Corsair Dominator Platinum - Cant reach 3000mhz

[entropic remnants]

Just now, Ebkleks said:

"soc" is what exactly im realy a newbie in such things cause its my first build

SOC = "System On a Chip" and is used in this case to refer to the Ryzen on-board memory controller or "UMC".

 

This voltage went out of control on early Crosshair VI boards and actually fried mobo's.  Bad juju.

 

AMD's memory controller is a very early design for them -- their first DDR4 controller.  It benefits from a locked higher voltage but under no circumstances go higher than 1.2 and most recommend no higher than 1.15 to make sure spikes don't over-volt the thing.

[Ebkleks]

24 minutes ago, entropic remnants said:

Oh, Tom, you brought me back from a nice dream where I did not read forums and get mad at trolls and fan-boi types, lol.  I have been off the forums for awhile now and only came back to some with the release of Ryzen.. and then I've avoided them again.  So full disclosure: cranky old man who doesn't suffer fools gladly, even myself, lol.

 

Okay, off the soap box now.

 

First off, something about Corsair: they market the same product with different chips in order to hit demand.  Corsair sells more memory than anyone else.  Part of knowing what to do is knowing what you have and that can be an issue.  One thing you will notice is that Corsair kits often have a "version number" attached and that's an indication of what's used.  You don't see this so much on other products that tend to have only one hardware type per SKU.

 

Dominator Platinum have apparently used both Samsung and Hynix chips, but the ones I'm familiar with use Samsung D-die, E-die or B-die chips.  The E-die versions are dual-rank (due to lower density) and harder to get "up to speed".  I'm using CMD16GX4M2B3200C16 version 4.24 which are Samsung, dual-rank, E-die.  Version 4.31 of my kit are supposed to be B-die.  

 

I have been able to get these E-die variants up to speed since beta BIOS 9943/9945 and the succeeding beta BIOS which are based on AGESA 1.0.0.6.  I'm currently running beta 1403 with stable 3200 mhz memory on a low-spec 1700 non-X which will only go to 3800 mhz reliably.  It does involve manually setting many of the sub-timings exposed with the 1.0.0.6 AGESA.

 

What I will do is update a post I did on the ASUS forums about this with my latest information and put it on this forum.  But you can see my original post on how to do this over here: 

 

https://rog.asus.com/forum/showthread.php?93472-Get-Corsair-Dominator-CMD16GX4M2B3200C16-Up-to-Speed

 

You do NOT need Thaiphoon Burner as you can see the subtimings you need from the Intel XMP profile right in the Crosshair VI BIOS.

 

I'm not at sure why AMD doesn't read the XMP data on the sticks and fill in the appropriate sub-timings for you.  These current BIOS just are that smart yet.  Even when you do this some "tweaking is involved to get other timings optimized.

 

Here's my current results attached both in terms of performance, and the actual timings I have.  I'll make a new post here on what I did when I have the time, but the information I've given you can get you down the road.  Again, if you don't want to use the AGESA 1.0.0.6 beta BIOS to get there, you'll have to wait until ASUS s-l-o-w-l-y rolls out new production BIOS for the Crosshair VI.

 

 

 

  Reveal hidden contents

 

 

first of all thank you for taking your time to help, the thing is im a total newbie in this i realy dont understand the timings or their meaning.. dont have the slithest clue what your talking about

 

[Ebkleks]

2 minutes ago, Ebkleks said:

first of all thank you for taking your time to help, the thing is im a total newbie in this i realy dont understand the timings or their meaning.. dont have the slithest clue what your talking about

 

its my first build btw..

[entropic remnants]

Just now, Ebkleks said:

first of all thank you for taking your time to help, the thing is im a total newbie in this i realy dont understand the timings or their meaning.. dont have the slithest clue what your talking about

 

Understood.  When I do the full post on this hopefully it will result in a "step-by-step" you can try to see if you can get there.

 

By the way, this is one of the problems I have with ASUS and AMD: from their marketing, this is all easy and you are buying the highest quality system and it looks plug-and-play -- but man that's not reality right now.

 

I get mad at trolls who shame the newbies, and the manufacturers who neither apologize nor fix their shite in a timely manner.  Like I said: I'm old and crotchety.

[PCGuy_5960]

5 hours ago, Ebkleks said:

MSI x370 Titanium Gaming

Update your BIOS

[entropic remnants]

Oh, and also: most of us don't understand how those memory timings work, lol.  We research people who do, try variations, and arrive at working solutions.  

 

I'm an electrical engineer but no experience in chip design or memory so I'm stumbling through this myself -- you're not alone.  

 

Hang in there and over time you'll feel like you have a better idea how to manage these things and some of what those timings mean in terms of operation even if you don't have a technical understanding.

