Question RAM Resets itself to 2133 upon cold boot?

[hharriswilliam]

Hey! So this is my 2nd ever build, and it seems that others have a similar issue from searching EVERYWHERE, long story short;

My RAM (Gskill Ripjaws V 2x8GBs) is rated to run at CL16-18-18-38 @ 1.35V. I don't have any issues having to make changes in the bios, as the computer boots quick and fine. A few issues though; upon shutting down, if I were to boot the PC after, ONLY the RAM clocks reset to CL15, the standard clocks. I'm assuming the mobo isn't retaining my settings due to possibly a bad battery or a PSU. I haven't had any issues with this system ever since I've assembled it (about a week ago)

EDIT: I've tried the A1/B1 configurations versus the A2/B2 slots, as my mobo states in the manual they should be in the 2nd/4th slots.


I've also tried using both the XMP 2.0 profile, a mix of the XMP profile and custom settings, and simply custom settings to DDR4 3200 @ 16-18-18-18-38 @ 1.35V & @ 1.360V.

I've updated the BIOS to the latest one (1.70), I've considered trying the beta BIOS, but I'll wait until the official ones drop.

I'm assuming this could be a CMOS issue, MOBO or a PSU issue.

My specs are as follows:

AMD Ryzen 5 3600
AsRock Phantom Gaming 4 X570
Gskill Ripjaws V (F4-3200C16D-16GVRB)
Western Digital Blue SSD 250GB
Corsair 430M, Bronze Rated
Geforce 1050 TI

---------------------------------------

A few other notes:

I moved this PSU from my previous system (a FX-6300 system)
No really bad issues other than it restarting once during boot-up, only to realize ONLY the memory clocks have reset to 2133, instead of 3200.
The CPU/DRAM LEDs turn on, but then quickly switch off to the VGA/HDD lights.
I've checked the slots and cables, everything's firmly in place.



I'm planning on upgrading the GPU and PSU very soon, but I'm just trying to get a grip on what the possible issues could be. I'm gonna attempt replacing the CMOS battery coin .

Any ideas from what I've mentioned?


EDITED

 

[damric]

Darkbreeze is 100% right. Always follow what the motherboard says for slots because they wire them such for better signalling. This matters much more now than in past years because of the pure speed of the data going down the traces.

However, I think you can do better with this memory overclock if you try harder. I'm on a really poor second hand x370 pulling incredible speed with Hynix RAM. You just got to have to patience and enough gumption to get it done.

I hope this helps:

https://forums.tomshardware.com/threads/ryzen-memory-speed-tuning-you-can-do-it.3542552/

 

[hharriswilliam]

At this point, I'd BE inclined to say it's a motherboard issue. That memory, according to G.Skill, is 100% compatible with that board based on G.Skill's memory configurator and G.Skill pretty well knows their business when it comes to what boards a given memory kit is approved for. I'd trust their word way before I'd trust the word of the motherboard manufacturer or their little bitty QVL list.

To begin with, that's not a terrific board. It's rather low end for THIS particular chipset. Even so, that board and that memory kit SHOULD be fine together, so something else is definitely going on and I think I'd start by contacting ASRock and G.Skill to see what they have to say about it. Point them to this thread if necessary. They are certainly both welcome to chime in here with their input if they wish. I know G.Skill participates here occasionally although I've never seen any official representatives from ASRock on this forum.

So I fixed the issue.


It was the Power Supply.
I had a CX430M Corsair PSU since about 2013, so I guess it was starting to fail.
I ended up buying the Corsair CX650M.

Nonetheless, I booted the PC up after putting it in, and it completely booted to Windows. Check the RAM and it was 2133. So I shut down, and applied the XMP settings > Booted to Windows.

THEN, I shut down my PC, waited 5 seconds, and powered on with 3200mhz. NO ISSUES!


I assumed it was the Power Supply, but nonetheless my issue has been solved. Thanks so much everyone!!

 

[Endre]

That's wrong. It absolutely matters. And there are very specific reasons for it which I won't get into because they are too technical to be explained, and understood, here, but suffice to say that the architecture is SPECIFICALLY designed for those DIMM slots, 2 and 4, to be used FIRST. Some boards won't even WORK properly if you populate the first and third DIMM slot with nothing in the second and fourth slots. I've seen this more times in more threads than I can recall although I'm sure I can point you to some of them is you require emperical evidence.

