[Sammelthread] Intel DDR5 RAM OC Thread

Can you improve the IMC by grinding the "Die"? Does the Mem then go up, or is it just the wrong value in the BIOS?
Wouldn't it be easier to just buy a better CPU?
I can't play the lottery, I buy the parts as a companny, can't just send it back if I don't like it. ☺️
It's not a bad CPU at all, max stable is 8533, same as my KF.

I would buy prebinned CPU, but no one can give me invoice from that. 🤣
 
Wenn Du diese Anzeige nicht sehen willst, registriere Dich und/oder logge Dich ein.
Es gibt noch eine Option. Kaufen, testen und auf dem Gebrauchtmarkt verkaufen.

Nachdem ich ein paar Tage im OCN unterwegs war, frage ich mich, wieso ihr euch nicht dort austauscht. Das, was ihr hier in englisch ausbreitet, ist dort den ganzen Tag Thema, habe ich das Gefühl. Diese ganzen Loadline/Köpfen/Kühler Geschichten haben doch recht wenig mit Mem OC zu tun. Passt doch eher in ein Wakü Thread oder OCN Intel OC Thread.
 
Zuletzt bearbeitet:
considering the material removed from the die.
I can't imagine I removed even 100 microns yet. Maybe 50.
Got confused :-)
Indium should have been between 250-350 microns (0.25-0.35mm)
Well ~200 + silicon feets/ring of IHS.

Its a far too small range to measure unless you invest in mitutoyo measuring gear with a ruby ball (not worth for your task)
I wonder if those acrylic lapping supports still exist for LGA1700 chips. They had fine steps on them.
Interesting.
Well a diamond-crystal cant expand for sure.
Just to understand fully, this is:
~ a single crystaline synthetic grown crystal, like they use for LEDs
~ a copper ? sheet just with a diamond coating deposited on it
~ some other base material that you inspected to have good Z-Axis Wm/K rating, and hopefully low thermal transfer resistance.
Which is far more than it needs.

Which option shall i use in OCTool to link you the VF curve?
I'm sorry, but i didnt write the tool. I dont remember its dropdown position.
By the time you waited for the answer - you could look up how to use it and find the pages
There is no tutorial for it written out.
The curve parts probably sit near the Intel dropdown menu.
Take a look and you will find them. Explore the tool

The "tutorial" of it may lie back ~150 to now ~180 pages back
or if you track mine, zebra's , tibcsi's posts you will also find pictures.
I don't know :-)

Repeat it like couple posts ago how to use it.
Play with it, its straight forward & explore this thread.
You may find answers for other duplicate questions you have~
so took my 8 bar compressor and airgun and cleaned every socket with air pressure. And it booted. :)
But now I have stability issues again, so started to troubleshoot. Seriously, it really annoys me now, but it's my fault, pushing too hard.
The Blocks, i dont know
The pushing hard in perfection or lowering voltages, or buying stuff you dont need
~ although later part who knows might rule out some things for you
Non of those would be a waste.
Making an accident, oh well :) stuff happens.
It was already polished when I got it. Maybe too thin?
Eh, at worst not flat but tilted. Things you can measure, as waterblocks/aircoolers generally dont tilt easily.
But besides that, i dont think so.
Too thin would crack or make the lithography visible. Hope i used the right word
Basically would clearly show the inner life, haha.
There is space above it, but too low and things crack too easy.
I have some 1000 grit sandpaper here, I wonder should I check the die with a liner and try to polish it. What do you think? Maybe that would help with the MC SP fluctuation as you suggested before.
1000 grit is too little for crystal polishing. You'll just scratch it deeply
1000-1200, even 1600 is for metal polishing. Copper for example or nickel.
2000 may be for steel bars or other material you want shiny , like kettles. EDIT: (not all kettles are soft brass)

2000 should be the bare minimum for the crystal.
// XOCer like bisobiso for GPUs for example use flat knife polishing stones for this type of work. Else a Marble/Glas surface and high grit plus water droplets is the way to go (of course leveling required)
Sure you can start with 1000, but why even scratch it more than needed :)
Would only use that if i didnt know how to remove indium better.