 

[Ebkleks]

Just now, PCGuy_5960 said:

Update your BIOS

alredy on newest BIOS

[entropic remnants]

8 minutes ago, PCGuy_5960 said:

Update your BIOS

See, this is how forums waste peoples time: that's been said, and it doesn't work for all the memory variations.  Generic as heck there and not particularly useful.

 

Sorry to be a troll myself, but c'mon -- do you have ANYTHING worth contributing better than THAT?  [excessive further insult deleted, lol]

[PCGuy_5960]

Just now, Ebkleks said:

alredy on newest BIOS

I think it's a motherboard problem, MSI's Ryzen BIOS is one of the worst AFAIK

Just now, entropic remnants said:

See, this is how forums waste peoples time: that's been said, and it doesn't work for all the memory variations.  Generic as heck there and not particularly useful.

 

Sorry to be a troll myself, but c'mon -- do you have ANYTHING worth contributing better than THAT?

Most of Ryzen's memory issues were fixed with BIOS updates, so this is why I suggested that

[entropic remnants]

1 minute ago, PCGuy_5960 said:

I think it's a motherboard problem, MSI's Ryzen BIOS is one of the worst AFAIK

Most of Ryzen's memory issues were fixed with BIOS updates, so this is why I suggested that

I apologize then, but you should have read some of the posts -- already suggested.

 

No, Ryzen's memory issues are FAR from sorted, lol.

 

Intel is SOOOOOO far ahead of AMD in DDR4.  It's actually at times shocking.  And this from someone who actually likes the AMD platform, lol.

[PCGuy_5960]

5 minutes ago, entropic remnants said:

I apologize then, but you should have read some of the posts -- already suggested.

 

No, Ryzen's memory issues are FAR from sorted, lol.

They are kinda sorta fixed  At launch people were having trouble getting their system to post with 2666MHz RAM and now there are a lot of people who are able to get 3000MHz stable.

 

These issues aren't fixed, but the situation has improved A LOT

[entropic remnants]

Just now, PCGuy_5960 said:

They were kinda sorta fixed  At launch people were having trouble getting their system to post with 2666MHz RAM and now there are a lot of people who are able to get 3000MHz stable.

 

These issues aren't fixed, but the situation has improved A LOT

Somewhat true: but not for certain DDR4 chips (as opposed to DIMM makers).  The same chips which would run at "rated overclock" (an interesting term in and of itself) on Intel won't on AMD without large amounts of hand tweaking.  I have to tweak about 10 parameters that SHOULD either come from the XMP profile (read automatically by the BIOS, FFS), or get set less conservatively during training.  This is true even on 1.0.0.6.  In fact, you can't even get some sticks up to speed prior to that and it's still not a production BIOS for my mobo.

 

That's the situation.  You're right it's better, but it's NOT fixed by any stretch of the imagination.

[Ebkleks]

15 minutes ago, entropic remnants said:

Somewhat true: but not for certain DDR4 chips (as opposed to DIMM makers).  The same chips which would run at "rated overclock" (an interesting term in and of itself) on Intel won't on AMD without large amounts of hand tweaking.  I have to tweak about 10 parameters that SHOULD either come from the XMP profile (read automatically by the BIOS, FFS), or get set less conservatively during training.  This is true even on 1.0.0.6.  In fact, you can't even get some sticks up to speed prior to that and it's still not a production BIOS for my mobo.

 

That's the situation.  You're right it's better, but it's NOT fixed by any stretch of the imagination.

what i dont understand is, some people have the same specs and all but they can post at 3000mhz wich i realy dont understand why... can a motherboard do this much chaos...

[entropic remnants]

Just now, Ebkleks said:

what i dont understand is, some people have the same specs and all but they can post at 3000mhz wich i realy dont understand why... can a motherboard do this much chaos...

Are you sure they have the same VERSION of the Corsair DRAM?  As I mentioned, that's critical and seldom mentioned despite that.  The same Corsair part number can have different chips on it depending on the version.

 

For instance, what version are yours?  It's written on the label on the DIMM and also on the box.

 

Where it's NOT told to you is in the listing to buy the memory.  Mobo manufacturers often have a "qualified vendor list" or QVL (or something like that) to tell you what to buy for best results.  For the Corsair memory, at least on the ASUS Crosshair VI QVL, it mentions a version for the Corsair parts -- and only for the Corsair parts because they seem to be the only ones who work things this way.

 

It makes it very hard to buy compatible Corsair memory unless you're looking at it in a store, get lucky, or some vendor I haven't seen yet TELLS you what version the memory is -- which they generally DON'T.

 

PITA

[entropic remnants]

Also, keep in mind that technically ANYTHING over 2133 mhz is an "overclock".  Intel did a good job with their XMP profile and memory training systems which result in all the Corsair variations "just working".  AMD hasn't gotten that far yet.

 

So, yes, a mobo/CPU can result in chaos when their memory adaptations are not robust enough.

 

Let me see if I can put this in a clearer light.