Certainly, it CAN work in those slots, but it is not preferred, it is NOT what the manufacturers, ALL of them, at least on dual channel Intel and AMD DDR4 motherboards, recommend by way of THEIR population rules, nor does it comply with the recommendations in the JEDEC DDR4 SDRAM standards technical documentation.

It is certainly MORE than just because using those slots affords better clearance for most aftermarket CPU coolers, although that too, is an obvious consideration.

Here they recommend slots A1 + B1.
They even say it’s wrong populating slots A2 + B2 first!
I don’t think it matters, as long as you’re putting them in “the same color”, and being on dual-channel mode.

https://superuser.com/questions/691692/which-slots-are-proper-to-install-2x8gb-of-ram

 

[hharriswilliam]

It may be a problem with your motherboard. Motherboard may be weak or burnt. Otherwise it can be from memory

My problem has been solved, see the post above

 

[Darkbreeze]

Here they recommend slots A1 + B1.
They even say it’s wrong populating slots A2 + B2 first!
I don’t think it matters, as long as you’re putting them in “the same color”, and being on dual-channel mode.

https://superuser.com/questions/691692/which-slots-are-proper-to-install-2x8gb-of-ram

That is because they are idiots that don't know anything about memory architecture. ALL, AAAAALLLLL motherboards that fall within the purview of what I have been indicating, SAY that channel A is A1 and A2 and that channel B is B1 and B2, and that is correct. Those slots DO belong to those channels. What people like you, and those who posted that misinformation in that thread you linked to do NOT understand, is that in order to use memory in DUAL CHANNEL, you must populate the memory in one slot from channel A and one slot from channel B, hence DUAL channel. If both DIMMs were in A1 and A2 then they would both be in the same channel and you could not HAVE dual channel. If you look at the graphical guide for that board it CLEARLY shows DIMM slots A2 and B2 highlighted in the same color and slots A1 and B1 highlighted with the same color. It's not rocket science, although I can fully appreciate why so many people across the web misinterpret the channel designations to indicate those are the channels that should be used for dual channel and I wish motherboard manufacturers would simply STOP putting it in the user manuals so they stop confusing people who don't understand WHY it is there in the first place. It is not needed or necessary to be there. The only thing the consumer needs to know is the population rules, not which DIMM slots belong to which channel.

Regardless, the fact remains, A2 and B2 are to be used for a single pair of sticks. That does not change for any modern dual channel board regardless of whether it is Intel or AMD. Triple, quad and other multiple channel architectures will have their own specific population guidelines but we are not concerned here with HEDT platforms anyhow.

 

[Darkbreeze]

So I fixed the issue.


It was the Power Supply.
I had a CX430M Corsair PSU since about 2013, so I guess it was starting to fail.
I ended up buying the Corsair CX650M.

Nonetheless, I booted the PC up after putting it in, and it completely booted to Windows. Check the RAM and it was 2133. So I shut down, and applied the XMP settings > Booted to Windows.

THEN, I shut down my PC, waited 5 seconds, and powered on with 3200mhz. NO ISSUES!


I assumed it was the Power Supply, but nonetheless my issue has been solved. Thanks so much everyone!!

Normally, for almost any problem, the PSU is the first thing I go after. Knowing that your CX430w wasn't a super bad unit and never having seen a PSU work fine in all other regards but cause memory to reset, I'd have never considered that as the problem on this. Good call. In the future, knowing the age of hardware COULD be helpful in diagnosing problems because IF I had known that PSU was 6 years old I'd have told you it probably needed to be replaced as a matter of course anyhow. Glad you figured it out and can move on now though.

 

[hharriswilliam]

Normally, for almost any problem, the PSU is the first thing I go after. Knowing that your CX430w wasn't a super bad unit and never having seen a PSU work fine in all other regards but cause memory to reset, I'd have never considered that as the problem on this. Good call. In the future, knowing the age of hardware COULD be helpful in diagnosing problems because IF I had known that PSU was 6 years old I'd have told you it probably needed to be replaced as a matter of course anyhow. Glad you figured it out and can move on now though.

