Overclock.net banner
13621 - 13640 of 14146 Posts

·
Registered
Joined
·
152 Posts
What timings can I try tightening with DR Hynix CJR? I had to increase tRDWR from 8 to 10, increase VDIMM to 1.38v, otherwise I had a lot of errors on XMP settings.
Still get an occasional error 0 with 1usmus and error 1 with Universal profiles. Tried increasing tWTRS from 5 to 6 but seems to error out sometimes.

Increasing frequency is difficult, 3733 I get WHEA 19 unless I increase vSOC but then I get a lot of error 6 and BSOD
 

Attachments

·
Iconoclast
Joined
·
31,568 Posts
What timings can I try tightening with DR Hynix CJR? I had to increase tRDWR from 8 to 10, increase VDIMM to 1.38v, otherwise I had a lot of errors on XMP settings.
Still get an occasional error 0 with 1usmus and error 1 with Universal profiles. Tried increasing tWTRS from 5 to 6 but seems to error out sometimes.

Increasing frequency is difficult, 3733 I get WHEA 19 unless I increase vSOC but then I get a lot of error 6 and BSOD
tRCDWR, tRC, tRRD_S & _L, tFAW, and tRTP/tWR can probably be tightened significantly. tWTR_S & L should be able to handle a bit tighter. Sometimes setting RttNom to 34 or 40ohm can help.

A lot of mediocre bin CJR cannot handle tight primaries without active cooling or excessively loose subtimings. Sometimes the best results are achieved by loosening the primaries and tightening up everything else as far as is practicable.

This is one of my budget Timetec CJR DR kits at 1.32v:


That said, errors on XMP settings sounds like a serious misconfiguration somewhere, or defective memory.
 

·
Registered
Joined
·
42 Posts
i habe a question if i changed my CLDO VDDP Voltage = 0.9v go my automatic the tPHYRDL from 26 to 28? are this normal i have read the 26 are better then 28 are this right any help about the tPHYRDL?
 

·
Registered
Joined
·
152 Posts
tRCDWR, tRC, tRRD_S & _L, tFAW, and tRTP/tWR can probably be tightened significantly. tWTR_S & L should be able to handle a bit tighter. Sometimes setting RttNom to 34 or 40ohm can help.
I've previously tried tRRD_S/L and tFAW 6,6,24 and increased tFAW to some other value I can't remember, both with errors. The problem is I still don't have a 100% stable baseline. I don't know if changing some parameters introduces new errors or are they the same old issues still lingering.

I'll start with tRCDWR and see how it goes.

That said, errors on XMP settings sounds like a serious misconfiguration somewhere, or defective memory.
I've seen other people with the same issues with TM5. I can pass y-cruncher 10+ hours, HCI Memtest 1000%, RamTest ~6000% and Memtest86+without errors with XMP, only TM5 poses a problem.
 

·
Registered
Joined
·
2,291 Posts
I've previously tried tRRD_S/L and tFAW 6,6,24 and increased tFAW to some other value I can't remember, both with errors. The problem is I still don't have a 100% stable baseline. I don't know if changing some parameters introduces new errors or are they the same old issues still lingering.

I'll start with tRCDWR and see how it goes.


I've seen other people with the same issues with TM5. I can pass y-cruncher 10+ hours, HCI Memtest 1000%, RamTest ~6000% and Memtest86+without errors with XMP, only TM5 poses a problem.
Which TM5 config are you using ??
 

·
Registered
Joined
·
2,291 Posts

·
Registered
Joined
·
152 Posts
With the screenshot above (modified XMP profile, tRDWR from 8 to 10, increase VDIMM to 1.38v) sometimes I get error 0 usually after 15+ cycles, sometimes I can pass 25 cycles without errors.
I'm suspecting tRRD and tWTR, but from what I can see both are pretty loose.
 

·
Registered
Joined
·
2,291 Posts
With the screenshot above (modified XMP profile, tRDWR from 8 to 10, increase VDIMM to 1.38v) sometimes I get error 0 usually after 15+ cycles, sometimes I can pass 25 cycles without errors.
I'm suspecting tRRD and tWTR, but from what I can see both are pretty loose.
Thats not the 1usmus v3 config though is it ?
 

·
Registered
Joined
·
2,291 Posts
No, meant with those ZenTimings settings
OK im with you,

I dont have experience with DR Hynix CJR, though a couple of generic suggestions.

First are you able to increase RttPark to RZQ/2 or RZQ/3 ?

Also in combination with the RttPark change, I would first decrease voltage and if that does not work increase it slightly

I remember tweaking a set of Corsair CMK32GX4M2D3600C18 modules and less voltage was better, anything above 1.38v and I was getting weird stability issues, hence the reason I suggest weakening RttPark.

Hopefully CJR owners can chime in

:)
 

·
Iconoclast
Joined
·
31,568 Posts
I've seen other people with the same issues with TM5. I can pass y-cruncher 10+ hours, HCI Memtest 1000%, RamTest ~6000% and Memtest86+without errors with XMP, only TM5 poses a problem.
That's why we're using TM5. The errors it detects are real and an actually stable memory subsystem will pass it.