I can believe MC SP calculations (changes) may happen by a colder state (0°C vs 20°C roomtemp)
I can not believe it has any sort of duration (the testing) for thermals to matter, soo contact thermal transfer wouldnt be such a big issue
Only uneven pressure would be. But too much pressure on socket also is not too forgiving in the long term. Those are springs after all.
Yet also, thats a far too low chance, given all are factored in when manufacturing as usage and wear tollerances.
 
Zuletzt bearbeitet:
The Blocks, i dont know
The pushing hard in perfection or lowering voltages, or buying stuff you dont need
~ although later part who knows might rule out some things for you
Non of those would be a waste.
Making an accident, oh well :) stuff happens
Now cleaned the dimm slots again and seems better. Got always error 2 in TM5, seems like it's a good test to find hardware seating issues too. Voltage stability was not good at the golden fingers. Y didn't run a cycle at all. Now it's running 30 minutes ago.
Also made a patented anti bending solution for memories
Anhang anzeigen IMG_3536.jpeg
I will buy the Der8auer block for sure. I am also sure it won't be any better in temps, but maybe in IMC stability it could help.
2000 should be the bare minimum for the crystal.
then I will get a 2000 grit paper, maybe I have it here. On OCN they used 1000 grit I believe.
I can believe MC SP calculations (changes) may happen by a colder state (0°C vs 20°C roomtemp)
Safedisk told me the same, but temp is between 16-18C, I meen the coolant.
 
Zuletzt bearbeitet:
then I will get a 2000 grit paper, maybe I have it here. On OCN they used 1000 grit I believe.
Grit papers make only sense if you have a marble or glas surface
Else they make no sense.
1000+ ones mostly, unless exceptions exist ~ are titled wet sand papers

i still would recommend the knife polishing stone (flat not curved) for direct die
But yes ~2000 should work

Low grit logically is for metallic materials
here we are polishing crystals and emeralds :)
We dont like to have a coarse grain

Be sure to get Fujifilm prescale papers (LLW and LLLW) sample packs
You can try to lap by eyes and feeling, but still need something to check your work
Careful with the SMDs ~ but that goes without saying. :-)
Safedisk told me the same, but temp is between 16-18C, I meen the coolant.
Its just if you compare user setups (which is hard) or if you compare your own binning
That room temp slightly might influence the first value, and of course the bios version
But nothing else really. Its a too short test.
 
Zuletzt bearbeitet:
Mit leih Kit hab ich angefangen. 8400 wollte ich, aber mein Mobo will das nicht leider. Jetzt C38, spaeter versuche ich es C36 erreichen.
 

Anhänge

  • 8200C38-49-25-49-69-704.png
    8200C38-49-25-49-69-704.png
    316,7 KB · Aufrufe: 53
snaphsot0004.png


Im MP zum Verkauf. :bigok:
 
Super_PI_32M_250.863.jpg


Das habe ich bisher hinbekommen mit den 1.62v. Ich hab kein Plan, ob 24GB M-Die so mit Spannung skalieren, wie es A-Die machen. Werde es demnächst irgendwann mal rausfinden. ;)
 
Interesting.
Well a diamond-crystal cant expand for sure.
Just to understand fully, this is:
~ a single crystaline synthetic grown crystal, like they use for LEDs
~ a copper ? sheet just with a diamond coating deposited on it
~ some other base material that you inspected to have good Z-Axis Wm/K rating, and hopefully low thermal transfer resistance.
No metal; just manufactured diamond. The manufacturer advertised them for electronics heatsink applications. It has some visual artifacts/imperfections but is almost transparent.
 
DDR5 Intel@anta777 Config ist brutal. Manchmal über 8 Watt.
50°C unter Stress. Raumtemp ist 26.9°C.
Extreme Sommer hier.
Ich hab gestern auch mal den neuen Anta777 drüber geprügelt. Wasser Temp. 32 Grad. auf den RAM 35,5. Es prügelt ebenfalls die CPU auf 230W zum Teil...ist schon Saftig.
 