I mentioned in the beginning I moved it over from my older PC, which was a FX-6300 system. Perhaps that could've gave some insight, but I planning on upgrading it because I also plan to upgrade the GPU.

Thanks SO much though for your input, learned some stuff too in the long run!

 

[damric]

Yeah the old CX (before 2015) units were really bad. Corsair only rated them for 30'C and 2 year warranty, and they hold the record for most RMAs of a PSU model. The new ones are much better, like night and day, and actually meet newest ATX specification, rated at 50'C.

 

[Endre]

That is because they are idiots that don't know anything about memory architecture. ALL, AAAAALLLLL motherboards that fall within the purview of what I have been indicating, SAY that channel A is A1 and A2 and that channel B is B1 and B2, and that is correct. Those slots DO belong to those channels. What people like you, and those who posted that misinformation in that thread you linked to do NOT understand, is that in order to use memory in DUAL CHANNEL, you must populate the memory in one slot from channel A and one slot from channel B, hence DUAL channel. If both DIMMs were in A1 and A2 then they would both be in the same channel and you could not HAVE dual channel. If you look at the graphical guide for that board it CLEARLY shows DIMM slots A2 and B2 highlighted in the same color and slots A1 and B1 highlighted with the same color. It's not rocket science, although I can fully appreciate why so many people across the web misinterpret the channel designations to indicate those are the channels that should be used for dual channel and I wish motherboard manufacturers would simply STOP putting it in the user manuals so they stop confusing people who don't understand WHY it is there in the first place. It is not needed or necessary to be there. The only thing the consumer needs to know is the population rules, not which DIMM slots belong to which channel.



Regardless, the fact remains, A2 and B2 are to be used for a single pair of sticks. That does not change for any modern dual channel board regardless of whether it is Intel or AMD. Triple, quad and other multiple channel architectures will have their own specific population guidelines but we are not concerned here with HEDT platforms anyhow.

Preferring A2+B2 over A1+B1 is probably your own preference.
By calling the guys from the link that I posted “idiots” is not a proof that they are wrong and that you are right!
I didn’t see A2+B2 being the “correct combination” either in my motherboard’s manual or on the internet.
The only “wrong positions” are if you populate A1+A2, leaving channel B empty; or B1+B2, leaving channel A empty!
Have a nice day!

 

[Darkbreeze]

Yes, it is. Because they clearly have no clue. You show me ANY, ANNNNYYYY, motherboard manual for any dual channel consumer motherboard using DDR4, or even high density DDR3, that says something different and I'll retract that statement. But you can't, so that's not going to happen.

 

[Endre]

Yes, it is. Because they clearly have no clue. You show me ANY, ANNNNYYYY, motherboard manual for any dual channel consumer motherboard using DDR4, or even high density DDR3, that says something different and I'll retract that statement. But you can't, so that's not going to happen.

Here’s an example that both combinations are correct.
This is the manual of Gigabyte Z390 Aorus Master. Page 17.

https://download.gigabyte.com/FileList/Manual/mb_manual_z390-aorus-master_1001_190219_e.pdf

 

[Darkbreeze]

That shows no such thing. It still shows A2 and B2 as the PREFERRED configuration, which is why it is listed first in the order of priority. Nobody has ever said that A1 and B1 won't EVER work, because they work on some boards, but they are not the preferred configuration based on the recommendation of the manufacturers population rules. If they were, they would be listed first. And as I said, on SOME boards, the implementation is such that populating A1 and B1 only just don't work, or they work, but there is no dual channel. Or they work, but something ELSE about the configuration is off such as "the memory resets itself to 2133mhz" among other possible configuration problems.

 

[Endre]

That shows no such thing. It still shows A2 and B2 as the PREFERRED configuration, which is why it is listed first in the order of priority. Nobody has ever said that A1 and B1 won't EVER work, because they work on some boards, but they are not the preferred configuration based on the recommendation of the manufacturers population rules. If they were, they would be listed first. And as I said, on SOME boards, the implementation is such that populating A1 and B1 only just don't work, or they work, but there is no dual channel. Or they work, but something ELSE about the configuration is off such as "the memory resets itself to 2133mhz" among other possible configuration problems.

