so konnte ich ac ll höher stellen.
Bitte nicht lachen, es ist ac54xP (zwei Kerne bis 57x), 43x E, und 47x R.
Ich denke, es ist super eingestellt, weil für meinen 165Hz Monitor habe ich stabil 0.1% fps.
Ich brauche kein ac57 oder noch höhere Frequenz. Video - 162fps cap
Das reicht mir, und zu frieden.
Aber wenn wir reden, ich habe eine komische Frage.
Absolut verstehe ich nicht.
Ich kann Y Cruncher stable machen. Aber das ist nicht TM5 stable.
Wenn ich TM5 stable mache, das ist nicht Y stable.
Unterschied ist groß.
Y wird stable, wenn ich extreme wenig Spannungen einstelle.
Ich möchte schon mein alte 10°C Keller Kühler zurück, und dann weiter.
Ich nutze jetzt 8000c36 TM5 u. HCI stable Profil, aber es geht ca. 15-20 min Y.
Nicht so schön so ein Profil, die kann kein Y stable, aber kein eizieges mal Problem hatte.
Alles geht schön mit niedrigem Spannungen auf 800-5400MHz allcore.
Wenn man die letzten Stufen nicht nutzt (58x 60x 60x 60x), Curve komisch, aber ich habe etwas gebastelt
Gehts gut meine ich.
Beitrag automatisch zusammengeführt:
Edit/:
High Voltage
Mid Voltage
Low Voltage
High and mid passt TM5, aber Low Voltage passt nur Y Cruncher. Über eine Stunden.
Etwas noch da zu.
High u. Mid mit fix rtts, low voltages profil habe ich alles auto. Odt u rtts.
Tx, vddq, u bischen mc ist unterschied.
Sorry für das späteres Bilder.
Handy habe ich alles.
Supercool using the original ILM
With the ILM you secure the cold plate and then you screw the block housing on the top of the cold plate.
With this you get the original contact pressure what Intel designed.
Most users on OCN has this block, like Sugi and many others.
It's a little bit messy right now, since the loop is still in test phase.
Wenn du mich fragst, natürlich beides. Aber bei mir, TM5 25 Rund bringt sehr gute stabilität.
Y manchmal 15 min und vt3 error, manchmal kann laufen 40+ minutes.
Mit diesem Low Voltages Profil, gehts immer 70+ minutes.
Aber Spannungen so niedrig, mein Testspiel cp77 benchmark crash...
Das ich verstehe nicht. Jede hat Probleme mit Y Cruncher. Aber wenn ich so niedrig stelle, lauft es lang.
Ich sage es sehr lautlos: TM5 25 Rund, HCI 1.5 Hour, Aida Kombi test (Cache, Mem, CPU, FPU) 1 Hour, das ist schon nutzbar.
Beim NASA kann sein, das ist nicht genug stable, aber für einen gamer mit seiner 60-90W Power, reicht.
Lautlos
Supercool using the original ILM
With the ILM you secure the cold plate and then you screw the block housing on the top of the cold plate.
With this you get the original contact pressure what Intel designed.
Most users on OCN has this block, like Sugi and many others.
It's a little bit messy right now, since the loop is still in test phase.
Ah , ive heard to not use QDC's on the Iceman, due to the weight on the ram slots.
hmmm
i imagined it like the Mycron, directly with pads onto the board, and just backplate clamped.
I mean you'll figure it out, but its currently not in a good state
Ah , ive heard to not use QDC's on the Iceman, due to the weight on the ram slots.
hmmm
i imagined it like the Mycron, directly with pads onto the board, and just backplate clamped.
I mean you'll figure it out, but its currently not in a good state
Nope. I can still revert to Iceman, but I would be more happy with this particular block.
Not easy. Maybe I can even try this block with a contact frame. ☺️
Nope. I can still revert to Iceman, but I would be more happy with this particular block.
Not easy. Maybe I can even try this block with a contact frame. ☺️
That's its goal We don't run forced clock to focus on applications We run enhanced DVFS or how Intel™ calls it now. Enhanced Boosting within their parameters. Have you re'run MC-SP test? Different conditions on SP training , may need a re'run or at very least a CMOS + Profile re'load. In case the CPU wasnt detected as swapped. In case you were almost enough. May you also please show the current Curve. What very likely happens, is that you boost higher due to TVB and peak states not being ok What may also happen is that hidden limiters and hidden clock boosts higher. Its all a margin topic at the end. CPU is not stupid. I would need to go over your last 4-5 screenshots and compare VIDs to see a change, but it very likely is just a V/F x Thermal curve variable. Mem by itself doesn't change discharge properties as such "little" changes. Neither the crystal is that thermal sensitive. If we talked about 40° to 0°C , then maybe - but barely and rarely. I don't believe Do you have some sort of error logs, anything i can work with ? What ME are you on right now? Update there would have messed with curve. 5x MC SP test, checking pressure Then current curve visual example Then we'll see afterwards, optimally "errors" that appear. There have been now enough reports that 1002 base for 9901/2 is alright ~ G4 base. But also that its different and ppl struggle with peak clock stabilizing. We might start here, with sanitychecks above~ Unless there is a new bios again. EDIT: Nice new V-SA. Anhang anzeigen 979439
Hey man, does mc SP mean anything in general? I have mine at 88-89, does this mean anything regarding voltages, SA or imc? What is your experience with this parameter.