If XMP settings aren't stable and you're sure there aren't problems elsewhere (CPU or IMC), then the memory is defective by virtue of not being stable at it's rated specifications. Unfortunately, this is not terribly uncommon. Most module builders seem to leave very little in the way of margins for their various bins...banking on few of their customers really checking things out.
 

·
Registered
Joined
·
9 Posts
Hi, it's been a while since I've been there, too busy with work.
I got my cousin's PC, he has 4x8 Crucial Micron E-Die C9BVK, I noticed the voltage scaling for tRC and tRFC is zero. Also I am annoying because I cannot follow the basic rule tRC = tRP + tRAS because below 50 impossible to boot. Same for tRFC, below 600 impossible to boot.
I peeled a lot of message, but the topic before too quickly to be able to find an answer to my question: What rule should I apply for tRC?
If in passing you see any inconsistency in the timings let me know. On the image only tRC and tRFC are not optimized, in auto.
@Veii Not that the opinion of others does not interest me far from there, but I must admit that your explanations are always clear and precise, if ever you are still among us, a quick little analysis is no refusal.
Welcome to all who can enlighten me on the EDie Micron.

View attachment 2528467
These are my settings try to see if they work for you
Light Product Black Font Screenshot
 

·
Premium Member
Joined
·
6,445 Posts
it is more correct to increase the testing time - up 2000%-10000%, rather than the number of cycles

For you:
tRAS=23 max tRAS for tRCconstanta - better
tCWL=14
RDWR=8
WRRD=4 or 5
tRFC=256 or 248

Why RCDRD<RP ?
RP=16
and
decrease RC with RAS
tCWL 12 gives me better write and copy in AIDA, tWRRD at 1 better too, tRDWR needs to be at 10 for tCWL 12, tRFC is according to the tRFC calculator, tRP 16 tRAS 26 gives me worst results and harder to get stable than my settings at 19-23 which started out as an experiment suggested by @Veii and I find really works well.

I DID change tRAS to 23 though, helps with latency a bit to have tRP+tRAS=tRC.

The TM5 is 3 cycles of Universal at 1000% I ran overnight.

Font Screenshot Software Computer Electronic device


Font Screenshot Multimedia Software Electronic device
 

·
Registered
Joined
·
152 Posts
First are you able to increase RttPark to RZQ/2 or RZQ/3 ?
Tried with RttPark RZQ/2 (120Ohm in BIOS I think), left VDIMM @1.38v; passed 10 cycles of Universal and 25 cycles of 1usmus, which is the first time I could pass both consecutively without errors. Now I'm running RamTest until 20000+% just in case. Afterwards will try with XMP settings + modified RttPark to rule it out as the only problematic setting.
Next, decreasing tRRD, tWTR etc. Fingers crossed it behaves better with new RttPark.
That's why we're using TM5. The errors it detects are real and an actually stable memory subsystem will pass it.

If XMP settings aren't stable and you're sure there aren't problems elsewhere (CPU or IMC), then the memory is defective by virtue of not being stable at it's rated specifications. Unfortunately, this is not terribly uncommon. Most module builders seem to leave very little in the way of margins for their various bins...banking on few of their customers really checking things out.
I'd agree with you but the way I see it there are two issues here:
-XMP profile doesn't dictate secondary or tertiary timings, that's the motherboards job. It's highly likely my MSI B550I ITX doesn't have a robust memory system and secondary/tertiary timings are at fault.
-Sending the RAM back for warranty won't yield any results as they'd probably just run memtest86+ and call them fault free...
 

·
Iconoclast
Joined
·
31,568 Posts
-XMP profile doesn't dictate secondary or tertiary timings, that's the motherboards job. It's highly likely my MSI B550I ITX doesn't have a robust memory system and secondary/tertiary timings are at fault.
XMP 2.0 can include more and usually has at least the tRCs and RRDs and TFAW, while the JEDEC or AGESA formulae for the rest should work. Of course, none of this is a given and manual tuning should be better anyway.

Your MSI B550I should be a fairly solid board for memory OCing though. MSI usually keeps up with firmware pretty well and only having two slots is an advantage.

-Sending the RAM back for warranty won't yield any results as they'd probably just run memtest86+ and call them fault free...
True.
 

·
Registered
Joined
·
42 Posts
Blue Grille Font Mesh Electronic device


this are what i get on windows 11 with my 16-16-16-16-32-48 are this okay? i see much people got low latence then me but i dont know why
 

·
Overclock the World
Joined
·
3,283 Posts
View attachment 2528532

this are what i get on windows 11 with my 16-16-16-16-32-48 are this okay? i see much people got low latence then me but i dont know why
Try this patch, and report back if it helps please
kb numbers can be just written out, mb needs to be converted down to kb
Be sure to switch it to decimal instead of hex
Also grab the Win 11 chipset drivers
Rectangle Font Screenshot Parallel Software
Win+R , type WinVer , just to know what you run
 
13621 - 13640 of 14146 Posts
Top