tibcsi0407
Semiprofi
I will let it to run more time now and send you. Will have a lot of errors for sure. 🤣
Let's see how it works.
Beitrag automatisch zusammengeführt:
What do you think? Too much Vdimm?
Zuletzt bearbeitet:
[Sammelthread] Intel DDR5 RAM OC Thread
- Ersteller even.de
- Erstellt am
Let's see how it works.
Less VDDQ_MEM ~ more droop
DIMM PCB Crash, or SNR issue
Before that, enable RX-DFE
and dont change anything else
Just to validate.
tibcsi0407
Semiprofi
Thanks.
Should I change the TX too to keep my delta?
Edit
It was on auto, so I enabled it.
With start fresh i didnt mean , throw away the 8400 profile
Sorry.
I ment , scale up
Throw away the old profiles that have no V/F change nor SA/TX/MC change.
Also use:
tibcsi0407
Semiprofi
I did not load the old profile, this is new with our V/F.
The 8600 profile I had had similar Voltages like my 8400C36.
Don't check the timings.
This was a HT off profile with only short Y and TM5 test. (that was my stable before, now 90 min VST/VT3 is the stable line for me)
But this profile shows just you can run itThis was a HT off profile with only short Y and TM5 test. (that was my stable before, now 90 min VST/VT3 is the stable line for me)
Not that it was stable
I dont now what to do with that one~
"Throw this one away"
It was tested too short
EDIT:
What it may also show, is that IMC had enough voltage for itself
When half of the CPU doesnt run 🤭
tibcsi0407
Semiprofi
Yes, the new is completely different, I learned a lot since and now I am using your TRFC mini as a base for timings.
My RAM can do that, the question is my IMC capable? Or am I capable?
Stopped the old TM5 at 5. round and started again in the one you sent me. Is there any difference, or just the look?
IMC is, else it wouldnt pass neither TM5 nor VST
A bit of difference on resource management, and on the config too.
IMC is only as capable, as voltage or power headroom you have.
Or rather give it.
tibcsi0407
Semiprofi
That's what I am thinking on.
We will see. 3 rd round will pass soon, still no issues.
We can still play with RTT's and ODT it's still on auto.
Step by step, i need to check on DFE state first.
8800 may need per channel RON figuring out, if both channels are identical or one is weaker
EDIT:
And 8600 may need some training improvements later down the line
tibcsi0407
Semiprofi
Now I am in 6th cycle, error free atm. The error 4 came in 7th cycle before.
8800 seems impossible for me with these sticks. But who knows.
Still ok ?Now I am in 6th cycle, error free atm. The error 4 came in 7th cycle before.
8800 seems impossible for me with these sticks. But who knows.
tibcsi0407
Semiprofi
Yeah, we're at halfway. My magical VDDQ -- TX delta works here too. LOL
But we still have some cycles left, so we will see.
I am even doing some work (with a lot of save button) on my other screen.
Easy work 🤭
But in all seriousness,
~ why is RX DFE on stock not sufficient.
I'd rather ask that first.
Its not like it changes too much between IC SKUs.
Even a soft CTLE+DFE 3 tap (1 TX + 1 RX) would be fine
I personally like 3 on RX, but still
Why is this not standard behavior.
tibcsi0407
Semiprofi
I don't know, but that could be the part of the retraining issues. I just turned it on from auto and didn't get error 4 yet with the same voltages. MRC Fastboot is disabled of course. (14th cycle now, still error free.)
Beitrag automatisch zusammengeführt:
1st one came at 1:07, it's error 13. Could be too low SA?
Zuletzt bearbeitet:
I know i waste your time a bit
But once TM5 finishes with proof
I would like you to put back RX DFE to auto, and try to lower VDDQ_MEM a bit.
Adapt VDDQ_CPU.
#4 is still an overcurrent crash for whatever RTTs the Board trains.
It would be good to try and fix that first, before using extra help from DFE for higher clock onwards.
RX (receiver) is on MEM
It should crash sub 30min probably again.
Its one of the VDDQ's i think.