Es ist jetzt Kit von einen Freund. Leider kann/will ich nicht so einen sau teuer kaufen. 8400c40.
Wundert mich, gehts wie Traum. Musste ich garnichts mit RTTs, ODTs spielen.
Alles auto. Es geht einmal frei.
Nur Datein von Veii's Excel eingegeben, Spannungen ausgesucht, 8200c36 48 68 672 mit extreme Temperature durchgegangen.
Neue TM5 50°C. Es ist nur mit einen 140mm Ventillator gekühlt. 1600rpm :d vollgas.
Ich habe kein Lust sie zurückgeben.
Leider Y crasht später, wegen weiniger Vcore.
24/7 ich nutze weniger Spannung wie Y stable.
1.08V drop ist knapp für vst/vt3. Egal, Kit ist sehr gut...
V1 Apex kann handle stable 8200. 8400 unmöglich denke ich. Natürlich Y stable.
 

Anhänge

  • 8200-C36-48-24-48-68-672.png
    8200-C36-48-24-48-68-672.png
    295 KB · Aufrufe: 1.058
  • 20240520_195502.jpg
    20240520_195502.jpg
    1,6 MB · Aufrufe: 581
DDR5 Intel@anta777 Config ist brutal. Manchmal über 8 Watt.
Discharge config * vs consistency and error correction check ~ config
HCI/Karhu sind nicht besonders anders zu der bekannten discharge config.
Für das schwache TM5, nehme ich 1usmus's config. Für thermal stability, Karhu/HCI/Dangwang oder umso besser GSAT/Stressapptest
Thermal stability der CPU , dann natürlich y-cruncher sowie andere Render-Applications bzw welche mit dynamischer Last (Spiele).

* Einer der Gründe weswegen man mit mehreren Tools gegentesten sollte.
Eines für jeden Job.

Zebra_hun, bitte verwende die 0.13 Version nicht, bis klar wurde ob diese sicher ist.
Die Änderungen in dieser sind nicht mehr open-source !
Die geteilte 0.12.3 ist gut genug und wurde für ~3-4 Monate von mir gegengetestet, bevor diese öffentllich wurde.

SerJ der RUS Developer von TM5 , wollte nie und möchte nicht seinen Code verkaufen.
Das Projekt bleibt privat. Weswegen man sich entschied die Änderungen nicht mehr mitzuteilen und closed-source zu gehen?
Ich vertraue 0.13 vorerst nicht.
 
Zuletzt bearbeitet:
Btw, @Veii , I stabilized 8400c34 with RxDfe disabled, albeit with fixed voltage. It was really, really difficult. Full RTTs/RONs, Read Equalization +14 & +7, all slopes still AUTO.

Now I'm trying to get it stable with adaptive voltage.
 
Read Equalization +14 & +7
Hmmmmm, its far away of what i'd do
Thats a strong Skew.
Maybe Encore is very different or slopes are strange for it to be soo shifted.
But would not use them bellow 8800.
They are clock specific too.
Beitrag automatisch zusammengeführt:

Now I'm trying to get it stable with adaptive voltage.
You know what i hold of that topic :)

Intel spec https://www.hardwareluxx.de/communi...-ram-oc-thread.1306827/page-492#post-30388402
Our shenanigans are on 0.6 AC , given ASUS kept it at 0.5 for months on Auto LLC. Before it was 0.8 on Auto , and sometimes it was 1.1ohm

Push memOC , increase AC let IA supply more and keep own correction curve
Or remake curve on 1.1ohm to the lowest possible and then play with it till 1.7ohm ~ when you need bit more juice for memOC and Rings part.
Above 1.1ohm i believe it will hit ICCMAX close to always (math) & also keep hitting 1.6 VR MAX ~ soo i dont know :)

Thinking about both things at the same time, isnt a good idea.
You were at 8533 //EDIT: Yea i'm bad with names, it wasnt you, it was tibcsi
without active enablement of RX-DFE. At whatever Auto preset it was defined.