Please look again at the plan from page 17.
There is no preference for A2+B2.
Is either A1+B1 or A2+B2.

 

[Darkbreeze]

No it isn't. Like I said, you, and the people who posted that advisement, simply do not understand how it works. I tried to explain it to you, but clearly you are not understanding it. As I said, the A1 + B1 and A2 + B2 are CHANNELS. You need TWO channels in order to have DUAL CHANNEL. You CANNOT have DUAL CHANNEL if you are using two DIMM slots from the same channel. I'm not sure why that is so hard to understand, but if you look at the population rules graphic for that motherboard on page 17, hopefully you will understand better.

If you install a DIMM in slots A1 and A2 or in slots B1 and B2, then you are only populating both DIMMs of the same channel, and there is NO possible way DUAL CHANNEL can happen. Technically, if you populated slots A1 and B2, you COULD, theoretically, have dual channel because you are in separate channels, however the architecture is not designed to pair those channels together. Only A1 with B1 and A2 with B2, and A2 with B2 (Or second DIMM slot with fourth DIMM slot, regarding of what name is used) adhere to the population rules based on how the architectures are designed according to specification and standards. That is all.

 

[Endre]

No it isn't. Like I said, you, and the people who posted that advisement, simply do not understand how it works. I tried to explain it to you, but clearly you are not understanding it. As I said, the A1 + B1 and A2 + B2 are CHANNELS. You need TWO channels in order to have DUAL CHANNEL. You CANNOT have DUAL CHANNEL if you are using two DIMM slots from the same channel. I'm not sure why that is so hard to understand, but if you look at the population rules graphic for that motherboard on page 17, hopefully you will understand better.

If you install a DIMM in slots A1 and A2 or in slots B1 and B2, then you are only populating both DIMMs of the same channel, and there is NO possible way DUAL CHANNEL can happen. Technically, if you populated slots A1 and B2, you COULD, theoretically, have dual channel because you are in separate channels, however the architecture is not designed to pair those channels together. Only A1 with B1 and A2 with B2, and A2 with B2 (Or second DIMM slot with fourth DIMM slot, regarding of what name is used) adhere to the population rules based on how the architectures are designed according to specification and standards. That is all.

I do understand how dual channel works.
Channel A contains A1 & A2
Channel B contains B1 & B2.
You’ll need to populate both channels to have dual channel enabled.
I agree with 99% of all that you’ve said.
I also didn’t agree with the guys from the link that I posted, saying that in dual channel mode A2+B2 is a wrong combination.
But I also don’t see A1+B1 being wrong either.

Again, I agree with all that you’ve said except of the claim that in dual channel mode A2+B2 should be populated first.

Have a good day!
Let’s just leave it here.
We both have said enough.

 

[gskill support]

I usually say second and fourth slots away from CPU because I do recall some motherboards that list these as A1 B1 (first). As far as whether first and third slots are OK to use, it varies with motherboard. However in most cases the second and fourth slots will perform better, but first and third may be able to work fine.

 

[Darkbreeze]

JEDEC states explicitly that for each channel, DIMM 0, which is the FURTHEST DIMM slot from the CPU for each channel (Has nothing to do with actual DIMM slot naming scheme) is, by standard, the recommended DIMM slot. That part isn't open for discussion. They create the standard and partners comply with it.

This is why:

From the electrical engineering point of view, the DRAM dies provide adjustable bus termination and putting the only active DIMM for each channel on each channel's last slot puts them at the end of the bus where those bus terminations will be most effective at mitigating signal reflections, which should give you the likely best chance at trouble-free operation - the 5mm unterminated bus stub from an unpopulated DIMM slot before the bus termination is less disruptive than a 10-15mm stub at the end of the bus from the extra PCB trace distance and connector.

Support for multiple DIMMs per channel almost got scrapped from the DDR4 spec due to such signal integrity concerns.

DIMM slots which are the second and fourth, for dual channel 4 slot boards, ARE the slots furthest, at the end of the bus.