Hey man, does mc SP mean anything in general? I have mine at 88-89, does this mean anything regarding voltages, SA or imc? What is your experience with this parameter.
The parameter has to factor in CPU leakage factor to be anywhere near reliable
I can not talk about Dev's secrets in how it works, but if it should work accurately it needs to test with different voltages ~ matching cpu leakage factor
Because the only fused data to go by that is exposed (i think) is V/F curve points
Training for SP is done in bootable factors.
Per MC and altogether.
Thermals will not really influence the result unless its lets say -50°C or lower.
Because of the sole duration it tests & crystal being quite thermally stable.
A high value has high potential to boot XYZ, but non of those mean it will be stable on it.
A high memOC may be happening and the value will reflect that, but it will not reflect that anything inside the CPU is keeping up with it.
It will not change based on ICCMAX user settings and has nothing to do with cache or FCLK limitations of the sample.
At the end,
It is a value based on X testet voltage, boot test.
Now what the enduser needs to set to reach this boot test target, will vary.
At the very best it can be used as a consistency check. If whatever value stays the same within the same bios between mounts.
Normal CPU degradation is unlikely to change MC SP, because if, then first it could show adjustments on V/F points.
MC SP has nothing to do with cores either, hence its an isolated variable.
So if one looks on it that way, then the value can be any value.
You would compare your sample vs your sample.
Cross comparison between samples is complicated, because they alone will not have the same V/F curve and also not the same leakage factors.
Where leakage factor differences only start to matter on both extreme ends. Low/Hi thermals, low/hi clock, low/hi voltage requirements.
Neither of those being an isolated factor and non of those dictating "whats better". Its usecase specific.
Hence SP between non similar V/F curve samples, should not be compared.
Like between different SKUs should not be compared.
Is it useless, no
Will it help you reach that clock it tests at, also no
Does it make sense to compare a 1.4 V/F 14900K vs a 1.48v V/F 14900K @ same MC SP , absolutely no.
// EDIT: If tool tests well, it may be weighted and match quality and even match clock, but usable targets to reach this clock between both samples. Will very a lot.
// Target voltage and strain factor will be strongly different between those two, even if "potentially" they can reach the same cookie-sp state
// To throw in another bone, it will not say that as made-up example MC-SP 88 equals 8800/8800 MT/s. It could have been 9200/8400 too. Or 9600/8000. Between MC-Links.
If you factor in all that, it is usable to speed up checks of "is mount good", "is mount consistent" & with plenty of personal data "is sample vs sample @ nearly identical V/F better or not"
Better or not due to whatever reason. It can show stars, it can be cookies, it can be anything.
The value is only comparable within similar situations & should be testet multiple times, given its not a fused but floating value.
It is not a leaderboard value, because leaderboards on leakage-factors aka ASIC% ~ just can not function.
// EDIT: And because it again is not a readout value but a floating testet value. It has variables;
It can only function as leaderboard value @ target sample VID & same mount or Mainboard @ target bios.
All those are variables that will mess with it. Thermal is unlikely to be one unless we talk again about extreme sides of the testing (-50° vs +50° ambient).
Beitrag automatisch zusammengeführt:
Soo if i would be a co-designer in this,
I very much would show two values instead of one. (MC 88/88)
To not only assist mounting but show link quality between both MC-Links or at least channels.
It would very much speed up time to know what is usable on DDR4 and Gear4 XOC.
What the community makes out of this values then is up to them. It would be more usable data tho;
Hey man, does mc SP mean anything in general? I have mine at 88-89, does this mean anything regarding voltages, SA or imc? What is your experience with this parameter.
P.S. It’s mostly SA, hence I haven’t found it to be very reliable in regard to 24/7 or even bench stable. The SA connection is the reason why many SA bugged 14th Gen CPUs have a high MC SP.
Die meisten Treiber bekomme ich mittlerweile über Windows Update.
Und im schlimmsten Fall installiert man Armoury Crate, installiert die Treiber und kickt es wieder.
Zero Day awaiting
Haha.
Its so so, not very happy to partner "ASUSClientLicense" telemetry with everything (memtweakIT included)
But as long as you are given a choice
The idea will fail tho, because you need ethernet drivers to get ethernet support. Local installs.
Especially on 3rd party "Killer/Aquantia" situation. And from the perspective of normal non tech-savy users.
This goes to the enterprise/workstation market too , proArt and similar markets who run their systems through Hypervisors and Firewalls.
They'll shoot themselves in the foot with that, just smile and watch~
Zero Day awaiting
Haha.
It's so so, not very happy to partner "ASUSClientLicense" telemetry with everything (memtweakIT included)
But as long as you are given a choice
The idea will fail tho, because you need ethernet drivers to get ethernet support. Local installs.
Especially on 3rd party "Killer/Aquantia" situation. And from the perspective of normal non-tech-savvy users.
This goes to the enterprise/workstation market too, proArt and similar markets who run their systems through Hypervisors and Firewalls.
They'll shoot themselves in the foot with that, just smile and watch~
Zero reason
Its just telemetry dll.
Installs automatically on all boards, when installing their package.
Tools (exe) have their own DLLs,
Installer just pushes bloat services with it, but thats been the common case here.
Sitting back and watching~~
Time will fix.
Zero reasons
It's just telemetry dll.
Installs automatically on all boards, when installing their package.
Tools (exe) have their own DLLs,
Installer just pushes bloat services with it, but thats been the common case here.
Sitting back and watching~~
Time wants to be fixed.