Messing with ReadEQ delay (duration and starting point before and till reflections)
Basically messing with reflections pulse-size ~ is not something i would recommend at all.
Its far too early for this. 8400-8600 doesnt need Receiver side DFE presets unless DIMM IC & PCB is bad
(your RTTs work to fix powering, not use advanced filtering techniques ib a good board to polish a tu*d)

8600 no active RX-DFE preset.
You know the drill :d
Has 8533 gotten too hard with your CPU ?
// EDIT: Nvm , but yes nothing bellow 8400 for this. Need RX-DFE for later, not now.

Fixed slopes comes before EQ skewing.
RTT & ODT comes before slopes
Beitrag automatisch zusammengeführt:

@Darkthrone it was around there
+/- some pages, talk with tibcsi about the curve . Jump around :)
Main main introduction was before that, pages of talk with zebra_hun.
Somewhere near the page 390-400 should be a huge writeup , dropdown menu's with linked posts of the past (categorized)

@CarSalesman
RX EQ +1/-11 :)
+2 / -15 or -8 :)
- location near my last public OCN post, when i shared the Biosmod, there you have some info
But dont use it now. Pulse distance and wideness is something you tune muuuch later.
Slopes starting point and ODT for it, matter faar too much. Its a floating distance

There are more things around DFE and EQ, which you can not change.
ASUS higherup's don't like fun.

Best you can try is Start at +1 to +2 , Stop as floating Training-enabled value ~ but its tooo early for that
Please dont build anything on a skewed foundation. Same topic as to not build anything on a PLL setup.
Basically messing with reflections pulse-size ~ is not something i would recommend at all.
Its far too early for this
Doing so , will cause a situation of "mem stability is achieved, Internally it breaks y-cruncher cant pass"
Or the other side, with good skew ~ "y-cruncher keeps stability but mem slopes get messed up and mem is impossible to be stable"

Dont skew with this value. Faar too early :giggle:
Also if we get too far with those Start/Stop presets, ASUS might take away this option too. . .
Would cause a power imbalance, you know~~ 🤭
 
Zuletzt bearbeitet:
Hmmmmm, its far away of what i'd do
Thats a strong Skew.
Maybe Encore is very different or slopes are strange for it to be soo shifted.
But would not use them bellow 8800.
They are clock specific too.
Beitrag automatisch zusammengeführt:


You know what i hold of that topic :)

Intel spec https://www.hardwareluxx.de/communi...-ram-oc-thread.1306827/page-492#post-30388402
Our shenanigans are on 0.6 AC , given ASUS kept it at 0.5 for months on Auto LLC. Before it was 0.8 on Auto , and sometimes it was 1.1ohm

Push memOC , increase AC let IA supply more and keep own correction curve
Or remake curve on 1.1ohm to the lowest possible and then play with it till 1.7ohm ~ when you need bit more juice for memOC and Rings part.
Above 1.1ohm i believe it will hit ICCMAX close to always (math) & also keep hitting 1.6 VR MAX ~ soo i dont know :)

Thinking about both things at the same time, isnt a good idea.
You were at 8533 //EDIT: Yea i'm bad with names, it wasnt you, it was tibcsi
without active enablement of RX-DFE. At whatever Auto preset it was defined.

Messing with ReadEQ delay (duration and starting point before and till reflections)
Basically messing with reflections pulse-size ~ is not something i would recommend at all.
Its far too early for this. 8400-8600 doesnt need Receiver side DFE presets unless DIMM IC & PCB is bad
(your RTTs work to fix powering, not use advanced filtering techniques ib a good board to polish a tu*d)

8600 no active RX-DFE preset.
You know the drill :d
Has 8533 gotten too hard with your CPU ?
// EDIT: Nvm , but yes nothing bellow 8400 for this. Need RX-DFE for later, not now.

Fixed slopes comes before EQ skewing.
RTT & ODT comes before slopes
They took away debug menu, too!

Jerks!