Whether other slots "work" or not probably depends on the board and the implementation. Maybe even depends on the memory itself or the strength of the IMC on a given part. It's still not the "recommended" configuration by any of the big four and does not comply with JEDEC standard recommendations.

 

[Endre]

JEDEC states explicitly that for each channel, DIMM 0, which is the FURTHEST DIMM slot from the CPU for each channel (Has nothing to do with actual DIMM slot naming scheme) is, by standard, the recommended DIMM slot. That part isn't open for discussion. They create the standard and partners comply with it.

This is why:



DIMM slots which are the second and fourth, for dual channel 4 slot boards, ARE the slots furthest, at the end of the bus.

Whether other slots "work" or not probably depends on the board and the implementation. Maybe even depends on the memory itself or the strength of the IMC on a given part. It's still not the "recommended" configuration by any of the big four and does not comply with JEDEC standard recommendations.

OK.
Now, this is a very detailed answer with clear guidelines to why this is so.
In this case: YOU were right.
I didn’t know all of these, and my mobo’s manual doesn’t talk about them.
I’m not stubborn.
I’ll take your advise.
I’ll move my DIMMs from slots 1+3 to slots 2+4.
I don’t expect to see a big difference in performance, but if that’s the correct configuration, then I’ll do it!
Thanks!

PS: you live, you learn!

 

[damric]

Also learn about daisy chain and T topology if you are interested in how the dimm slots are wired. They do act differently depending on RAM placement as far as speed.

 

[Endre]

Also learn about daisy chain and T topology if you are interested in how the dimm slots are wired. They do act differently depending on RAM placement as far as speed.

OK.
I’ll search infos on that too.
Thanks.

 

[Endre]

Also learn about daisy chain and T topology if you are interested in how the dimm slots are wired. They do act differently depending on RAM placement as far as speed.

Can you, please, put a good link on this subject, for Intel Z390 chipset, or Gigabyte Z390 Aorus Master?

 

[damric]

Can you, please, put a good link on this subject, for Intel Z390 chipset, or Gigabyte Z390 Aorus Master?

I'm not familiar with that board. Email the manufacturer or search reviews and hope that it is mentioned. Generally the most expensive boards are T topology.

 

[gskill support]

OK.
Now, this is a very detailed answer with clear guidelines to why this is so.
In this case: YOU were right.
I didn’t know all of these, and my mobo’s manual doesn’t talk about them.
I’m not stubborn.
I’ll take your advise.
I’ll move my DIMMs from slots 1+3 to slots 2+4.
I don’t expect to see a big difference in performance, but if that’s the correct configuration, then I’ll do it!
Thanks!

PS: you live, you learn!

If you would like you can run some benchmark tests to see what the average performance is for both configurations, and others. The results would be interesting to see and for you to know.

 

[Endre]

If you would like you can run some benchmark tests to see what the average performance is for both configurations, and others. The results would be interesting to see and for you to know.

Alright!
So, I moved my memory DIMMs from slots 1+3 (A1+B1) to 2+4 (A2+B2).
I ran the following benchmarks:
Cinebench R15, Cinebench R20, Geekbench 4.3.3, PassMark Performance Test 9.0 Memory Mark, CPU-Z 1.9, and Windows’ rating tool: the “winsat formal” command in Command Prompt.
All of these tests were ran on both memory configurations: A1+B1 and A2+B2.

AND THE RESULT IS...
THE SAME!

Yeah, that’s right folks!
There’s 0% gain or loss from moving my DIMMs into slots 2+4!

 

[Darkbreeze]

Performance isn't the issue, or in question here. It is purely a matter of avoiding configuration errors, instability, degraded signal integrity and trouble free operation in general without random and unexplained memory errors from signal disruptions. Has nothing whatsoever to do with benchmarks, FPS or any other performance indicator. And since not every board is identical, there are going to be differences in the level to which these factors might be present or non-existent. So while it's great that you are able to run a pair of sticks in both configurations without any obvious problems while running your synthetic benchmarks, it's really not an indicator of whether or not it might be given a specific workload or extended period of sustained usage. And to be honest, nothing you ran even includes anything I would recommend running when testing memory configurations anyhow.

Even so, a sample of one, as we well know, does not ever "tell the tale" anyhow.