Else its just badly trained DQ (training itself can be bad too vs no training)
I'm not sure between both. On of both based on magical RTTs ASUS-HQ uses.
I somewhy bet too much RTT_WR causes that, as thats been recently a trend for 16gb dimms.
But anywho, one of the VDDQ's it likely is. Causing Dimm PCB Crash.
tibcsi0407
Semiprofi
Sent a new pic with error 13 while you wrote this comment.I know i waste your time a bit
But once TM5 finishes with proof
I would like you to put back RX DFE to auto, and try to lower VDDQ_MEM a bit.
Adapt VDDQ_CPU.
#4 is still an overcurrent crash for whatever RTTs the Board trains.
It would be good to try and fix that first, before using extra help from DFE for higher clock onwards.
RX (receiver) is on MEM
Lower further? Isn't it too small already?
You are not wasting my time, I am the one who is wasting your time.
If you didnt stop it let it collect one more at least.Sent a new pic with error 13 while you wrote this comment.
Lower further? Isn't it too small already?
Ignore last comment suggestion. Its not stable still.
Lets keep bandaid for now
#13 is still an SNR issue.
EDIT:
if no new error in ~15min
Then lets go down voltage rabbit hole.
First verify me that y-cruncher at least does 2 cycles, or 3. Just something to work off from.
Then what happens if you put SA to 1.16 and run TM5 again
tibcsi0407
Semiprofi
I will let it finish. I will leave it here and come back later and send it to you.
I have fun, don't worry.You are not wasting my time, I am the one who is wasting your time.
Want to see 8800.
Until later then~
Have to go out 🤭
It shouldn't BSOD.
Chance is too low.
Just hoping for no #0 spam.
Beitrag automatisch zusammengeführt:
Sheet says increase ODT (value) or start messing with RTTs.
Soo likely SA bit higher. But i want to tripple verify if lower SA wont fix it and if something else starts crashing. *
Should rule out training inconsistency too. Can also just be bad DQ-VREF.
But i see big changes on the Bios since 0071
Maybe tuning is better now.
We'll see~
Step by step.
EDIT:
* May needs bigger distance on VDDQ_CPU. Should tell by itself with lower SA.
Cant differentiate between MEM or CPU ~ VDDQ issue. Outside of sheet reporting ODT issue.
Zuletzt bearbeitet:
Wer Lust hat kann auch hier mit mal rumtesten, https://www.techpowerup.com/317943/...ve-settings-ecc-support-for-select-ryzen-cpus. Schönen Abend.
tibcsi0407
Semiprofi
It was frozen when I came home with black screen 🤣
Something should be really wrong. Maybe it spammed error 0's.
With 1.12 SA I got insta freeze.
1.15 SA lots of error 5.
Let's try it with 1.20V?
Zuletzt bearbeitet:
Way to big steps
Log.txt should have logged the errors
Under 1.2, 1.2 is a switching point
Over 1.15
tibcsi0407
Semiprofi
It was on 1.18 with the long test. Maybe TX needs change and SA should be on 1.18V.
Exactly, 1.19 will make a change
1.2 will make too many changes. It will mess with VDD and VDDQ_CPU . Both. too many variables.
1.17 or 1.165 is just about max , that will also make a change
Did you find log.txt ?
tibcsi0407
Semiprofi
Now just started it with 1.45V VDDQ and 1.30V TX.
Will report back and use your suggestions if it won't work.
Afternoons are always crazy, kid is at home. ☺️
tibcsi0407
Semiprofi
Sometimes it's hard to find the balance. ☺️Focus on your life, less on this field here.
Whenever you can
I'm already in Dad's age.
The child always has priority. Like without even a doubt
Your hobby is the other child. Although the Partner usually has equal prio. 50:50
Glad you're not an XOCer. That would hurt separating time for that.
tibcsi0407
Semiprofi
Yes, it is true. Little girl is everything for me. ☺️I'm already in Dad's age.
The child always has priority. Like without even a doubt
Your hobby is the other child. Although the Partner usually has equal prio. 50:50
Glad you're not an XOCer. That would hurt separating time for that.