I asked safedisk for BIOS with debug menu, lol he linked me a test BIOS without debug menu. BeCareful24 told me I am completely shooting in the dark without it and now I yearn for it.
 
Das Aorus Pro X ist draußen.
Wie ist das gemeint oder zu verstehen?

Soweit ich weiß, wurde das Board aber schon letzten Oktober oder so vorgestellt und haben tue ich es seit März. Muss nur das neu Bios mal aufspielen, aber das wollte ich verschieben, bis ich den anderen Speicher habe.
 
Wie ist das gemeint oder zu verstehen?

Soweit ich weiß, wurde das Board aber schon letzten Oktober oder so vorgestellt und haben tue ich es seit März. Muss nur das neu Bios mal aufspielen, aber das wollte ich verschieben, bis ich den anderen Speicher habe.
Das ich es ausgebaut habe und jetzt verkaufe.
 
They took away debug menu, too!

Jerks!

I asked safedisk for BIOS with debug menu, lol he linked me a test BIOS without debug menu. BeCareful24 told me I am completely shooting in the dark without it and now I yearn for it.
I am certain there are many people there who do their best.
Still love my OLED Zenbook. There are couple released products which show their craftmanship
But with a big company, there are many employee's who are a bit more rough.
Because company's Ideology appears to seek for the stars. This may or may not cause one to forget where they came from and forget caretake of community.
Its not the Bios/FW department here and its not the Intel department exclusively either.

Hard to say,
There is no need to defame when some internal people have a difficult perspective of their product.
But on a flagship Board with a community that pays its high price ~ one would think they keep their important (non common) tuning options public.

Unfortunately those findings were always there in the Bios since AlderLake. Probably a testing menu for their engineers
// which also was fully fine, given you actually need to read and understand what the options do, let alone their shorted name even means
Now that somebody made it public, it became a bad thing. But fine maybe intel wasnt happy (which makes little sense on this context) , soo higherup forced it to be gone
It was removed then. But it was not fully wiped away // out of public's eye, out of mind

Now later that we helped to access it with proprietary tools , that was also not good enough
Not because it would bother anybody, but this time it appears like some ego was hurt.
Soo its gone from every ASUS Board, while it was accessibly for years without issues.

One could still say "Intel doesnt want it" , but people employed decided to go out of their way to fully remove it and any trace of it.
Causing a situation of "what makes ASUS/ROG Bios even special at this point, when you have to rebuild it fully with own options"
Its unfortunate, but it is what it is :)

Soo when you can't differentiate their Bios from other Boardpartners except hidden tuning
And you are forced to create those options from scratch. Where is the difference between ASUS and any other Boardpartner :)
Yea i dont know :-)
Its fine. Own taken decision;
// for me this only speaks as
// "if i cant differentiate you, Why should i bother with your foundation and help marketing."
// At the end a Bios is everywhere the same and talks to one FW. It's on the Team & ODM to tune it.
// Soo unless i love the Hardware side ~ i don't know :-)"

Unfortunately there are no 1 DPC rival releases that match it on the engineering side.
Nothing cheaper or without own unhappy decisions taken on its design.
Beitrag automatisch zusammengeführt:

Yop, verstanden ;)
Danke das Infos.
https://www.overclock.net/threads/o...94772/page-953?post_id=29238198#post-29238198
DangWang 1.0
You target 2000% , but may be ok with 1000% pass.
Its different from 5.0 and newer

On the same level as Karhu 20K %
Overkill but needed for discharge tests (discharge, noise, thermal tests)

TM5 is fully ok for what it can do.
Things it doesnt do well have to be covered by other applications


Actually while we are on this mood
I can share finally some mod. But no debug menu. Just access to some more curve tuning options
Unsure about 14th KS support, although there is nothing really to support.

Cap VR MAX to 1600mV as always, to prevent CPU self killing itself due to buggy curve
Cap IA TDC near 300A max or near 260-270A range ~ see previous posts and YT comment.
Cap IA ICC MAX to 400A , at very worst 420, but 400A is a good value.