 

So lets say you buy 3200 mhz, 16-18-18-36 memory and they give you a rated voltage.  You THINK that's all you need to get an overclock.  You believe those timings "specify" your memory.

 

But there are other parameters "hidden" in the Intel XMP data on the DIMM's "SPD" flash memory that defines additional critical timing parameters that should be in place before training.  Training is basically running the memory through some test sequences and setting other technical-and-only-to-be-grokked-by-the-priesthood timings to fine tune it.  Training is not an exhaustive test though.

 

Most Ryzen mobos even at AGESA 1.0.0.6 do not appear to be using the XMP timings, but are substituting their own defaults from a table of some kind.  What you end up with does not match the XMP on many memory sticks.

 

What this means is that for a given clock speed, memory training will not result in a satisfactory memory overclock and/or stability.  Intel made memory folks specify those parameters for a reason -- and AMD and many if not most mobo makers are not following suit.

 

 

[entropic remnants]

Okay, one more thing to illustrate what I'm talking about.  See the attachment.

 

Here's the complete specification of my memory stored in a flash area on the DIMM itself.  Note the "Extreme Memory Profile v2.0" at the bottom. and the line I've highlighted.  That line gives a "complete" memory specification for the stick.

 

Besides the standard 16-18-18-36 timings, it specifies values for these items -- and you can look up what they mean:

 

tRC

tRFC1 (sometimes simply "tRFC")

tRFC2

tRFC4

tRRDL

tRRDS

tFAW

 

In practice, all those extra timings need to be set BEFORE training.  The basics of what I did to get my memory up to speed is to fill that information in by hand in the BIOS and Bob's-your-uncle, my memory worked.

 

I followed the insanely long Ryzen overclocking discussion on overclock .net (over 2000 pages, yikes!) to learn from masters how to further tweak for best performance -- a time consuming occupation I'll tell you.

 

EDIT: This thread is the one I watch: http://www.overclock.net/t/1624603/rog-crosshair-vi-overclocking-thread#post_25889679 -- that first post is not being updated by elmor much or at all, so it's keep up with the blitz over there or go mad.

 

[entropic remnants]

Started work on a tutorial: 

Not finished but I'll get to it.

[Ebkleks]

On 6/27/2017 at 7:32 PM, entropic remnants said:

Started work on a tutorial: 

Not finished but I'll get to it.

still stuck with the same problem no method works

 

[Ebkleks]

On 6/27/2017 at 10:49 AM, TheRandomness said:

Try remove one stick and XMP again. Might just be the memory failing to train properly.

none of the method worked for me.. no progress at all

[TheRandomness]

10 minutes ago, Ebkleks said:

none of the method worked for me.. no progress at all

Well, I've run out of ideas, sadly. Do you have any other RAM sticks to test? Or another motherboard to test with? Because you can try RMA if it's not working.

[Ebkleks]

50 minutes ago, TheRandomness said:

Well, I've run out of ideas, sadly. Do you have any other RAM sticks to test? Or another motherboard to test with? Because you can try RMA if it's not working.

Nope dont have any other motherboard...

and whats RMA?

[TheRandomness]

Just now, Ebkleks said:

Nope dont have any other motherboard...

and whats RMA?

Sending the motherboard and/or RAM back to the seller so they can test and see if its faulty, then sending you a new kit or refunding you if they were faulty.

[entropic remnants]

RMA = Return Merchandise Authorization.

 

You have basically two choices: some vendors will simply take the merchandise back within a certain time period but there can be a "restocking charge" and you'll pay shipping.  For that you deal with who you bought it from and it varies from vendor to vendor.

 

For an RMA you'll need to register the board/memory with whomever makes it to activate the warranty.  Once that's done, you contact them to arrange to return it and they give you an RMA Number that you must use when returning it so they know what to do with it.

 

Good luck and sorry you didn't have success.

 

[Ebkleks]

On 6/30/2017 at 1:33 AM, entropic remnants said:

RMA = Return Merchandise Authorization.

 

You have basically two choices: some vendors will simply take the merchandise back within a certain time period but there can be a "restocking charge" and you'll pay shipping.  For that you deal with who you bought it from and it varies from vendor to vendor.

 

For an RMA you'll need to register the board/memory with whomever makes it to activate the warranty.  Once that's done, you contact them to arrange to return it and they give you an RMA Number that you must use when returning it so they know what to do with it.

 

Good luck and sorry you didn't have success.

 

hello once again

I just got a new update for the BIOS version: 7A31v17

But i dont know if i shoulld install it or brings more problems...

 

[entropic remnants]

6 hours ago, Ebkleks said:

hello once again

I just got a new update for the BIOS version: 7A31v17

But i dont know if i shoulld install it or brings more problems...

 

Well, I'm running the ASUS Crosshair VI Hero so my BIOS versions are different.  There's no way to answer that question easily I'm afraid.  Perhaps someone with your motherboard might chime in on that.