Have fun :-)
APEX Encore ~ base on 0018 // you can keep it on the other eeprom due to lower ME state (if you want to fool with CPU OC)
The Dev prefers to stay private;
https://wormhole.app/ddAq7#567gQkl9HCURQfK0BvoxXA should be for 3 digit downloads online, lets see
 
Zuletzt bearbeitet:
OCN vermisst euch.
 
Yop, verstanden ;)
Danke das Infos.

du solltest aufhören auf diesen ... zu hören ... nur mal so als tipp

die meisten leute die ahnung haben schreiben hier nichts mehr (zb. @Phoenix2000 , gibt auch noch viele andere) wegen genau ... . ... hält sich für etwas besseres , behauptet dinge die nicht stimmen, hat keinerlei einsehen wenn fehler begangen werden.

... denk mal drüber nach

- hat ja wohl ein grund wieso "alle" leute die bisher hier auf ... gehört haben kein stabiles & gutes ram oc haben.




... sagt es ist ein virus ... "100+" leute die es nutzen beweisen das gegenteil. das karhu addon hat auch "positive" virenmeldungen genauso wie alte tm5 versionen ...
 
Zuletzt bearbeitet:
Unfortunately there are no 1 DPC rival releases that match it on the engineering side.
Nothing cheaper or without own unhappy decisions taken on its design.
There is the little Asrock Z790i Lightning.
For next gen I will buy Asrock board I believe. I loved my X370 Taichi and now it seems they woke up from the 3 years of sleeping. :)
 
There is the little Asrock Z790i Lightning.
For next gen I will buy Asrock board I believe. I loved my X370 Taichi and now it seems they woke up from the 3 years of sleeping. :)
Waait, define woke up :d
Ahh ASRock~

In the X3/470 time Shimizu_OC [JP] was on their side (unsure about previous time)
Later he moved to MSI , probably near 500 series , and the last i heard about him was that he made? or joined an own Lab.
He was such a sweet & polite person (to me) and did well on the marketing side too (XOC a bit)

mm mm, during early Zen times ASRock did a lot having the most unlocked Bioses
And often not listening to AMDs orders (well publicly yes, but thats not the whole story) haha

On AM5 they became AMDs favorite , so to say. InHouse OCer Team of AMD and generally AMD picked them, to what i know;
Currently like on 500 series, they remain their same. "It works,, works great but is booring" ~ title.
It must be hard, because it isnt like the old days of the stories i'm told with/of Nick Shih and the custom Intel Boards.

Stories and old days 😌
But yes, its hard because they didnt have a good time with their Aqua series and the Z790 one never got public.
Looking forward to Core Ultra's one, but in the current state, their Taichi Carrara is lovely on AM5's side.
Beautiful board & good engineering. Just high EU price~

Its a high risk to make those OC Boards, and Splave doesnt seem to push much either recent months.
But maybe its a good time now, given people look for other options.
On the RMA & Tech support side, official or not so official ~ they always get an A+? S- , from me.
Well i was ignored here and there, but on the contrary also helped with danger of the person loosing their job due to an NDA topic.
// when i bricked my ITX Board twice, due to them having Bios stability trouble
Soo it evens out. People's readable part on "consumer support" so far they nail. (y)

EDIT:
To add, on AM5 currently
Their 118€ B650M-HDV ~ also nails 8000MT/s right now (stable). Some push ~8200.
They must be happy, because it doesnt need flashy RGB nor big marketing to make a great product.
I just feel on the bios tuning side it was a little slow, but yes they got their stuff now together ~ i feel.

EDIT2:
Sharing the same Bios Branch between X370 and X470 Taichi ~ due to being lazy and poor single? fw-dev
Us doing complete Frankenstein mods turning X370 to X470 , like permanent converts 😆
//
because AMD couldnt give us proper 3000/5000 support on the overkill X370 Taichi.
Or on my B550 PG-ITX, putting a Gigabyte or Biostar bios on it, because why not 🤭🤭
It was also a great time with them, mmmmm ~ i must be PR nightmare X'D
I mean AMD locked at the end everything down thanks to us, but still :geek: always fun times;
 
Zuletzt bearbeitet:
Welche Spannungen kommen denn beim IMC an?
SA - CPU intern
VDDQTX - CPU zu mainboard?
VDDQ - Memory zu mainboard?
MEM - CPU - mainboard - memory?

Oder wie kann man sich das vorstellen?

Ich frage mich, ob ne hohe VDD den IMC degraden kann.

PS: PLL Termination voltage von 1.05 auf 1.1 zu setzen verringert den VDD Bedarf. Das fasziniert mich... dachte VTT sei CPU intern und VDD memory intern - aber irgendwie spielen die zusammen.
 
Ich frage mich, ob ne hohe VDD den IMC degraden kann.
Technisch eigentlich unmöglich. Da die Spannung auf dem dimm selbst generiert wird.

Mit der PLL ist interessant. Die Frage ist wieviel VDD.
Ergibt für mich aber irgendwie keinen Sinn. Sollte ja eine cpu Spannung sein, wenn ich mich nicht ganz vertue und diese dient ja zu kappen des Signals. Aber da kann ich auch komplett daneben liegen 🤣.
 
Dachte ich bisher auch so... hab das mit der PLL Term ausm OCN und es senkt mir den VDD Bedarf tatsächlich um ca. 30mV. Kann dann ja nur mit einem saubereren Clock-Signal o.Ä. zutun haben.

Ich hatte bisher 1.63V VDD, jetzt 1.6. Alle anderen Spannungen sind unsportlich bei mir. 8600 ginge wohl auch aber mit arg Fummelei und viel VDD bzw. nicht mit C36, was es dann wieder unattraktiv macht.
 
Du kannst die CL38 auch flott machen. Einfach tWRRD_sg/dg und tCWL senken, bis es nimmer geht.
 
Welche Spannungen kommen denn beim IMC an?
SA - CPU intern
VDDQTX - CPU zu mainboard?
VDDQ - Memory zu mainboard?
MEM - CPU - mainboard - memory?

Oder wie kann man sich das vorstellen?
Voltage explanation sheet ~ LUXX Edition

"Der IMC" ist nicht einer.
Es ist ein multi-stage design mit mehreren Soft-IP Blöcken.

SA = intern
VDD(Q)_CPU (intern lower level & ausgehend, benützt um VREF zu erstellen)
VDD(2)_IMC (intern, low level & ausgehend als fertige VREF)
VDDQ_MEM (innerhalb des DIMM-PCBs, erstellt vom PMIC)
VDD_MEM (innerhalb des DIMM-PCBs, erstellt vom PMIC)

Syncronisierungsplatz ~ DIMM Slots.
Ich frage mich, ob ne hohe VDD den IMC degraden kann.

PS: PLL Termination voltage von 1.05 auf 1.1 zu setzen verringert den VDD Bedarf. Das fasziniert mich... dachte VTT sei CPU intern und VDD memory intern - aber irgendwie spielen die zusammen.
Welche VDD, es gibt 3.

Zu hohe VDD(Q)_CPU, ja.
Zu hohe VDD(2)_IMC , fast nein. Data Rail.
Zu hohe VDD_MEM , nein. DRAM kann nicht degraden, nur die Gold Fingers charren/anbrennen welches du reinigen kannst.

PLL Terminations ändern die min/max voltage range.
Nicht über 0°C zu empfehlen.
EDIT:
Nur für schlechte Boards zu empfehlen.
Termination PLL ändert den SA stand, welches ODT ändert. Sprich alle anderen Spannungen werden abgeändert.
Hohe PLL erlaubt niedrigeren VDDCR_SA bzw niedrigere IVR VDD2_IMC , jedoch nicht zu empfehlen.
Nicht für ambient szenarien. Definitiv nicht für Daily-Systems.
Kann dann ja nur mit einem saubereren Clock-Signal o.Ä. zutun haben.
Nein.
Im gegenteil.

(V) selber bedeutet nichts.
 
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